Published on 2026-07-25
Why are the least skilled people often the most sure they're right?
A 1999 study found the worst performers on a test rated themselves near the top. Confidence and competence, it turns out, don't always travel together.
You've sat in that meeting. Someone who clearly hasn't done the reading talks over everyone else with total certainty, while the person who actually knows the subject hedges every sentence with 'I think' and 'probably'. It feels backwards, and it is, but it isn't random. There's a documented pattern behind it, and it was measured directly, on paper, with a test.
A grammar test that measured something else entirely
In 1999, the psychologists David Dunning and Justin Kruger, working at Cornell University, ran a set of studies on undergraduates. They tested students on grammar, on logical reasoning, and, in one version, on humour, asking sixty five of them to rate how funny a set of jokes was and comparing their judgement against professional comedians. After each test, every student was asked one more question: how well do you think you did, compared to everyone else who took it?
The students who landed in the bottom quarter of actual scores, in the twelfth percentile, on average guessed they'd scored around the sixty second percentile. Not a small miss. A gap of fifty percentile points between where they ranked and where they believed they ranked. Meanwhile, the students who scored in the top quarter did the opposite: they slightly underestimated themselves, guessing somewhere around the seventieth to seventy fifth percentile despite having genuinely outscored almost everyone in the room.
The gap wasn't between confidence and modesty. It was between two different jobs the mind has to do: performing a task, and judging your own performance on it.
The missing skill has a name
Dunning and Kruger's explanation centres on metacognition, the ability to step back and evaluate your own thinking. Grammar, logic and humour all require a certain competence to perform. It turns out they also require a related but separate competence just to judge how well you performed, and that second skill is exactly what weak performers are missing. If you don't know enough about grammar to spot your own errors, you also don't know enough about grammar to notice that you're making them. The same blind spot that produces the mistakes hides the mistakes from view.
The researchers tested this directly. They gave the lowest scoring students a short training session in logical reasoning. Once their actual skill improved, their self-assessments dropped, and dropped accurately. Learning the subject was what let them finally see how shaky their earlier answers had been. Competence, in other words, is partly what teaches you to doubt yourself in the right places.
Not a story about a journey
One popular version of this idea pictures confidence as a curve: a beginner starts wildly overconfident, then crashes into doubt, then climbs slowly back to earned mastery. That image is memorable, and it isn't quite what the original research measured. Dunning and Kruger's studies were a snapshot, one test, one moment, comparing people at different skill levels at the same point in time, not the same person tracked across months of learning a skill. Treat the finding for what it actually is: at any given moment, the people with the least skill in a domain are, on average, the ones least equipped to notice that gap in themselves. That's unsettling enough without adding a tidy narrative arc that the data doesn't quite support.
It's also worth being honest about the limit here: this isn't a licence to dismiss anyone who sounds sure of themselves, and it isn't a reason to trust every hesitant hedge as automatic proof of expertise. Confident people are sometimes confident because they've earned it. The pattern is a tendency, well documented but not universal, and it says the most useful thing about the people closest to a topic's edge, not about every confident voice in the room.
There's a flip side worth naming too: genuine experts often carry a private worry that everyone else must know what they know, so their own hedging can read, wrongly, as a lack of authority. That mismatch is part of why loud certainty travels so well in a room and quiet expertise so often doesn't. Neither volume proves anything on its own. What would actually settle the question is the same thing in both cases: checking the claim against the work, not against how it was delivered.
Here's the harder question this raises, and it points at you as much as anyone else: in whatever you feel most sure about right now, how would you know if that certainty came from real command of the subject rather than simply not knowing enough about it to see the gaps? There's no test score to check that against in daily life, no comedians rating your jokes. Just the habit of asking the question before someone else has to ask it for you.
Sources
- Unskilled and Unaware of It (Kruger and Dunning, 1999, University of Michigan archive) (sites.lsa.umich.edu)
- Dunning-Kruger effect (EBSCO Research Starters) (ebsco.com)