Mount Stupid Is Real: What the Dunning-Kruger Effect Actually Says

The Paper That Became a Punchline

In 1999, two Cornell psychologists named Justin Kruger and David Dunning ran a series of experiments on undergraduate students. They tested the students’ abilities in three domains: logical reasoning, English grammar, and humor. Then they asked the students to estimate how well they had done.

The students who scored in the bottom quartile had a peculiar problem. They consistently overestimated their performance. Not by a little. The worst-performing students believed they had scored, on average, around the 62nd percentile. They had actually scored around the 12th.

This mismatch between actual and perceived performance became the foundation of one of the most cited and most misunderstood papers in modern psychology. Kruger and Dunning published their findings in the Journal of Personality and Social Psychology in 1999, under the title “Unskilled and Unaware of It.” The paper has been cited over 6,000 times and almost certainly read by far fewer people than have confidently summarized its conclusions.

Before explaining what the Dunning-Kruger effect actually says, it is worth pausing on what has happened to it since. The paper produced a theory that became a talking point that became a meme that became a universal tool for explaining why the other person in any disagreement is incompetent and overconfident. Which is, ironically, a perfect illustration of the very bias it describes.

What Kruger and Dunning Actually Found

The 1999 paper made several specific claims worth going through carefully, because the popular summary loses most of them.

Finding 1: Incompetent people overestimate their own ability

This is the part everyone knows. Students who performed in the bottom quartile on tests of grammar and logical reasoning believed they had done significantly better than they had. In the logical reasoning study, bottom-quartile participants estimated they had scored at the 61st percentile when they had actually scored at the 12th.

Kruger and Dunning’s explanation drew directly on the work of John Flavell, the Stanford psychologist who coined the term metacognition in 1979. Flavell defined metacognition as thinking about one’s own thinking, specifically the capacity to monitor and evaluate one’s own cognitive processes. Kruger and Dunning argued that the skills required to perform well at a task are often the same skills required to recognize good performance. If you lack the knowledge to solve a problem correctly, you also lack the knowledge to recognize that your solution is wrong.

Finding 2: Competent people underestimate their relative ability

This finding gets almost no coverage in popular discussions, which is a shame because it is equally interesting. Students who scored in the top quartile consistently underestimated their relative performance. They assumed the tasks were straightforward for everyone, which led them to assume their peers had done just as well. In the logical reasoning study, top-quartile participants estimated they had scored at around the 68th percentile. They had actually scored at the 86th.

So the Dunning-Kruger effect, properly understood, describes a double distortion. Low performers overestimate upward. High performers underestimate downward. The confident person and the expert are both wrong about where they stand, just in opposite directions.

Finding 3: Training improves both performance and calibration

In a follow-up phase, participants who received training in logical reasoning performed better and also became more accurate in assessing their performance. This is the finding that has the most practical implications and the least popular coverage: competence and self-awareness develop together. You cannot simply tell someone they are overestimating themselves and expect it to help. You have to teach them the subject. The awareness follows.

Mount Stupid: Where the Graph Came From

If you have seen the Dunning-Kruger curve online, the one that shows a steep peak of confidence early in learning, a sharp drop into a “valley of despair,” and then a slow rise toward genuine expertise, that graph is not from the 1999 paper. It does not appear anywhere in Kruger and Dunning’s research.

The graph is a folk model, a conceptual diagram that spread online and was gradually attributed to the original study through repetition. The term “Mount Stupid” for the peak of early overconfidence was popularized on the internet and later widely associated with the Dunning-Kruger effect, but it is not Kruger and Dunning’s language.

This does not mean the model is wrong. It means it is an elaboration rather than a finding. Kruger and Dunning’s study was a cross-sectional snapshot comparing novices and experts. The curve model implies a developmental trajectory, suggesting that all learners move through these phases as their expertise grows. That developmental claim has some empirical support from separate research on skill acquisition, but it is not what the 1999 paper demonstrated.

The reason this matters is that the graph is now the primary way most people encounter the concept. But it represents a significantly tidied-up, more intuitive version of what is actually a messier and more interesting set of findings.

