Gilles Gignac

Who they are

This page is built on the 2020 paper. Its title and journal place it in research on intelligence and individual differences, which is as much as the vault verified about his field. His institution and wider career were not checked, so none is stated.

The paper argues that the usual test of the Dunning-Kruger hypothesis, comparing self-assessed and measured ability across quarters of a sample, is confounded by the better-than-average effect and by regression to the mean. It proposes two alternatives, the Glejser test of heteroscedasticity and quadratic regression, and shows with simulated data that the quartile method can produce a Dunning-Kruger-like pattern when none is present.

The sample was 929 community participants, about half women, with a mean age of about 27. Each gave a single self-rating of intelligence and sat the Advanced Progressive Matrices. Self-rated IQ averaged about 124 against a measured average of about 102. Sorted into quarters, the gap between the two shrank from about 33 points in the bottom quarter to about 13 in the top. Neither alternative test found an effect. The authors concluded that the effect is mostly, rather than entirely, a statistical artefact, and noted the limits of using one ability measure.

The 2023 exchange with Hiller

In 2023 Avram Hiller published a comment on the 2020 paper in Intelligence (volume 97, article 101732). He argued that the finding of equal prediction error across levels of intelligence does not show that the Dunning-Kruger effect is merely a statistical artefact, and that the result probably comes from the authors' conversion of a comparative self-rating of intelligence onto a linear scale, when a normal scale would have fitted better. He also cautioned researchers working with self-assessed intelligence to watch for measurement problems when ordinal ratings are treated as interval data. Gignac and Zajenkowski replied the same year in Intelligence (volume 97, article 101733), under the title Still no Dunning-Kruger effect. They re-ran the analysis with the transformation Hiller suggested and again found no support for the effect, noted that the one study they could find using a percentile-based approach also largely failed to support it, and encouraged researchers to keep developing new ways to test the hypothesis. The record of the comment and the reply was checked at Crossref, PhilArchive and the University of Western Australia repository on 2026-10-25; the vault read the abstracts and summaries, not the full texts.

Key papers

Gignac, G. E., & 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.

Gignac, G. E., & Zajenkowski, M. (2023), "Still no Dunning-Kruger effect: A reply to Hiller," Intelligence, 97, 101733.

Key ideas

The quartile method used in the original studies can create the Dunning-Kruger shape from the better-than-average effect and regression to the mean alone.

Tests that do not depend on grouping, such as the Glejser test and quadratic regression, are better suited to individual-differences data.

The effect's true size may be much smaller than reported. The conclusion was challenged in 2023 by Dunkel and colleagues, who repeated the analyses on a nationally representative sample of 13,977 and found a significant but small effect.

---