Brain age test that will not invent an age

Because there is no function that could produce one honestly, and it would have to be invented. To convert a score into an age you need a published curve of that score against age, collected under conditions a browser reproduces, and with enough precision that a four-minute measurement can be located on it. This battery has one part with a real curve — vocabulary — and that curve runs upward, so a good score on it would make you older, not younger. The two parts that would make you younger have no curve a browser can use. An age printed from that mixture would be a weighted average whose weights were chosen by whoever built the page, and the visitor would have no way of knowing that the choice, not their performance, produced the number.

Usually from a formula somebody wrote down, and it is worth understanding the shape of it. The common construction is to take a score, map the observed range of scores linearly onto a plausible-looking age range — twenty at the top, seventy at the bottom — and print the result. That produces a number that reacts to performance and has no relationship to any measured population, which is why the same performance yields wildly different ages on different sites and why the ages cluster suspiciously around flattering values. A second, subtler version scores against a real published norm table and forgets to check its date, which the Flynn effect alone makes wrong by several years. The page shows that figure beside its source so you can see the size of the problem.

Because it is the counterexample the rest of the battery needs. If everything measured here declined with age, an age estimate would at least be internally coherent, and the argument against printing one would be about precision rather than about direction. Vocabulary breaks that: a large crowdsourced study puts the average twenty-year-old at about forty-two thousand lemmas and the average sixty-year-old at about forty-eight thousand, roughly a word and a half a day of net gain over four decades. Any battery containing both a speeded task and a word task is measuring two things that disagree about which way age goes, and putting them in the same page makes the disagreement visible instead of hiding it inside a single number.

Because remembering a sequence you were shown seconds ago is a different question from what you can do with material in front of you, and this page keeps to the second. That boundary is also why part two is built so that every premise stays visible while you answer: a reasoning item that requires you to hold three facts in your head returns a score that moves with working memory, and a page reporting that as reasoning has mislabelled it. If a memory measurement is what you are after, it is not that this page thinks it unimportant — it is that a good one needs its own design, its own scoring and its own page.

Because the thing an attention block would be claiming to measure takes longer than this whole battery to appear. Sustained-attention performance does not fall in the first minute or two; the decline classical studies describe emerges over tens of minutes of watch-keeping, so a browser test that ends in ninety seconds has stopped before its own phenomenon starts, and any figure it printed would be about the first two minutes of a task rather than about attention. The eight-second attention span figure that circulates widely is worse than imprecise: it has no primary source at all, and the citation chain behind it terminates in nothing.

Not on the basis of this page, and the reason is that a single measurement has no baseline to be worse than. Detecting a change in cognition requires two measurements of the same person on the same measure, separated in time, or a clinician with instruments and your history — the whole difficulty of the field is that the between-person spread is far larger than the change that matters within one person. A poor score here is also produced by a small screen, an unfamiliar keyboard, a bad night, a headache, or having answered part one while thinking about part three. Where a real change has occurred, the thing worth acting on is what you observed about yourself, and the place for it is a consulting room.

Only part one, and only weakly. The reasoning and word items are the same sixteen and eight on every run — the order shuffles, the items do not — so a second attempt at those partly measures what you learned from the answers the results panel gave you, which is practice rather than ability. Part one regenerates its digit strings from a fresh seed every run and has no content to remember, so it is the part where two runs are comparable. Even there, a difference between two runs is only interesting if it is larger than the spread within each of them, which is why the page prints intervals rather than bare scores.