The Replication Problem Nobody Talks About

In 2020, Gilles Gignac at the University of Western Australia and Marcin Zajenkowski at the University of Warsaw published a reanalysis of the Dunning-Kruger effect in the journal Intelligence, and their findings created a stir among researchers that has barely registered in popular coverage.

Their argument is technical but worth understanding. When you ask people to estimate their own performance and then compare that estimate to their actual performance, you will almost always find that poor performers overestimate and good performers underestimate. This pattern can emerge purely from random measurement error, a phenomenon known as regression to the mean, without any psychological mechanism driving it. When Gignac and Zajenkowski applied corrections for this issue, the effect size shrank substantially.

Other researchers have pushed back on this critique, arguing that the statistical corrections applied were themselves debatable. The debate is ongoing. But it has made the picture more complicated than either side of the popular argument tends to acknowledge.

David Dunning, in a follow-up paper in Advances in Experimental Social Psychology, described a broader principle: that people rely on the very knowledge and skills they are trying to evaluate in order to make their evaluations. Everywhere that knowledge is limited, self-assessment will be distorted. And knowledge is limited everywhere. Ignorance, in his account, is not an occasional problem but a persistent background condition.

The Dunning-Kruger Effect in the Workplace

The workplace is where the Dunning-Kruger effect gets most urgently applied and most frequently misapplied.

Research on team cognition found that high-performing teams are distinguished not by the average expertise of their members but by the accuracy of each member’s understanding of their own and their teammates’ capabilities. The term used in that literature is “shared mental models.” Teams fail when members overestimate their own contribution and underestimate others’, when they assume they understand a problem that another person has spent years working on.

The workplace version of the Dunning-Kruger problem is not primarily about the junior employee who is overconfident in their first month. It is about the escalation of commitment to bad decisions by people who are senior enough to have stopped receiving honest feedback. The more authority someone accumulates, the more their social environment shapes their confidence in ways that have nothing to do with their actual competence.

This is a point that Amy Edmondson at Harvard Business School has made in her work on psychological safety: environments where it is socially costly to say “I don’t know” or “I think that’s wrong” are environments where the Dunning-Kruger effect is systematically amplified rather than corrected.

What the Dunning-Kruger Effect Is Not

Several things are commonly attributed to the Dunning-Kruger effect that the research does not support.

It is not a finding that stupid people are more confident than smart people. The studies compared novices and experts within a domain, not people of different general intelligence levels. A person who is highly competent in their professional field can be a Dunning-Kruger-style novice in another domain. The effect is domain-specific, not a personality trait.

It is not primarily about arrogance. The overconfident novice in Kruger and Dunning’s experiments was not someone being stubborn or defensive. They were someone who genuinely did not know what they did not know. The corrective is not humility as a character virtue. The corrective, as Kruger and Dunning’s training experiments showed, is knowledge.

It is not a reliable way to identify whether you have it. The honest answer is inconvenient: the very feature of the effect that makes it interesting is that it is not directly introspectable. The person on Mount Stupid feels exactly as confident and competent as they would feel if they were actually an expert. That is the problem.

Metacognition: The Skill That Changes Things

John Flavell introduced the concept of metacognition in 1979 in a paper in American Psychologist, and the term has since become central to educational psychology, cognitive science, and the study of expertise.

Metacognition is thinking about your own thinking. Not in a vague, introspective sense, but in the specific, technical sense of monitoring your own cognitive processes, evaluating their reliability, and adjusting them when they are going wrong. It is the skill that allows an expert to say “I am not sure about this, I should check” while a novice says “I’m confident” about the same type of problem.

Research building on Flavell’s framework found that metacognitive ability is one of the strongest predictors of academic achievement, stronger than IQ in some studies, because it governs how effectively people learn from experience. Students with high metacognitive skills not only know more but know when they do not know, which means they seek information in the right places and at the right times.

The path out of Mount Stupid is not an act of will. It is an act of learning. The person who becomes genuinely more accurate in their self-assessments is usually the person who has gone deep enough in a field to understand the gap between novice and expert performance, not the person who has adopted an attitude of general humility.

Flavell’s later work on metacognitive experiences described how skilled thinkers differ not just in what they know but in how they interpret the feeling of uncertainty. A novice who encounters confusion tends to interpret it as a signal that the answer is nearby. An expert who encounters confusion tends to interpret it as a signal that their current model might be wrong. The confusion feels the same. The interpretation differs entirely.

