Every celebrity IQ score you have read is made up

Cognitive testing

Type almost any famous name into a search box, add the letters IQ, and a number comes back. Precise-looking, too: 155, 152, 160. Then try chasing one of them back toward a source. You run out of pages first.

Search as long as you like. There is no verifiable published IQ score for Elon Musk, and none for Mark Zuckerberg, Barack Obama, Jim Carrey or Lisa Kudrow. Those scores do not exist at any value.

Where the numbers come from

Someone invented a good share of them. A writer typed a figure into a listicle a decade ago, the next listicle borrowed it, the one after that borrowed it again, and round it went. Repeat that enough times and the number collects citations. Follow one. You land on another site carrying the same number and naming nobody.

Other writers back-calculate a figure from something unrelated. An old SAT score, a chess rating, a university place, an anecdote about someone's childhood. Conversion tables that turn test scores into IQ estimates circulate online, and the people reaching for them sound a good deal surer than the tables deserve. Reach back a few decades and the ground has shifted anyway: publishers rescored the tests, and the populations changed underneath them.

And some pin a number on the wrong person entirely, or on a character that person once played.

Repeat a number a thousand times and you have still measured nothing. What you have is a rumour with a decimal point in it.

Why nobody publishes their real one

A qualified psychologist sits down with you, one to one, and administers an IQ assessment. What comes out of it counts as health information. You would not toss that into a magazine interview, and a psychologist who intends to keep a licence will not hand a client's score to anyone who asks.

That leaves a figure attached to a living person with two possible origins, neither of them flattering. Either the person said it about themselves, which makes it unverified self-report, or somebody invented it.

What the number would mean if it existed

The listicles skip this part. An IQ score is a position in a distribution. It tells you how far you sit from the average of the group the publisher standardised the test on, measured in standard deviations and then rescaled so that the average reads 100. It counts out no units of intelligence.

Everything else follows from that. A score of 130 does not hand you thirty units of anything; it sits two standard deviations above the mean, around the 98th percentile. And because the population defines the scale, the population matters. Scores drifted upward across the twentieth century by roughly three points a decade, so psychologists keep restandardising the tests. Set a score from 1975 next to one from today and you are not comparing like with like.

Quote a score without its scale and you have told the reader nothing. Some tests use a standard deviation of 15, others 16, others 24. One person, one level of performance, and the figure that comes out reads 132, 134 or 151 depending on which test they sat. The internet quotes all three as though they were the same claim.

The signs-of-high-intelligence genre

The other half of that search results page is lists. Seven traits, ten signs, fifteen strange indicators. You cannot measure most of the items on them and you cannot disprove them either; the rest flatter you, which is what the page is selling. "You overthink things" and "you talk to yourself" fit most of the people reading them, which is rather the point.

A few things do correlate with measured ability, modestly and on average: reading a lot, vocabulary size, working memory capacity, reaction time. Those two qualifiers carry most of the weight in that sentence. None of the four picks out an individual. A correlation drawn from thousands of people tells you almost nothing about the one person sitting in front of you.

What we do here instead

Our reasoning test reports a range instead of a single figure, and the result page says plainly that you cannot turn that range into a percentile. Arithmetic forces our hand there. A percentile needs a normative sample, we hold no normative sample, and a figure shaped like an IQ score would invite the exact conversion this article has spent its length complaining about.

We can tell you how you do on twenty generated reasoning puzzles. That is the whole offer. For a real IQ assessment you book a psychologist and keep the appointment, and most people who go looking for the number want something else entirely, which the number was never going to supply.