IQ test: method

What it measures

Fluid reasoning (matrices), working memory (digit span), processing speed (symbol-digit coding), visual-spatial ability (mental rotation) and verbal ability (synonyms, analogies, odd one out). Only reasoning is placed on an IQ-type scale.

Procedure

Before starting, the player can give an age band (optional: under 16, 16-17, 18-29, 30-44, 45-59, 60 or older); without it, the run is not used for future norms. 1) Reasoning: a fixed form of 14 Open Matrices Item Bank (OMIB) items, the same for everyone, easiest first (OMIB items 5, 17, 2, 9, 1, 54, 112, 20, 91, 126, 169, 170, 214, 213), after 2 practice items (6 and 11) with solutions. The player builds the missing ninth cell from 20 elements (5 groups of 4) instead of choosing from options; each item has a 90 s limit, and the cell built so far is submitted at the limit. 2) Working memory: digit span forward (start 3, maximum 12 digits) and backward (start 2, maximum 10), digits shown on screen for 1,000 ms + 600 ms per digit, typed recall, 2 tries per length, two misses at one length end that part; no digit repeats twice in a row. 3) Processing speed: symbol-digit coding for 90 s, 9 symbols from a set of 12 with a new key every run, 6 practice items with feedback; every answer moves on. 4) Visual-spatial: same/mirror mental rotation with a fixed set of 16 pairs (the same for everyone), 2 practice pairs, 8 s per pair. 5) Verbal: 15 original items (English: 6 synonyms, 5 analogies, 4 odd-one-out; 5 options each). A few unscored trial items of new types (a generated matrix, a number series, a harder rotation, a letter series, an analogy) are placed at seeded positions, never first, and are not counted. The server rebuilds the whole run from the seed to score it.

Parameters

Scoring

Reasoning: ability theta is estimated by expected a posteriori (EAP) estimation under the two-parameter logistic IRT model with the published OMIB item parameters and a standard normal prior; the posterior SD is its standard error. Theta is moved onto the IQ scale as IQ = 100 + 15 x (mu + sigma x theta), where mu = 0.7 (SE 0.2) and sigma = 0.88 (SE 0.03) describe the OMIB calibration group in general-population SD units. These values are modelled from education statistics, not measured. The error of the IQ estimate combines the person's SE and the uncertainty of mu and sigma; the shown range is +/- 1.645 SE (90%), capped at 70 and 130. Under-16s see raw scores only. Other sections show raw scores with 'comparison coming': forward and backward span, symbol-digit correct answers in 90 s (and time per correct answer), rotation accuracy and mean RT on correct pairs, verbal number correct. No overall IQ is shown until every section has norms. A retake note uses a practice gain of about 5 points.

Valid-run rules

Every section must be finished, and the test version must match the version the run started with. Quality information is stored with the run: tab switches, fast answers (a matrix item built in under 3 s, a choice item under 1.2 s, a rotation answer under 300 ms), reasoning timeouts, whether the run was resumed, and whether it is a first attempt (browser record plus earlier runs from the same visitor id or account). A run counts towards future norms only if it is a first attempt, has no tab switches, at most one fast answer, and an age band of 16 or older. All-correct reasoning runs are flagged, because the item keys are public. Every scored run is stored with its input device class (touch, mouse, keyboard), page language, scorer version and runner-config hash. Quality flags (for example many runs from one connection within an hour, a missing visitor id, an automated browser, impossibly fast answers) and earlier attempts by the same visitor, account or connection are recorded with the row. Such runs are kept, but they are left out of any comparison with other players. Flags never change the player's own result.

Differences from the classic task

Clinical IQ tests are given in person by a trained examiner, with many more subtests and norms from a representative sample. This test is self-administered and unproctored, and only the reasoning section has a population comparison, through a modelled offset. OMIB items were calibrated without a time limit, so the 90 s limit makes them slightly harder here. Digit span is shown on screen, not read aloud. The verbal items are our own and have no norms.

Limitations

The IQ range for reasoning rests on a modelled, not measured, link between the calibration group (applicants to German medical schools) and the general population. Unproctored online scores average about a fifth of a standard deviation higher than proctored ones, and retakes gain from practice. Screen size, input device and browser timing affect the timed sections. Verbal scores must not be compared across languages. Not a clinical or official IQ test; not for school, work or medical decisions.

Languages and review status

Reasoning, working memory, speed and spatial sections are language-independent. The verbal section uses items written in each language; some adapted from the English items. Available in 29 languages: English, Arabic, Bengali, Chinese (Simplified and Traditional), Dutch, Filipino, French, German, Hebrew, Hindi, Indonesian, Italian, Japanese, Korean, Marathi, Persian, Polish, Portuguese, Russian, Serbian, Spanish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu and Vietnamese. In other languages the verbal section is skipped. Instructions are in all site languages.