How Do You Know If You Have the Dunning-Kruger Effect?

This is the question that most popular summaries rarely answer honestly. The answer is that you mostly cannot tell from the inside.

You are more likely to be at the overconfidence peak when you have recently learned enough about something to follow the conversation but not yet enough to understand what the best practitioners in that field consider hard. The internet is particularly efficient at producing this state. A few hours of reading can make complex domains feel navigable, because the introductory material is written to be accessible and does not convey the texture of the problems that remain.

You are also more likely to be overconfident when you are receiving mostly positive feedback in a social environment rather than the corrective feedback of actually attempting things. Dunning described this in his follow-up paper as one of the key conditions for inflated self-assessment: environments that reinforce a sense of competence without actually testing it.

And there is one more test, which is the most useful. When you encounter a piece of information that contradicts your current understanding of something you feel confident about, notice your first reaction. Is it curiosity, a pull toward finding out if the new information changes the picture? Or is it dismissal, an instinct to explain why the new information does not count?

The curiosity response is a sign of metacognitive health. The dismissal response is a sign you might want to look more carefully at what you think you know.

Wondering where you fall on the curiosity-to-certainty spectrum? Take the One Good Thing Thinking Style Quiz. It takes about three minutes and maps your natural tendencies across curiosity, openness, and reflective thinking.

The Part of the Dunning-Kruger Story That Actually Matters

The popular version of the Dunning-Kruger effect has turned into a way of explaining other people. You see it in comment sections, in arguments, in the smug tone of someone who has recently read a Wikipedia article about cognitive biases. It has become a tool for dismissal rather than reflection.

Dunning, to his credit, has consistently pushed against this use. In his widely-cited essay “We Are All Confident Idiots”, he wrote that the lesson of his research is not that some people are incompetent and we should identify them. The lesson is that we are all incompetent in more domains than we realize, that confidence is a poor guide to actual knowledge, and that the best response to this is not cynicism about human cognition but investment in the habits and environments that improve metacognitive accuracy.

The research on metacognition, from Flavell forward, suggests that the skill of accurately assessing one’s own knowledge is one of the most trainable and most neglected skills in formal education. Schools teach content. They rarely teach students how to monitor their own understanding of that content. The student who leaves with an A and no awareness of where their understanding has edges is arguably less prepared for continued learning than the student who leaves with a B and a clear sense of what they do not yet understand.

The drop from peak confidence to the valley of despair is not a sign of failure. It is a sign of learning. The moment when something reveals itself as more complex than you thought is the moment when metacognitive development begins in earnest. And it is where the pause before certainty snaps back into place, the most fertile cognitive space there is, becomes genuinely open rather than rhetorical.

This connects directly to what we explored in how System 1 and System 2 thinking operate under digital conditions: the environment either makes metacognitive monitoring easier or harder, and most digital environments make it significantly harder.

Frequently Asked Questions

Further Reading

• Kruger, J. & Dunning, D. (1999). Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments. Journal of Personality and Social Psychology, 77(6), 1121–1134.
• Gignac, G. E. & Zajenkowski, M. (2020). The Dunning-Kruger Effect Is (Mostly) a Statistical Artefact. Intelligence, 80, 101449.
• Flavell, J. H. (1979). Metacognition and Cognitive Monitoring: A New Area of Cognitive-Developmental Inquiry. American Psychologist, 34(10), 906–911.
• Edmondson, A. C. (1999). Psychological Safety and Learning Behavior in Work Teams. Administrative Science Quarterly, 44(2), 350–383.
• Brown, A. L. (1987). Metacognition, Executive Control, Self-Regulation, and Other More Mysterious Mechanisms. In Metacognition, Motivation, and Understanding. Lawrence Erlbaum.
• Salas, E., Sims, D. E. & Burke, C. S. (2005). Is There a 'Big Five' in Teamwork? Small Group Research, 36(5), 555–599.
• Dunning, D. (2011). The Dunning-Kruger Effect: On Being Ignorant of One's Own Ignorance. Advances in Experimental Social Psychology, 44, 247–296.
• Dunning, D. (2014). We Are All Confident Idiots. Pacific Standard.