Assessment disclaimer

Lernio is not a clinical or diagnostic assessment, and it is not a medical device. It is a short, self-administered, unsupervised test taken on your own device. It produces an estimate of cognitive performance for training purposes. It is not a substitute for an assessment administered by a qualified professional, and it must not be used for any educational, medical, employment, immigration, forensic or other decision that calls for one.

Five abilities are scored separately — working memory, logical reasoning, pattern recognition, processing speed and spatial reasoning — from questions that are timed, because how quickly you answer is part of what is being measured rather than an afterthought. A category is only reported when enough questions were answered to say anything about it; below that threshold it is flagged as unreliable rather than shown as a number.

Responses are scored with an item-response model, not by counting correct answers. Each question carries its own difficulty and discrimination, so a hard question answered correctly counts for more than an easy one, and two people with the same number of correct answers can legitimately receive different estimates.

The model produces an ability estimate together with a standard error. The estimate is placed on a conventional scale with a mean of 100 and a standard deviation of 15, and the standard error becomes the reported range. Displayed values are held within defensible bounds: an estimate far outside the range a short test can support is not credible, so it is not printed as though it were.

A short assessment cannot pin a number down exactly. Reporting a single figure would imply a precision that does not exist. Every estimate is therefore reported with the interval it could plausibly fall in, and the interval is the result — the midpoint is not more true than its ends. Where the evidence is especially thin, the report says so instead of quietly narrowing the claim.

The percentile compares your ability estimate against a modelled normal distribution, not against a measured sample of people who have taken this test. It answers “where does this estimate sit on the assumed scale”, not “how many real people did you beat”.

[Norming basis — the reference population the scale is anchored to, and whether a collected sample will replace the modelled distribution before launch].

The conditions of an unsupervised test are not controlled, and they matter. Tiredness, interruptions, a small screen, a slow connection, a touchscreen instead of a keyboard, illness, medication, stress, and simply how much attention you gave it can all move a result — sometimes by more than a genuine change in ability would. A single sitting is a measurement of one sitting.

Retaking soon after a previous attempt mostly measures how well you remember the questions, which is why a retake is only offered after a cooldown.

Training improves performance on the tasks you train, and that progress is tracked separately as a Cognitive Performance Score. We do not claim that playing games raises a validated measure of intelligence. Your baseline estimate does not move because you trained — only a new assessment can change it, and the two numbers are kept apart on purpose so that improvement in one is never presented as improvement in the other.

Timed questions and visual patterns can disadvantage people whose vision, motor control or attention differ from what the format assumes. That is a limitation of the format, not a finding about the person, and a result affected by it should be read with that in mind. [Accommodations available, and how to request one].

Tell us. Contact [support email address] with the date of your assessment. We can look at how the score was produced, and we can re-score an assessment when the scoring model itself is corrected — every result records the version of the model that produced it, so this is traceable rather than a matter of opinion.

This disclaimer forms part of the terms of service.