The question is transfer, not improvement
Nobody disputes that practising a cognitive task makes you better at that task. Do a digit span drill every day and your digit span goes up. The big claim is different. It says the gain spreads to things you did not train: memory in daily life, focus at work, protection against mental ageing. Researchers call this transfer. Near transfer means gains on very similar tasks. Far transfer means gains on truly different skills, and in daily life.
Near transfer is real but modest. Far transfer is where the commercial claims live, and where the evidence is thinnest. Almost every argument about brain training comes down to somebody presenting near transfer as if it were far transfer.
The large trials
The most striking test of scale came from a study run with the BBC. Owen and colleagues (2010) had 11,430 people do six weeks of online brain training. A control group spent the same time using the internet to answer quiz questions. The trained groups improved on the training tasks. On untrained cognitive tests, the authors reported no evidence of transfer, even to tasks that were closely related to those trained.
The biggest review is Simons and colleagues (2016), a survey of the whole brain-training field, published in Psychological Science in the Public Interest. Their conclusion: there is plenty of proof that people get better at the trained tasks. There is less proof of near transfer. And there is little good proof that training helps everyday thinking. They were blunt about study quality too. Many positive results came from small groups, weak comparisons, and tests that looked just like the training.
The control-group problem deserves its own sentence. Say your control group does nothing, while your training group logs into a colorful app every day and gets praised by it. Some of the difference you then see is hope and motivation, not brainpower. Studies with active control groups — people doing an equally engaging but different activity — consistently show smaller effects than studies with passive controls.
Where the picture is more favorable
Two things keep this from being a flat dismissal.
First, specific, narrow training with a specific target sometimes does hold up. The ACTIVE trial was a large, careful study of brain training in older adults, reported by Ball and colleagues (2002). It found that the trained skills were still better two years later. What has been argued about ever since is how far those gains extended into everyday activities, and by how much.
Second, "no far transfer" is not the same as "no value". Getting better at a memory game is a real skill. It is like getting better at an instrument: real progress, even though it does not make you smarter overall. What fails here is the marketing claim, not the activity.
And the things that do reliably support cognition are unglamorous and well evidenced: sleep, cardiovascular exercise, treating hearing loss, social engagement, managing blood pressure. They are harder to sell as an app.
Why this site measures instead of training you
This is a testing site, not a training programme, and the distinction is deliberate. We do not claim that repeating these tests will make you sharper, because we would not be able to support that claim. What practice here does do is make you better at these tests. Keep that in mind when you read your own history. A big jump over your first few runs is mostly you learning how the test works.
The same honesty governs the numbers. Where a published normative distribution matches what a test does — measured the way you actually play it — we show a percentile and link the study. Where the match is not honest, we show your score and nothing else, even when famous data exists: reaction time cites a large laboratory study as reference but carries no percentile, because browser and device delay would quietly bend the comparison. Making up a curve would be easy, and it would make the site look most trustworthy exactly where it deserves the least trust. Eight of the nine tests here have no percentile for that reason, and the methodology page says which and why.
If you want a benchmark, this is one. If you want cognitive improvement, the evidence points at your sleep schedule and your running shoes before it points at any screen, including this one.
How to read a brain training claim
A few questions that separate a real result from a press release:
- What was the control group doing? If they were doing nothing, expect the effect to shrink when compared against an active control.
- Does the outcome measure resemble the training? If the trained task and the tested task are near-twins, you are looking at near transfer being sold as far transfer.
- How many people, and were they randomly assigned? Small, loosely run studies produce big results that do not hold up.
- Did the benefit show up in daily life, or only on another test? This is the claim that matters and the one most often skipped.
- Who paid for it? Not disqualifying, but industry-funded trials of the funder's own product deserve a closer read.