The Dunning-Kruger Paper Was About Quartiles, Not Confidence

In December 1999, Justin Kruger and David Dunning, both at Cornell, published 'Unskilled and Unaware of It' in the Journal of Personality and Social Psychology. The premise had come from a Pittsburgh bank robbery a few years earlier: McArthur Wheeler, who had rubbed lemon juice on his face under the impression it would render him invisible to surveillance cameras, was caught the same day. Dunning, struck by the article, wondered whether incompetence systematically masks itself.

The study ran four tasks: humor (rating jokes), grammar, logical reasoning, and a self-test of debating-style argumentation. Participants — Cornell undergraduates — rated their own performance against peers afterward. The headline figure: people scoring in the bottom quartile estimated their performance was, on average, near the 62nd percentile. They were not strangers to the topic. They were, according to the authors' interpretation, missing the very skills required to evaluate their own answers.

The top quartile showed a mirror error in the other direction. They underestimated themselves, mostly because they assumed their peers were as competent as they were. Adjusting their self-assessment after seeing other students' answers corrected the underestimate quickly. The bottom quartile, given the same exposure, did not improve.

The paper has since been simplified into a meme — the curve where confidence drops off a cliff at 'Mt. Stupid.' That curve does not appear in the original. The actual data show roughly straight lines: self-assessment rises gently with skill while real performance rises steeply, and the gap closes only at the top. Statisticians have argued for years over how much of the effect is real psychology and how much is regression to the mean. The paper's place in the canon survives both critiques. Most of its citations get the headline wrong.