Is the Dunning-Kruger effect real? Partly. People who score lowest on a test do tend to rate themselves too highly, but much of the famous pattern is a statistical illusion, and new research suggests that once the numbers are handled more carefully, the most able people are the most overconfident, not the least.
The idea comes from a 1999 paper by Justin Kruger and David Dunning, and it has become internet shorthand for the confident person who does not know what they are talking about. That shorthand is unkind and not very accurate.
Overconfidence itself is not in doubt. It turns up everywhere, from the way overconfident voters can make fake news more damaging to the colleague who is sure a project will take a week. The real question is who is most overconfident.
What did Kruger and Dunning actually find?
Kruger and Dunning ran four studies using tests of humour, grammar, and logic, published in the Journal of Personality and Social Psychology. In their 1999 paper on unskilled and unaware self-assessment, people who scored in the bottom quarter overestimated how well they had done. In the logic study, the lowest scorers averaged the 12th percentile but guessed that their score sat at the 62nd.
People at the top tended to underestimate themselves a little. The authors argued that the skills needed to do well are also the skills needed to spot mistakes. The logic study had 45 students, so its bottom quarter was just 11 people.
Why some researchers call it a statistical artefact
The main criticism is that the pattern can appear even when nobody lacks self-insight. Test scores contain luck, so the lowest group includes people who had a bad day or hit unlucky questions, and the highest group includes people who had a good one. Their guesses reflect what they can usually do, so they look too high at the bottom and too low at the top, a quirk known as regression to the mean.
Several studies back this up. Gilles Gignac and Marcin Zajenkowski tested 929 adults on a reasoning test and a self-rating of intelligence, and found an essentially straight-line link, with no sign that low scorers misjudged themselves more. They concluded that the effect may be much smaller than first reported. A statistical model by Magnus and Peresetsky showed that statistics alone can produce the classic pattern.
Not everyone thinks the effect vanishes. In a 2023 replication by Curtis Dunkel and colleagues using a large, nationally representative sample, the effect was statistically significant but minimal, and the authors questioned its practical meaning.
What the 2026 study found
In a paper published online on 20th July 2026 in Psychological Review, Chris Dawson at the University of Bath and David de Meza at the London School of Economics set out an open access study of overconfidence and loss aversion. Its main aim is to explain why overconfidence might have evolved, but the authors also tested their ideas on data from an earlier study of about 4,000 online participants, who answered 20 grammar and 20 logic questions and then guessed how many they had got right.
In the grammar test, the average score was 10.17 out of 20 and the average guess was 12.49. Once the authors allowed for random error in both scores and guesses, overconfidence showed up at every level of performance and grew with it. As Dawson said in the University of Bath news release, overconfidence “is a universal human trait, but it is the most capable among us who exhibit it the most”.
A note of caution: this is one paper using one dataset, and it is new, so other teams have had little time to test it. The authors also say poor performers may sometimes lack the insight to recognise their own mistakes.
So is the Dunning-Kruger effect real?
The fairest answer is that a smaller version is real and the dramatic version is not. People do misjudge their abilities, and their guesses track their true skill only loosely: in the large 2023 replication, the correlation between measured and self-estimated intelligence was 0.24. But the meme version, in which the least skilled are uniquely clueless and confident, is not what the better evidence shows.
Is it the opposite of imposter syndrome?
Not quite, although the two are often paired. Imposter syndrome is the feeling of being a fraud despite evidence of competence, and it is common, with a 2021 UK study reporting that 77% of people experience it. The Dunning-Kruger effect describes a gap between how well someone did and how well they think they did. Neither is a medical diagnosis, and one person can feel both in different areas of life.
Why confident people still get promoted
Even if confidence does not signal ignorance, it signals something to others. Dawson and de Meza argue that because the cost of misplaced confidence is lower for people who are more able, confidence can work as a credible signal of ability. That is a theory, not a proven fact, but it fits what many people see at work: the person who speaks with certainty in a meeting is noticed first.
That might be the new starter who volunteers for the big presentation, or the experienced colleague who is sure a deadline is achievable. Either can be right or wrong, and the trouble starts when a team treats certainty as proof.
How to check your own blind spots
Nobody finds a blind spot just by looking harder, but a few habits help. Write down a prediction before you find out the result, such as how long a task will take, and then compare the two. Over time, the gap shows you where your guesses run high or low.
Ask for specific feedback from someone who knows the area well, and ask what you missed rather than whether it was fine. Learning more helps too, because Kruger and Dunning found that training people in a skill helped them recognise how limited their earlier answers had been. If you manage people, make “I don’t know” an ordinary thing to say.
FAQs
- Is the Dunning-Kruger effect debunked? Not entirely. One study found the effect may be much smaller than reported, a larger replication found it was significant but minimal, and the 2026 paper argues the classic pattern is a statistical illusion. “Debunked” is too strong, but the claim that the least skilled are the most overconfident no longer looks safe.
- Is it a mental illness? No. It describes a pattern in how people judge their own performance, and it is not a medical condition. If you are worried about your thinking or mood, speak to your GP.
- What is an example of the Dunning-Kruger effect at work? A new employee might rate their first report as excellent because they do not yet know what a strong one looks like, and then be surprised by the feedback. The 2026 findings suggest the same kind of overconfidence can show up in experienced staff too.
- Can smart people have it? Yes. In the 2026 analysis, overconfidence appeared at every level of performance and grew with it. Seeking feedback helps at any level.
Takeaway
The Dunning-Kruger effect is real enough to be worth knowing, but it is smaller and shakier than the meme suggests, and a 2026 paper argues that the most able are the most overconfident. The evidence so far is a mix of older studies, a large replication, and one new paper that others still need to test. The practical lesson holds either way: confidence is a poor guide to accuracy, so check your guesses against feedback and results.
