ACTIVE is the most consequential cognitive training trial ever run. It enrolled 2,832 healthy older adults at six US academic medical centers in 1999, randomized them to one of three short cognitive training programs or a no-contact control, and has tracked outcomes for more than 20 years. The signal that emerged in the long-term follow-up reframed the brain-training conversation: speed-of-processing training, delivered in just 10 sessions plus modest boosters, reduced the hazard of dementia by 29 percent at 10 years and by 33 percent for those who completed the booster sessions.
That finding is the strongest evidence in the cognitive-training literature, and it sits at the center of every honest discussion of whether brain training works. It is also frequently misrepresented, both by companies citing it to sell products it has no relationship with, and by skeptics dismissing it without engaging with the protocol.
This post walks through what the trial actually did, what it found, what it did not find, and how to read it.
The short answer: A 10-session program of computerized speed-and-attention training, delivered to healthy adults averaging 74 years old, was associated with 29 percent lower dementia risk a decade later. That is one trial, with one specific protocol, in one specific population. It is also the strongest signal the field has produced.
What was the ACTIVE trial actually testing?
ACTIVE was designed to answer one question. Could brief cognitive training in late life produce durable benefits in the cognitive abilities that matter for independent living: memory, reasoning, and processing speed? It was not designed to test commercial brain-training apps. It was designed to test whether the cognitive abilities themselves were trainable in a way that transferred to everyday function.
The 2,832 participants were randomized to one of four arms:
- Memory training, focused on episodic memory and verbal recall through mnemonic strategies (categorization, association, visualization).
- Reasoning training, focused on inductive reasoning and pattern recognition with novel problems.
- Speed-of-processing training, focused on visual divided attention, useful field of view, and rapid identification of stimuli at increasing eccentricity. This was the protocol that became the basis for what is now BrainHQ's "Double Decision" exercise.
- No-contact control, no training of any kind.
The initial dose was 10 sessions of about 60 to 75 minutes each, delivered in small groups over five to six weeks. Four-hour booster sessions were offered at 11 and 35 months. Total training time across the entire study was at most 30 hours.
What did the trial find at each follow-up?
The findings emerged in waves, and the headline shifted each time.
Ball et al., 2002, JAMA, year 2. All three training arms produced large and immediate gains on the trained ability, with effect sizes of roughly 1 standard deviation. There was little near-term transfer to other cognitive abilities or to self-reported daily function.
Willis et al., 2006, JAMA, year 5. The cognitive gains had decayed substantially but were still present. Reasoning training produced modest improvements in self-reported instrumental activities of daily living. Speed training produced modest improvements in everyday speed of processing tasks. No effect on dementia risk had emerged.
Rebok et al., 2014, JAGS, year 10. Cognitive gains in the trained abilities were still detectable 10 years out for reasoning and speed training (smaller for memory). Self-reported functional difficulty was lower in all three training arms versus controls. The effect on dementia or MCI had not been formally analyzed.
Edwards et al., 2017, Alzheimer's & Dementia, year 10. This was the result that changed the conversation. In a pre-specified survival analysis of the 2,802 participants with follow-up data, the speed-of-processing group had a 29 percent lower hazard of dementia compared with controls (HR 0.71, 95% CI 0.50-0.998). Those who completed the booster sessions had a 33 percent reduction. The memory and reasoning arms did not show significant dementia reductions.
"The risk of incident dementia was lower in the speed of processing training group than in the control group, but not in the memory or reasoning groups."
Edwards et al., 2017, Alzheimer's & Dementia
That dose-response relationship (more booster sessions, larger effect) is part of what gives the result its credibility. So is the fact that the speed protocol had previously been linked to better driving safety, fewer falls, and longer maintenance of instrumental daily functioning, all of which fit a coherent picture.
Why might speed-of-processing training matter at all?
Speed of processing is not really about typing fast. It is the rate at which the brain can detect, identify, and respond to visual information, particularly in the periphery and under divided-attention conditions. It is the foundational cognitive resource on which higher-order abilities depend. If processing speed declines, working memory, decision-making, and attentional control all become slower and noisier downstream.
The ACTIVE protocol trained two specific sub-abilities: divided visual attention and useful field of view. The exercises required identifying a central stimulus while simultaneously locating a peripheral target as the targets were presented faster and at wider eccentricities. The training adapts in real time to keep the user near their performance ceiling.
The mechanistic interpretation that fits best is that this kind of training rebuilds attentional control circuits that are vulnerable to age-related decline. The functional consequences (driving, fall risk, instrumental daily living) are exactly what one would predict if the underlying ability is being restored, not just the trained task improving.
What ACTIVE did not show
A few things worth being clear about.
It did not show that brain training in general prevents dementia. Two of three training arms did not produce a dementia-risk reduction. The result is specific to one protocol.
It did not show that commercial brain-training apps work. Most apps train abilities loosely related to ACTIVE's protocols, and the scientific near-vs-far transfer debate is still very much alive. The Lumosity FTC settlement in 2016 was about the gap between marketing claims and evidence; ACTIVE is not transferable evidence for unrelated products.
It did not establish causation cleanly. The 29 percent figure is a single trial's hazard ratio, and one of the field's open questions is whether this finding will replicate in independent samples. The Lancet Commission's 2024 report acknowledges cognitive training as "promising" but stops short of including it in the modifiable risk factors list, on the grounds that the evidence is not yet replicated at the level of the other 14.
It also did not test app-based delivery. ACTIVE was supervised, in-person, group-administered training. Whether equivalent effects survive when the same protocol is delivered through a phone is, technically, an open question. The available evidence suggests yes for near transfer, less clear for the dementia outcome.
What this means for what to actually do
If we take ACTIVE seriously, the practical implications are:
- Daily, brief, attention-heavy training is what the evidence supports, particularly programs that train divided attention and processing speed. This is part of why we believe in short daily sessions.
- The protocol matters more than the time spent. ACTIVE's intervention was about 10 hours plus boosters. More training is not necessarily better; the right kind of training is what produces transfer.
- Booster sessions appear to matter. Continued exposure over years, even at low doses, was associated with larger effects in the speed arm.
- Pair training with the modifiable lifestyle factors. ACTIVE-style training is a piece of the picture; exercise, sleep, hearing, and blood pressure control are the rest of it.
- Be skeptical of dementia-prevention claims from any product that is not delivering an ACTIVE-style protocol. The 2016 FTC settlement against Lumosity is the worked example of why.
ACTIVE is not the last word on brain training. It is the strongest signal in the literature, and it points in a specific direction: brief, frequent, attention-heavy, real-world-relevant training in healthy older adults. That is the bet the field is currently running.