Our IQ Assessment Methodology

At a glance

Where the test comes from

The test began as a paper-and-pencil instrument created by Peter Janicki, MSc Medical Psychology, during his university studies, and was later rebuilt as the online test on this site. Everything below describes the online version as it runs today; its norms come only from online sittings.

The model behind the test

The test is built on the Cattell-Horn-Carroll (CHC) theory of cognitive abilities, the most widely supported framework in modern intelligence research. CHC describes ability as a hierarchy: a general factor (g) at the top, supported by broad domains that are correlated with one another but measurably distinct. We score six of those domains: fluid reasoning, verbal reasoning, visual-spatial reasoning, quantitative reasoning, working memory, and processing speed.

Form length

There are two forms. The free short form is 12 items; the full form is 60 items. The median completed sitting took 2.8 minutes, with both forms counted together. The short form samples every domain but cannot measure any of them precisely, which is why its result is reported with a much wider confidence band than the full form.

How the items are chosen

The test is not adaptive: all items of a sitting are drawn when it starts, and which items you get never depends on your answers. Within each domain the draw takes two items at each of five difficulty levels, at random, avoiding items you saw in a recent sitting where the bank allows. The bank holds 60 approved items, as many as the full form asks for, so today every full-form sitting answers the same items. The short form is the first part of that draw: per domain, one easier and one harder item, each picked at random from the two at its level, so short-form sittings answer different items.

That is also why no reliability coefficient (Cronbach's α) is published. On the short form, sittings answer different items, so there is no single item set to compute it on; on the full form it could be computed and has not been yet.

How the overall estimate is combined

Domain scores are not averaged equally. Each contributes according to our own weights for how strongly that domain relates to general cognitive ability (informed by CHC theory, not copied from a test manual), so fluid and verbal reasoning weigh more heavily than processing speed rather than treating six unlike things as interchangeable.

What your score is measured against

Your result is an estimate of where you stand in our own sample of test-takers. We work out your position against everyone who has completed the same form, then show it two ways: as a percentile, and converted to the familiar mean-100, SD-15 scale as an estimated IQ, always with its range. The conversion makes the number readable. It does not make it a clinical IQ.

We want to be exact about the difference, because many free tests are not. A clinical IQ such as a Wechsler (WAIS) score means something because the test was standardised against a representative normative sample — a demographically controlled group recruited specifically for that purpose, which is slow and expensive work and is what you are paying for when a clinical instrument is administered. We have not done that. So our estimate is labelled as an estimate, is never shown without its range, and describes where you stand among the people who have taken this test, not in the population.

Why the result is a range

Every result is reported with a range rather than as a single value. The range is a fixed, deliberately conservative width: wider on the twelve-item short form than on the 60-item full form. It is not a measured standard error. If two ranges overlap, treat the difference between them as uncertain. Read the range, not its midpoint.

No time limit

The test on this site is untimed and scored on accuracy alone. That has one consequence worth stating: processing speed is, by definition, about working quickly, so without a clock its items measure accuracy on simple tasks rather than speed. It is also the domain with the smallest weight in the overall result, so the effect on your composite is small. Response times are still recorded, for one purpose — see the low-confidence flag below.

The numbers, stated exactly

The weights are our own choices of how much each broad ability should count toward general ability, informed by CHC theory; they are not taken from any test manual. On the full form the overall result is a rank: the weighted mean of your six domain accuracies is placed on the distribution of that same composite among everyone who has completed the full form (until enough sittings exist for that, it falls back to the weighted average of the six domain percentiles). On the short form the overall result is the weighted average of the six domain percentiles; an average of percentiles is not itself a percentile rank and pulls results slightly toward the middle, a known simplification we have left as it is.

• Comparison group. Each domain's percentile is computed against everyone who has completed the same form — here or on My Path, where the same test runs — excluding our own internal test runs. As of 16 October 2026 that is 96 completed sittings of the full form and 288 of the short form, which is why we call the sample small and expect percentiles to move as it grows. The norms are recalculated nightly.
• The range. The overall result is shown ±12 percentile points on the short form and ±7 on the full form, capped at the 1st and 99th. These are deliberately conservative fixed widths. They will be replaced by a measured standard error once the sample supports one.
• Short form composition. Two items per domain, one easier and one harder. Short-form norms count only sittings since 2026-09-10, when that composition was introduced, so results from the older, easier version do not mix in.
• Low-confidence flag. A result is marked low-confidence when too few items were answered or when answers came implausibly fast — judged both from per-item times and from the time our server observed between start and submit, which a browser cannot alter.

Item quality and bias

Items are reviewed for clarity and for any ambiguity in the intended answer. Per-item statistics (such as difficulty) need far more responses per item than we have so far, so they are not yet used. Pattern-based and numerical items carry less cultural and linguistic loading than verbal items; verbal items necessarily depend on how fluent you are in English. No cognitive test is culture-free, and we would distrust one that claimed to be.

Limitations — please read this part

This is a self-administered online screener for self-insight and curiosity. It is not a clinical or diagnostic assessment, and it must not be used for employment selection, educational placement, admissions, or any other decision that affects a person’s opportunities. Those uses require a professionally administered instrument, and in several jurisdictions using a test like this one for hiring carries real legal exposure.

Results also move for reasons that have nothing to do with ability: sleep, stress, illness, interruptions, the device you used, and simple familiarity if you have taken the test before. A single sitting is a snapshot taken under one set of conditions. If you need a formal evaluation — for a suspected learning difference, a giftedness assessment, or any clinical question — consult a licensed psychologist.