On January 6, 1995, a man in Pittsburgh walked into two banks and robbed them at gunpoint, wearing no mask, smiling at the security cameras. His name was McArthur Wheeler, and he was certain the cameras couldn’t see him: he had covered his face in lemon juice. Lemon juice is invisible ink, he reasoned, so it would make him invisible too. The police arrested him the same evening, and he stared at them in genuine disbelief. “But I wore the juice.”
🧠 The Robbery That Became a Study
The news reached David Dunning, a psychology professor at Cornell, and it planted a strange question: if Wheeler was too incompetent to be a bank robber, was he also too incompetent to know that he was incompetent? In 1999, Dunning and his student Justin Kruger tested the idea with logic puzzles, grammar tests, and humor ratings. Participants took the tests and then guessed how well they’d done. The result became one of psychology’s most quoted curves: the bottom quarter of performers wildly overestimated themselves, placing their skills around the 60th percentile when they actually sat near the 12th. The top quarter, by contrast, slightly underestimated themselves. The paper’s title said it all — Unskilled and Unaware of It — and the explanation was elegant: to know you’re bad at something, you need the very skill you’re bad at. The incompetent carry a double burden: they lack the ability, and they lack the ability to see that they lack it.
🤔 The Effect That Explains Itself
Here’s the twist that never makes it into the memes. The popular version — “stupid people are too stupid to know they’re stupid” — is a misreading on two counts. The effect is specific to individual skills, not general intelligence, and the worst performers are still less confident than the best ones; they’re just wrong by a much larger margin. And then there’s the deeper twist: the effect might not be a psychological mechanism at all. Test scores contain random noise, and when you sort people by those scores, the unlucky bottom performers naturally look like they overestimated themselves — a statistical illusion called regression toward the mean. In July 2026, a paper in Psychological Review by researchers at the University of Bath and LSE applied this correction to massive replication datasets and got an inversion: overconfidence turned out to be universal, and the most capable people were the most overconfident of all. The authors’ explanation isn’t ignorance — it’s signaling. High performers have the strongest incentive to project confidence, because confidence is how you broadcast competence you can’t directly display. The effect that taught us to mock the incompetent may have been, all along, a story about how the rest of us misunderstand statistics.
🔗 The Valley of Self-Doubt
If you’ve ever written anything, you know the curve personally. Beginners are radiantly confident; the drafts feel great. Then skill grows, taste sharpens, and suddenly every page looks terrible — the famous valley of self-doubt that makes writers quit right before they’d have gotten good. That dip isn’t failure; it’s the Dunning-Kruger curve working as advertised, the moment your ability to see the gap finally outpaces your ability to close it. It also explains something subtler about the self-knowledge industry: people who know themselves least are often the most certain they know themselves best. That confidence is what makes a personality reading feel so startlingly accurate — the reader arrives already sure of their own story, ready to nod at anything that fits it. And for anyone who works with surveys: self-assessments are the least reliable data you will ever collect. People are terrible at rating their own skill, which is precisely why the ratings are so interesting.
🎲 Ninety-Three Percent of Us
The bonus is a gift that keeps giving. Kruger and Dunning won the 2000 Ig Nobel Prize — the satirical award for improbable research — cited for “their modest report.” But the better party trick comes from a 1981 study by Ola Svenson: asked to rate their driving, 93% of American drivers said they were above average. Mathematically, at most half of them can be. The next time you’re in a car, try it — or ask a friend how they rank at anything they love. The odds are excellent they’ll say top half. And somewhere in Pittsburgh, a man who robbed two banks in a lemon-juice mask is still the purest example science ever found of the gap between what we believe about ourselves and what is true — even if, in the end, the gap turned out to be partly a trick of the numbers.