Dunning-Kruger Effect: What the Research Shows

The Dunning-Kruger effect is the finding that people who score lowest on a test tend to rate their own performance far higher than it was. In the 1999 study that named it, the bottom quarter of participants scored around the 12th percentile and estimated themselves at the 62nd (Kruger and Dunning, 1999). The authors' explanation is a "dual burden": the skills you need to do a task well are the same skills you need to judge how well you did it.

Two things you may have seen online are not part of that finding. The famous "Mount Stupid" graph isn't in the paper, and the effect is still debated among researchers. This guide covers what the study found, what the critics say, and how to check where your own knowledge really stands.

Key Takeaways

Morso turns any topic into 5-minute lessons you'll actually finish. Try it free →

• In the 1999 study, the lowest scorers (12th percentile) estimated themselves at the 62nd (Kruger and Dunning, 1999)

• The "Mount Stupid" graph is not in that paper

• Critics call part of the effect a statistical artifact, but a 2021 model comparison found it is not only that (Jansen et al., 2021)

• An outside score beats how confident you feel

What Is the Dunning-Kruger Effect?

It is the tendency for people with low skill at a task to overestimate how well they did on it. Psychologists David Dunning and Justin Kruger tested it in four studies with Cornell University undergraduates, using tests of humor, grammar and logical reasoning. The studies had 65, 45, 84 and 140 participants (Kruger and Dunning, 1999).

Participants took a test, then estimated how they ranked against their peers. People in the bottom quartile scored around the 12th percentile on average, yet placed themselves around the 62nd. The top quartile erred the other way: they underestimated their standing relative to their peers.

Three limits are worth keeping in mind. The study measured specific tests, not general intelligence. It measured estimated rank, not arrogance as a personality trait. And the participants were students, so the numbers are a starting point, not a law of nature.

Why Do Low Performers Overestimate Themselves?

Kruger and Dunning called it a "dual burden". People who are unskilled make more mistakes, and "their incompetence robs them of the metacognitive ability to realize it" (Kruger and Dunning, 1999). Metacognition means thinking about your own thinking. If you don't know the rules of grammar well, you also can't spot which of your answers break them.

The most useful result for learners comes from the fourth study. The authors gave some bottom-quartile participants a short training session on the logic task. Afterward, those participants (19 of them) rated their ability at the 44th percentile and their test performance at the 32nd. The authors reported that they "became more calibrated in every way" and credited improved metacognitive skill.

In plain terms: when people learned the skill, they also got better at judging themselves. That is the link to studying.

Is the Mount Stupid Graph Real?

No. The curve with a peak labeled "Mount Stupid", a "Valley of Despair" and a slow climb back up is not in the 1999 paper. The study compared different people at one moment, grouped by how they scored. A curve that follows one learner's confidence over weeks or months needs a different kind of study.

That kind of study exists. Sanchez and Dunning (2018) followed beginners across six studies as they learned two different probabilistic tasks. Participants did not start out overconfident. After roughly 9 to 14 learning trials, their confidence rose well past their accuracy, a pattern the authors named the "beginner's bubble". Then confidence leveled off while accuracy kept climbing, though in those studies accuracy never caught up with confidence.

So an early confidence surge is real. The dramatic crash into a valley is not what those data showed. We also could not find any study that gives a number of hours at which a "valley" begins, so treat any such number as a guess.

Is the Dunning-Kruger Effect Real or a Statistical Artifact?

Researchers on both sides have published, and the argument isn't settled. The critics' case goes like this. When people estimate their own scores with some noise, the lowest scorers will, on average, guess above their real score and the highest scorers below it, even if nobody lacks self-insight. This is called regression to the mean.

Nuhfer and colleagues (2016) used simulated random variables to show that noise alone can produce graphs like the original figure, as described by Magnus and Peresetsky (2022). Gignac and Zajenkowski (2020) titled their paper "The Dunning-Kruger effect is (mostly) a statistical artefact". Magnus and Peresetsky concluded there is "no need for a psychological explanation".

The response came from Jansen, Rafferty and Griffiths (2021). They built a model where people reason sensibly from noisy information, then compared it with a model where low performers are worse at judging their own answers, and tested both against a large replication. The data fit the second model better. In a commentary on the paper, Mazor and Fleming (2021) wrote that the effect is not merely a statistical artifact at the group level, while suggesting it may be a "single burden" that shows up twice, rather than two separate burdens.

What survives in every account is a plain, useful point: your sense of how well you know something can sit far from your real score, and the gap tends to be widest at the low end. The argument is about why.

What Does It Look Like Outside the Lab?

