The famous Dunning-Kruger effect may be backwards: once statistical noise is removed, the highest performers — not the lowest — show the greatest overconfidence.
Inside the research
Researchers at the University of Bath and the London School of Economics and Political Science (LSE) re-examined one of psychology's most repeated findings. For decades, studies have shown that people who score poorly on tests overpredict their performance by wide margins, while high scorers are more accurate. That pattern became the internet's favorite explanation for clueless confidence. The new paper, published in Psychological Review, argues the pattern was never about psychology at all.
Professor Chris Dawson (University of Bath) and Professor David de Meza (LSE) applied improved statistical modeling to data from thousands of participants across massive replication datasets. Their key insight: test scores are volatile. Luck, ambiguous questions, and day-to-day variation all pull scores up or down. When participants are sorted purely by their test score, a low score is unusually likely to reflect bad luck — so the person's actual ability looks higher than the score implies. A high score is unusually likely to reflect good luck, so ability looks lower. This is regression to the mean, and it mathematically manufactures the Dunning-Kruger pattern.
Once the researchers corrected for this bias in both test scores and self-assessments, the classic pattern vanished. What remained was the opposite: overconfidence appeared across all skill levels, and its size increased with competence. The best performers were the most overconfident.
"Overconfidence is a universal human trait, but it is the most capable among us who exhibit it the most." — Professor Chris Dawson, University of Bath
The authors also propose an evolutionary explanation. If competence is hard for others to observe, high performers may have the strongest incentive to project confidence — a behavioral strategy that signals ability.
Why this matters for your brain
This isn't just a stats argument. It changes how you should read your own self-assessments. If you score lower than expected on a test, that single result may understate your true ability — the same mathematical trap applies to you. If you score high, don't assume your confidence is fully earned. Even top performers systematically overrate themselves, according to this corrected model.
It also means you should never judge a person's whole ability from one test session or one bad day. One score is a noisy signal; patterns over time are far more informative.
What you can do
- Retest rather than trust a single result — repeated scores average out luck.
- Track your estimate versus actual score over several tries, not one sitting.
- Before big decisions, ask whether your confidence exceeds your evidence.
- Treat unusually low scores as possible noise, not proof of ceiling.
Source: Neuroscience News
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.