What IQ tests measure, and what they don't

By hmmm.me editorial team · Published April 19, 2026

IQ is a strange number.

It is one of the most studied variables in psychology. It predicts things — academic performance, certain kinds of job performance, risk of certain medical conditions — more reliably than almost anything else on the self-report side of the field. It is also, in the hands of the internet, a number that has been weaponised, memed, taken personally, and handed out to anyone who can answer twenty matrix puzzles in five minutes.

This is an attempt to say what IQ actually is, what it isn’t, and how much any score really tells you.

A very short history

In 1905, a French psychologist named Alfred Binet was asked by the Paris public schools to develop a way to identify children who needed extra help. He built a battery of age-graded problems. A child’s “mental age” could be compared against their chronological age to flag who was struggling. It was a diagnostic tool, designed to be used carefully, with the child’s development in view.

William Stern later suggested dividing mental age by chronological age and multiplying by 100 — the intelligence quotient. Lewis Terman, at Stanford, adapted the Binet test for American use. The Stanford-Binet was born. David Wechsler later added the adult-focused WAIS, which remains the most widely used clinical IQ instrument today.

So: the original test was a careful, individually administered diagnostic tool. Everything that has happened in the hundred-and-twenty years since is a story of that tool being pulled in directions Binet would not have endorsed.

What an IQ test actually does

A modern test like the WAIS-IV samples a wide range of reasoning tasks across several domains: verbal comprehension, perceptual reasoning, working memory, processing speed. The tasks are carefully timed and scored against a standardisation sample — a large cross-section of the general population, matched for age and demographics.

Your score is not absolute. It is positional. “IQ 115” does not mean “you are 115 units of smart.” It means “compared to the age-matched population, you scored one standard deviation above the average.” The 100 is arbitrary — it is defined as the population mean, by construction. The standard deviation of 15 is also by construction.

Because the test is normed against a sample, scores get re-normed every few decades. This is because of a weird, well-documented phenomenon: raw IQ scores have been rising by about three points per decade for the last century. This is called the Flynn effect, after psychologist James Flynn, who was the one who dug through the data and pointed out that nobody had noticed it before.

Nobody is entirely sure why the Flynn effect happens. Better nutrition, more schooling, more abstract-thinking demands in daily life, smaller families, and reduced childhood disease all probably play a part. What it does mean is that a child born today is, on the raw data, scoring significantly higher on the same test than their grandparent did at the same age. And because IQ is normed to 100, the tests have to be re-calibrated periodically to keep the average where it is supposed to be.

What it predicts

IQ has been correlated with outcomes for about a century, and the picture is roughly this:

• Strongly predictive of performance on cognitively demanding tasks. School grades, standardised test scores, performance in information-heavy jobs.
• Moderately predictive of job performance in general. The effect shrinks as jobs become less cognitively demanding.
• Weakly predictive of income once you control for education. Smart kids with rich parents have an outside advantage.
• Not predictive of happiness, beyond the point where cognitive ability affects your access to stable work.
• Not predictive of wisdom, creativity beyond a modest threshold, or most interpersonal qualities.

The thing to take from this is not that IQ is all-powerful. It is that it is real, it matters, and it is also one variable among many.

What a quick online “IQ test” actually measures

A twenty-question quiz — like the one on this site — samples one or two of the reasoning types a real IQ test samples. It is not normed. Your score is calibrated against performance on these specific questions, not against a population sample.

What does a score from an online quiz tell you? Honestly, a little. It tells you roughly whether you found those puzzles easy, medium, or hard. It does not tell you where you would land on a clinical instrument. If you took the quiz twice on different days, you would get meaningfully different scores.

If you want a real measurement, you need a clinical psychologist, a couple of hours, and some money. For most people, most of the time, that is not worth doing.

On the cultural weight of the number

One last thing, because it is worth naming. IQ has a long, ugly history of being used as a tool to justify prejudice — about race, class, immigration, and who deserves what. Some of that history is still active. Careful psychologists today are very aware of it. Careless commentators are not.

If you score low on this quiz, you are not stupid. Every validated IQ measurement has wide error bars and large day-to-day variation. Twenty questions on the internet have much wider error bars still. And even a careful clinical measurement tells you about performance on specific cognitive tasks, which is a narrower thing than “how smart you are.”

If you score high, that is nice, and also not load-bearing. Cognitive ability matters, but the list of people who are smart and also unhappy, unhelpful, or stuck is long.

Take the puzzles as puzzles. Enjoy the ones you solve, be curious about the ones you did not, move on.

Take our quick puzzle set — 20 questions, about five minutes.