What is the Dunning-Kruger effect? And what it isn't

The Dunning-Kruger effect is the finding that people who perform worst at a task tend to overestimate their performance the most, partly because the skills needed to do something well are the same skills needed to judge whether you did it well. It is named after the psychologists David Dunning and Justin Kruger, who described it in 1999. It doesn't mean that the least skilled are the most confident.

In short

• In the 1999 study, people in the bottom quarter on tests of logic, grammar and humor scored at about the 12th percentile on average but guessed they were at the 62nd.
• Nearly everyone rated themselves above average; the low scorers were just the furthest off. The popular "Mount Stupid" chart, where beginners are more confident than experts, is not what the study found.
• Part of the pattern is a statistical artifact that appears even in random data, and researchers still argue over how much real effect is left.

What did Dunning and Kruger actually find?

In their 1999 paper, Unskilled and Unaware of It, Kruger and Dunning, then at Cornell University, gave students tests of humor, grammar and logical reasoning across four studies, then asked them to estimate how they did compared with their peers. According to the paper's abstract, the people who scored in the bottom quarter had test scores at the 12th percentile on average, but estimated themselves at the 62nd. Top scorers tended to underestimate their rank somewhat, partly because they assumed others found the test as easy as they did.

The authors called this a dual burden. People who lack a skill make mistakes, and the same lack keeps them from recognizing the mistakes. Judging whether a sentence is grammatical takes knowledge of grammar. In one study, they gave some of the students a short lesson in logic, and afterwards the low scorers among them rated their earlier performance more accurately, even though their earlier answers hadn't changed.

What does the famous chart get wrong?

A chart that circulates online shows confidence shooting up to a peak, often labeled "Mount Stupid", for people who know a little, crashing into a "valley of despair", and climbing slowly back for experts. That curve appears nowhere in the 1999 paper. What the paper's graphs actually show:

	The viral chart	The 1999 study
What's on the horizontal axis	experience or knowledge	actual test score, in quarters
Who is most confident	beginners, more than experts	top scorers, slightly; nearly everyone rated themselves above average
What the low scorers do	feel like experts	guess they are a little above average while scoring near the bottom
What it's about	a journey over time	one test, taken once
Is it evidence	no, a drawing	yes, four studies, all with Cornell students

In the study, the worst performers did not think they were better than the best performers. They thought they were somewhat above average. Their estimates of themselves rose with their real scores; they rose much less steeply.

Is the Dunning-Kruger effect just statistics?

Partly. Two ordinary forces can produce the same picture with no difference in self-insight at all. One is the better-than-average effect: most people rate themselves above the middle in most things. The other is regression to the mean: when two measures are imperfectly related, people who are extreme on one tend to be less extreme on the other.

Worked example: making the pattern from nothing

Imagine 100 people who know nothing about their own ability. Every one of them guesses they are at about the 60th percentile, give or take a little at random. Now sort them by their real test scores into four groups of 25.

• Bottom quarter: real average about the 12th percentile, guess about the 60th. Overestimate: about 48 points.
• Second quarter: real average about the 37th, guess about the 60th. Overestimate: about 23 points.
• Third quarter: real average about the 62nd, guess about the 60th. About right.
• Top quarter: real average about the 87th, guess about the 60th. Underestimate: about 27 points.

That is the classic Dunning-Kruger graph, produced by people who all have equally poor insight into themselves. A flat line of guesses, set above average, plotted against real scores, makes the low scorers look the most deluded. A 2016 study in the journal Numeracy by Edward Nuhfer and colleagues used random-number simulations to show how much of this kind of graph noise alone can draw.

Still argued

How much real effect is left is disputed. Gilles Gignac and Marcin Zajenkowski, in a 2020 paper in Intelligence titled The Dunning-Kruger effect is (mostly) a statistical artefact, tested intelligence scores and self-estimates from 929 adults with methods that avoid the artifact, found no statistically significant sign of it, and concluded that where the effect is real it may be much smaller than reported. On the other side, a 2021 study in Nature Human Behaviour by Rachel Jansen, Anna Rafferty and Thomas Griffiths replicated the original design with about 4,000 people in each of two studies and modeled the artifact explicitly; it still found that low performers were worse at telling whether their own answers were right, in grammar and logical reasoning. A fair summary: the effect is real but smaller than the famous graphs suggest, and much of the gap they show comes from everyone, not just the unskilled, overrating themselves.

How can you check your own confidence?

The useful lesson is not about other people. Everyone is a beginner at most things, and the effect, where it is real, is invisible from the inside. Some ways to check:

• Predict, then score. Before you check answers, write down how many you think you got right. Over time, the gap shows how well calibrated you are.
• Explain it back. Try to explain the subject without notes. Gaps that a sense of familiarity hides show up fast.
• Get feedback from someone who knows. Self-assessment improves most with outside correction, which is what the logic lesson in the 1999 study provided.
• Learn a little more. The study's own finding is that training improves self-judgment along with skill.
• Watch which examples you see. Judging your skill only against people who quit or failed, or only the famous successes, skews the comparison, as survivorship bias explains.

Two cautions. The effect is a finding about averages on specific tests, not a label for a person or a group, and using it to call someone stupid misreads it. And its mirror image, feeling like a fraud despite real skill, is a separate idea, impostor syndrome. When evidence challenges how good we think we are, the discomfort and the urge to explain it away belong to cognitive dissonance. The habits above are part of critical thinking and, further back, of the oldest advice in philosophy: know what you don't know.

Check yourself

• Without looking back: what did the bottom quarter in the 1999 study score, and what did they guess?
• How can a group of people with equally poor self-insight still produce the Dunning-Kruger graph?
• What is wrong with the "Mount Stupid" chart as a picture of the study?

Sources

• Unskilled and unaware of it: how difficulties in recognizing one's own incompetence lead to inflated self-assessments — Kruger and Dunning, Journal of Personality and Social Psychology, 1999
• Random number simulations reveal how random noise affects the measurements and graphical portrayals of self-assessed competency — Nuhfer and others, Numeracy, 2016
• The Dunning-Kruger effect is (mostly) a statistical artefact — Gignac and Zajenkowski, Intelligence, 2020
• A rational model of the Dunning–Kruger effect supports insensitivity to evidence in low performers — Jansen, Rafferty and Griffiths, Nature Human Behaviour, 2021