What the 1999 Dunning-Kruger Paper Found, and What Critics Say

Key Points

  • The 1999 Kruger and Dunning paper found that, across four experiments, people in the bottom quartile on tests of humor, grammar and logic estimated themselves at the 62nd percentile while their scores put them at the 12th.
  • A 2002 paper by Krueger and Mueller argued that statistical regression and a better-than-average tendency explain the pattern, and Kruger and Dunning replied that their experimental data do not fit that account.
  • David Dunning said in 2024 that the effect is not about general stupidity and applies to everyone in the areas where they lack expertise.
  • Later papers disagree: one derives the pattern as a statistical artifact, while a large replication supports weaker self-monitoring among low performers in grammar and logic.

Published .

A four-story rust-brown steel and glass building seen from a corner, with trees in leaf in front, a sidewalk and crosswalk, and a blue sky with white clouds.
Uris Hall, from Feeney Way, Cornell University, Ithaca, New York. Photo: P. Hughes, CC BY 4.0, via Wikimedia Commons, resized, photo source.

In 1999, Justin Kruger and David Dunning of Cornell University reported that people who scored in the bottom quartile (the lowest scoring quarter) on tests of humor, grammar and logic grossly overestimated their performance. Across four experiments, their test scores put them in the 12th percentile, but they estimated themselves to be in the 62nd1. The pattern became known as the Dunning-Kruger effect. The debate that followed has covered both why it happens and how large it is, as the sections below show.

A bank robber and a lemon

The paper opens with McArthur Wheeler, who in 1995 robbed two Pittsburgh banks in daylight with no visible disguise. When police showed him the surveillance tapes, he said, "But I wore the juice," apparently believing that lemon juice on his face would hide it from video cameras2. The authors used the story to make a point about a dual burden. People who are unskilled reach wrong conclusions, and the same lack of skill keeps them from seeing their mistakes2.

What the studies did

Study 1 asked 65 Cornell undergraduates to rate jokes, and compared their ratings with those of professional comedians. On average the students placed their ability to recognize what is funny in the 66th percentile, 16 points above the actual average2. Those in the bottom quartile overestimated their percentile ranking by 46 points, while those in the top quartile slightly underestimated their ability2.

Study 2 gave 45 Cornell undergraduates a 20-item logical reasoning test, and Study 3 used a grammar test with 84 participants2. In Study 4, the authors trained participants in logical reasoning, and the training helped those who had scored low recognize the limits of their ability2. The authors linked the gap between estimates and scores to weak metacognitive skill, meaning the capacity to distinguish accuracy from error1.

What the effect is not

In a 2024 interview, Dunning said the effect is not about general stupidity. It applies to everyone in the areas where they lack expertise, he said, because the skill needed to spot a mistake is the same skill they are missing8.

The debate over why

The 1999 paper itself acknowledged that overestimation by low scorers could be taken as a mathematical certainty, since someone with a low score has more room to overestimate than to underestimate. It argued that if regression alone were responsible, the top quartile would miss by a comparable amount, and the data did not look like that2.

In 2002, Joachim Krueger and Ross Mueller reported that low scorers overestimate and high scorers slightly underestimate, but they found no evidence for the metacognitive explanation. They argued that statistical regression and a general better-than-average tendency explain the pattern, and that errors were no longer asymmetric when either was statistically removed3. Kruger and Dunning replied that Krueger and Mueller had replicated the basic findings but interpreted them differently, and that Krueger and Mueller used unreliable tests and inappropriate measures. They also argued that a regression account cannot explain their experimental data4.

In the same interview, Dunning said critics focus on the first two of the four studies in the 1999 paper, and that his group ran a nine-study series investigating regression to the mean8. In 2002, both sides agreed that low scorers overestimated, and the dispute was over why4.

Later work

In 2022, Jan Magnus and Anatoly Peresetsky offered a statistical explanation that, they said, does not require any psychological explanation, deriving the effect as a statistical artifact from a model that takes boundary constraints into account. They reported that the model fits the data almost perfectly5.

Curtis Dunkel, Joseph Nedelec and Dimitri van der Linden described an earlier paper in the journal Intelligence as finding no significant effect for self-rated intelligence once better suited analyses were used. Their own 2023 analysis of a more representative sample found a statistically significant effect, but a very small one, which they said brings its meaningfulness into question6.

In 2021, Robert Jansen, Anna Rafferty and Thomas Griffiths built a rational model showing that the pattern could be produced by prior beliefs about ability, by a relation between performance and skill at judging correctness, or both. They replicated the 1999 work with approximately 4,000 participants in each of two studies, and comparing two versions of their model supported low performers being less able to estimate whether they are correct in the domains of grammar and logical reasoning7. Read together, the three papers above point in different directions: some treat much of the pattern as a statistical artifact, and others find evidence for weaker self-monitoring among low performers in some domains.

Trivia

  1. At which university did Kruger and Dunning work when the 1999 paper was published? Cornell University1.
  2. At what percentile did the bottom quartile estimate themselves, on average? The 62nd1.
  3. Which bank robber opens the 1999 paper? McArthur Wheeler2.
  4. How many undergraduates took part in the humor study? 652.
  5. Which two researchers argued in 2002 that regression and a better-than-average tendency explain the pattern? Joachim Krueger and Ross Mueller3.

Sources

  1. Justin Kruger and David Dunning, Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments, Journal of Personality and Social Psychology, 1999 (abstract record). Accessed October 25, 2026.
  2. Justin Kruger and David Dunning, Unskilled and Unaware of It, Journal of Personality and Social Psychology, 77(6), 1121-1134, 1999 (full text copy). Accessed October 25, 2026.
  3. Joachim Krueger and Ross A. Mueller, Unskilled, unaware, or both? The better-than-average heuristic and statistical regression predict errors in estimates of own performance, Journal of Personality and Social Psychology, 2002. Accessed October 25, 2026.
  4. Justin Kruger and David Dunning, Unskilled and unaware, but why? A reply to Krueger and Mueller, Journal of Personality and Social Psychology, 2002. Accessed October 25, 2026.
  5. Jan R. Magnus and Anatoly A. Peresetsky, A Statistical Explanation of the Dunning-Kruger Effect, Frontiers in Psychology, 2022. Accessed October 25, 2026.
  6. Curtis S. Dunkel, Joseph Nedelec and Dimitri van der Linden, Reevaluating the Dunning-Kruger effect: A response to and replication of Gignac and Zajenkowski (2020), Intelligence, 96, 2023. Accessed October 25, 2026.
  7. Robert A. Jansen, Anna N. Rafferty and Thomas L. Griffiths, A rational model of the Dunning-Kruger effect supports insensitivity to evidence in low performers, Nature Human Behaviour, 2021. Accessed October 25, 2026.
  8. Interview with David Dunning, of Dunning-Kruger Fame, Open Mind interview by Corey S. Powell, April 5, 2024, reposted by The Conversable Economist, April 17, 2024. Accessed October 25, 2026.