Every claim a test makes rests on two things: where its questions come from, and how it turns answers into a number. Here is both, in plain terms. For the wider background — what IQ tests measure at all, and how the field got here — read how IQ tests work.
The questions
The test draws on the International Cognitive Ability Resource (ICAR), an open item pool developed by David Condon and William Revelle at Northwestern University and documented in their 2014 paper in Intelligence (vol. 43, pp. 52–64). ICAR exists precisely so that researchers — and sites like this one — don’t have to invent reasoning items from scratch. Its problems have been administered to tens of thousands of respondents and correlated against full-length batteries, the kind of validation a fresh question bank cannot claim.
We use four item families, chosen to sample different parts of reasoning rather than one narrow skill:
- Matrix reasoning. Raven’s-style grids where you infer the missing figure. These load most heavily on fluid intelligence — solving a novel problem with no prior knowledge — which is why this section has the most items of the four.
- Verbal reasoning. Analogies and logic expressed in words, leaning on crystallized ability: the vocabulary and knowledge you build over a lifetime.
- Numerical and letter series. Sequences where the task is to find the rule — a clean read on pattern detection and working memory.
- Three-dimensional rotation. Spatial items that ask you to track an object as it turns in space.
The spread is deliberate. Back in 1904, Charles Spearman noticed that performance across unrelated mental tasks tends to move together — the positive correlation he named g, the general factor. A score built on a single task type is easy to game and easy to mislead; four families produce a profile instead of a party trick.
How the score is calculated
Your raw result is the number of correct answers. Because the matrix section has the most items, it naturally carries the largest influence on that total. We then place that raw score on the familiar deviation-IQ scale, where the population average is fixed at 100 and one standard deviation is 15 points — the convention David Wechsler introduced and the one almost every modern battery still uses.
On that scale, 115 sits one standard deviation above the mean, and roughly 84% of people fall below it; 130 is two deviations up, near the 98th percentile. From the same normal curve we read off your percentile and assign one of six plain-language bands; the full band-by-band breakdown is in the IQ score chart. If you want to see the whole thing — from finishing the test to the result — in action, you can simply take it.
What the number is — and isn’t
Treat the result as an estimate with a margin of error, not a verdict. A short, unproctored test has a wider confidence interval than a two-hour supervised one. A fair rule of thumb is that your “true” score sits within about ten points either side of what you see. Sleep, caffeine, screen size, and whether you actually read each question all nudge the figure.
Scores also drift upward across generations — the Flynn effect, named after James Flynn, runs to roughly three points per decade. That is why norms need refreshing, and why a number with no date or context is close to meaningless. We don’t pretend the test measures everything either: it says nothing about creativity, emotional intelligence, conscientiousness, or judgement, and a high score has never once made anyone wise.
For anything consequential — a clinical question, a school placement — a licensed psychologist running a WAIS-IV or Stanford–Binet is the right call, and no online test substitutes for one. We repeat this in our terms of use because it genuinely matters.
For the human side of why we work this way, see about. If you care what happens to your answers, the privacy policy and cookie policy spell it out. Spotted an error in an item or a claim here? Tell us through contact — corrections are welcome.