An IQ test that shows you how the number is built

Each item in the bank carries two estimated properties: how difficult it is, and how well it discriminates — that is, how sharply it separates higher from lower ability. Getting a hard, highly discriminating item right tells the scoring model more than getting an easy one right.

The model behind this is a two-parameter logistic item response model. After your answers are in, the scorer finds the ability level that best explains the pattern of correct and incorrect responses — not the level implied by your raw total. Two people with the same number correct can receive different ranges if they solved different items.

Sixteen items produce a useful estimate but a wide one, which is why the short form reports a 95% range such as 98–112 rather than a single figure. The width is arithmetic: fewer items, less information, wider interval. The full scoring specification is on the methodology page.

An online estimate is not a supervised, professionally administered assessment and should never be used for clinical, educational or employment decisions. Test conditions — sleep, interruptions, screen size, familiarity with item formats — all move scores, and retaking the same items inflates results through practice rather than ability.

Treat your result as a well-reasoned estimate with stated uncertainty. The accuracy and limits page covers these constraints in full.