The Dunning-Kruger effect. · published 2 September 2026
The 1999 study by Kruger and Dunning is real and reported that bottom-quarter scorers overestimated their rank while top scorers underrated theirs — but the famous rising-and-falling confidence curve was drawn later for the internet, and critics from Krueger and Mueller in 2002 to Gignac and Zajenkowski in 2020 argue the pattern is largely regression to the mean, a dispute Dunning still contests, leaving one durable takeaway: people misjudge their own competence in both directions.
These are the working notes, unedited. Everything the video
states as fact comes from here, and nothing else may be stated as fact.
The sections on limits and on what the finding does not show are
part of the brief, not an afterthought — see
the method.
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).
What the paper reported
Across tests of humour, logical reasoning and grammar, participants who scored in the bottom quarter substantially overestimated their own percentile rank, while the highest scorers slightly underestimated theirs.
The popular version is that incompetent people are too incompetent to notice, usually stated as the less someone knows, the more certain they are. It is almost always illustrated with a curve rising to a peak of confidence, often labelled Mount Stupid, falling into a valley of despair and climbing again.
That curve appears nowhere in the 1999 paper. There is no such graph in the research. It was drawn later for the internet, and it is not a finding.
The stronger objection is statistical. Because self-assessments are noisy and imperfectly related to actual scores, comparing the two produces this pattern even when nobody is systematically deluded: the lowest scorers have nowhere to go but up and the highest nowhere to go but down. That is regression to the mean, combined with most people rating themselves a little above average. Krueger and Mueller raised this in 2002. Nuhfer and colleagues (2016, 2017), using measures designed to avoid the problem, did not find the classic pattern. Gignac and Zajenkowski argued in Intelligence (2020) that the effect is mostly a statistical artefact. A standard demonstration is that randomly generated data, containing no psychology at all, reproduces the familiar chart.
Dunning has responded that the phenomenon survives these criticisms, and the argument is not settled. It should be described as a live dispute, not as a finding that has been withdrawn.
What is not in dispute is that people are poor judges of their own competence in both directions, and that experts frequently underrate themselves relative to others.