One large study looked at news. Lyons and colleagues (2021) found that three in four Americans overestimated their relative ability to tell real from false news headlines, placing themselves 22 percentiles higher than warranted on average. The people who were most overconfident were more likely to visit untrustworthy websites and to say they'd share false content.

The authors note that their design can't show that overconfidence causes this. Treat it as a link, not a proven cause. It still shows why calibration matters when you're deciding what to believe and share.

How Can You Check Where Your Knowledge Really Stands?

Get a score from outside your own head. Feeling that you know something is a weak signal, because information that is easy to read gets mistaken for information you have learned (Bjork and Bjork, 2011). These four habits give you a better one:

1. Predict, then test. Before a quiz, write down the score you expect. The gap between that guess and your result is your calibration gap. This is a practical habit, not a result from the 1999 study.

2. Recall instead of rereading. In a classic experiment, students who read a passage once and took three recall tests remembered about 61% a week later, versus about 40% for students who read it four times (Roediger and Karpicke, 2006). Practice testing is one of only two techniques rated high-utility in the major review of study methods (Dunlosky et al., 2013). See active recall for the full method.

3. Explain it without notes. If you can't explain an idea in simple words, you've found a gap. The Feynman technique turns that into a routine.

4. Be most careful early. The beginner's bubble appeared after only 9 to 14 trials. The first few lessons in a new topic are when your confidence is least reliable, so test early and review on a schedule using spaced repetition.

For a wider set of exercises against overconfidence and other thinking errors, see critical thinking exercises.

How Do You Do This With Morso?

Morso is built around the outside-score habit. Type any topic, such as "cognitive biases" or "how memory works", and it generates a course of bite-sized lessons. Every lesson ends with a quiz, and about every three lessons a practice quiz brings back the questions you got wrong. XP, streaks and the Intellect Score give you a running number that doesn't depend on how sure you feel. Try it free on any topic.

To be clear about what is supported: the fourth study tested a short logic training session, not quizzes or apps. Using quiz results as your outside score is our practical inference from that result and from the research on practice testing. Kruger and Dunning didn't test it.

The Bottom Line

The Dunning-Kruger effect is a real pattern in the data, but it's narrower than the memes. It was measured on specific tests, the "Mount Stupid" curve isn't from the study, and researchers still disagree about how much of it is psychology and how much is statistics. The part that holds up for learners is simple: learning a skill improves your ability to judge it, and an outside score tells you more than a feeling of confidence.

If you want a place to start, pick one topic from this list of interesting topics to learn about, type it into Morso and see how your first quiz compares with your guess.

Sources

Kruger, J. and 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. Retrieved 2026-10-22. https://pubmed.ncbi.nlm.nih.gov/10626367/

Sanchez, C. and Dunning, D. (2018). Overconfidence among beginners: Is a little learning a dangerous thing? Journal of Personality and Social Psychology, 114(1). Retrieved 2026-10-22. https://psycnet.apa.org/record/2017-49272-001

Gignac, G. E. and Zajenkowski, M. (2020). The Dunning-Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data. Intelligence, 80, 101449. Retrieved 2026-10-22 (title and findings as cited in Magnus and Peresetsky, 2022). https://doi.org/10.1016/j.intell.2020.101449

Magnus, J. R. and Peresetsky, A. A. (2022). A statistical explanation of the Dunning-Kruger effect. Frontiers in Psychology. Retrieved 2026-10-22. https://pmc.ncbi.nlm.nih.gov/articles/PMC8992690/

Jansen, R. A., Rafferty, A. N. and Griffiths, T. L. (2021). A rational model of the Dunning-Kruger effect supports insensitivity to evidence in low performers. Nature Human Behaviour, 5(6). Retrieved 2026-10-22. https://doi.org/10.1038/s41562-021-01057-0

Mazor, M. and Fleming, S. M. (2021). The Dunning-Kruger effect revisited. Nature Human Behaviour, 5, 677-678. Retrieved 2026-10-22. https://doi.org/10.1038/s41562-021-01101-z

Lyons, B. A., Montgomery, J. M., Guess, A. M., Nyhan, B. and Reifler, J. (2021). Overconfidence in news judgments is associated with false news susceptibility. Proceedings of the National Academy of Sciences, 118(23), e2019527118. Retrieved 2026-10-22. https://www.pnas.org/doi/10.1073/pnas.2019527118

Bjork, E. L. and Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In Psychology and the Real World, 56-64.

Roediger, H. L. and Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249-255. Retrieved 2026-10-21. https://pubmed.ncbi.nlm.nih.gov/16507066/

Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J. and Willingham, D. T. (2013). Improving students' learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4-58. Retrieved 2026-10-21. https://journals.sagepub.com/doi/10.1177/1529100612453266