How our IQ test works

Every cell of a matrix has 4 features: shape, number of shapes, shading and direction. Each feature follows one of 4 rules across the grid: it stays the same everywhere, it changes once per row, it changes once per column, or it forms a Latin square where every row and every column holds each value once. The missing cell is the only one that satisfies all 4 rules.

Generating matrix problems from explicit rules is an established method in cognitive testing research, because it makes difficulty predictable: more changing features and more Latin-square rules make a harder item. Our level 1 items change one feature, our level 5 items change three, all as Latin squares.

Only triangles ever change direction. A square or a hexagon turned by 90 degrees looks the same, which would make two answer options identical.

A common way to write wrong options is to take the right answer and change one feature. That leaves a shortcut: the right answer is the option that shares the most features with the others, and people can find it without looking at the grid.

Our 8 options are every combination of 3 features, each set to the correct value or one wrong value. Each value then appears in exactly 4 of the 8 options, so counting features points nowhere. Where the grid shows another value for that feature, the wrong option uses it, so it looks plausible. A unit test checks the balance, the 8 distinct options and the single correct one on every item before each release.

Each series follows one written rule, from adding 4 at level 1 to adding the square numbers at level 5. The 5 wrong options are the results of plausible slips: applying the rule one step too far, using the wrong operation, or reading only the last two numbers.

Every person sees the same items in the same order, which is what makes two scores comparable. A solved item is worth 1 point, a wrong answer or a skipped item is worth 0, and items left when the 20 minutes run out count as unsolved. The result shows the total and the score at each level.

Your answers stay in your browser while you take the test. The result is encoded in the result link, so we never receive your answers.

An IQ score says how your raw score compares with a norm group, a large sample of people who took the same version. A score of 115 means you did better than about 84% of that group. Without a norm group a raw score cannot be placed on the IQ scale, so every IQ figure we show comes from a norm table built from people who took this exact version, and this page names the sample: who, how many and when.

• It measures fluid reasoning on two formats. Clinical tests add verbal, memory and processing-speed tasks, see types of IQ tests.
• Conditions are uncontrolled: distraction, a small screen or a second attempt all change a score.
• It is not a diagnostic or clinical assessment, and no school, employer or high-IQ society accepts it. Official IQ tests lists the ones that count.