The Flynn effect

The most complete count is the 2015 meta-analysis by Pietschnig and Voracek. It pooled 271 samples and almost 4 million participants from 31 countries, covering 1909 to 2013. It found gains of about 3 IQ points per decade overall. The gain depends on the kind of test.

Not obviously. Flynn himself argued that gains of this size, 15 points in a generation on some tests, could not mean that our grandparents were close to the borderline of intellectual disability by modern standards. He suggested that people had become better at one kind of thinking: abstract, scientific, test-style reasoning, as work and schooling asked for more of it.

The pattern supports that view. Gains are largest on fluid tests and smaller on vocabulary and knowledge. A 2013 meta-analysis by te Nijenhuis and van der Flier found that subtests that depend less on the general factor g showed larger gains (a correlation of -.38 across studies), which suggests the rise is not simply a rise in g. The question is still open, and researchers disagree about how much of the rise is real cognitive change and how much reflects test familiarity and changing item properties.

The most cited candidate causes are better nutrition and health, more and longer schooling, smaller families, and a world with more abstract, test-like demands. None of these has been shown to be the single cause, and the effect probably has several.

Because of the Flynn effect, an old test overstates a person's standing. Someone scoring 100 on 1990 norms would score lower on 2020 norms. Publishers therefore restandardise their tests every ten to fifteen years. The WAIS-5 was normed in 2023 to 2024. When 186 adults aged 16 to 90 took both it and the WAIS-IV, they averaged 1.9 points higher on the older test (Winter, Trudel and Kaufman, 2024). That is about 1.2 points per decade, well under the 3 points a decade that earlier Wechsler comparisons found.

The practical rule: when you compare two scores, check which edition produced each one. An old test flatters a score by a point or more for every ten years of age in its norms. This matters most for scores near a cut-off such as 130, where a few points decide eligibility. The Mensa eligibility checker shows the cut-offs by test.

In some countries, yes. Teasdale and Owen analysed tests of more than 500,000 young Danish men, tested between 1959 and 2004. Scores peaked in the late 1990s and then fell back to roughly early-1990s levels. Norwegian conscript data show a similar turn.

The best evidence on cause comes from Norway. Bratsberg and Rogeberg used about 730,000 test results from men born from 1962 to 1991. Scores rose until the cohorts born around 1975, then fell. Because they compared brothers within the same families, they could show that the rise and the fall both appear inside families. That rules out explanations based on who is having children or on immigration, and it points to environmental causes such as changes in schooling, media and reading habits, though the study does not identify which.

The picture is not uniform. Pietschnig and Voracek found that gains had already decreased in recent decades. A 2023 study of almost 400,000 American adults tested online from 2006 to 2018 found declines in several areas, including matrix reasoning and letter and number series, while 3D rotation scores rose. These are online samples, so they are not strictly comparable with norming samples.

The size of the correction depends on which estimate you trust. The long-run figure of about 3 points per decade comes from the meta-analysis of the whole century. The recent WAIS-IV to WAIS-5 comparison suggests about 1.2 points per decade for American adults today.

Worked example. Someone scored 130 on a test normed 20 years ago. At 3 points per decade the same performance would be worth about 124 on current norms. At 1.2 points per decade it would be worth about 127.6. Either way the person may now be on the wrong side of a cut-off, which is why institutions that use hard thresholds specify the edition.

Pietschnig and Voracek also found that the gains were stronger for adults than for children, and that they had decreased in more recent decades. A single correction factor is a rough device, and a score from a test normed ten years ago is a closer guide than one normed forty years ago.

Three things. First, a single IQ number belongs to a test edition and a year, so compare scores only on the same norms. Second, differences between generations on IQ tests are not a direct measure of ability, and are not evidence about who is "smarter". Third, the changes show that test performance responds to environment, which fits the evidence that heritability describes differences inside a group at one moment and does not fix a person's level.

For a wider guide to how scores are built, see IQ meaning, which also covers age, education and what IQ predicts.