Confidence is usually described as a fixed inner quality, something a person either projects or doesn’t. Two classic, well-established studies suggest something more specific and more useful: how well-founded a person’s confidence actually is depends heavily on real skill and on the exact nature of the task in front of them, not on some stable, general trait.
The foundational evidence comes from a 1999 study by Justin Kruger and David Dunning, published in the Journal of Personality and Social Psychology, the original research behind what’s now widely known as the Dunning-Kruger effect. Across four studies with Cornell undergraduates, testing skills like humor judgment, logical reasoning, and grammar, the researchers found that participants who scored in the bottom quarter on these tests dramatically overestimated their own performance, on average rating themselves around the 62nd percentile despite actually scoring closer to the 12th. Their explanation was specific: the same skills required to perform well on a task are often the same skills required to accurately judge how well you performed, so someone lacking the underlying competence often lacks the tools to recognize that shortfall in the first place. Critically, in a follow-up manipulation, the researchers actually trained participants and improved their real skill, and their self-assessments became noticeably more accurate as a result. Genuine competence, not just experience or exposure, was what improved the calibration.
A second, later study complicates the picture further by showing that overconfidence itself isn’t one single, consistent thing. In a 2008 study by Don Moore and Paul Healy, published in Psychological Review, the researchers ran an experiment with 82 participants who competed through 18 rounds of trivia questions varying in difficulty, and also reviewed and synthesized a wide body of existing overconfidence research. They distinguished three genuinely different phenomena that all get lumped together under “overconfidence”: overestimating your own raw performance, overplacing yourself relative to other people, and overprecision, being too certain your specific belief is correct. In their trivia experiment, the pattern flipped depending on task difficulty: on hard questions, people underestimated how they compared with others but still overestimated their raw score, while on easy questions the opposite pattern showed up. Confidence, in other words, wasn’t tracking one stable internal setting, it was shifting in specific, predictable ways depending on how hard the task actually was.
Together, these two studies offer a much more precise picture of what separates genuine confidence from the unfounded kind than the usual advice to simply “believe in yourself.” Kruger and Dunning’s research shows that real, accurate self-assessment depends on actually having the underlying skill, not just feeling sure, and that skill improvement is what closes the gap. Moore and Healy’s research shows that even well-intentioned confidence judgments shift in specific, systematic ways depending on the difficulty of what’s actually being attempted, meaning a person can be simultaneously overconfident in one sense and underconfident in another, on the very same task. Both point toward the same underlying idea: confidence that’s actually worth trusting comes from real competence checked against real, calibrated feedback, not from a general feeling of certainty.
It’s worth being honest about what these two studies don’t establish. Kruger and Dunning’s research was conducted specifically with university undergraduates on tasks like grammar and logic puzzles, so how precisely the same pattern applies to more complex, real-world skills like leadership or parenting is a reasonable extrapolation rather than something the original study tested directly. Moore and Healy’s trivia experiment is a controlled, artificial task, useful for isolating the mechanism cleanly, but it doesn’t capture every real-world situation where the stakes, time pressure, or social context might change how confidence gets calibrated. Both studies are also decades-old classics in a field that has continued to refine and debate these exact findings since, so they’re foundational rather than final words on the subject.
Within those honest limits, the research supports a specific, more useful version of “the art of confidence” than simply projecting certainty. It looks less like a performance and more like an ongoing calibration process, building real skill, getting honest feedback, and staying alert to the fact that confidence can be running high in one respect and low in another on the exact same task, sometimes without a person noticing either.
