Is IQ Debunked? What Critics Get Right and Wrong

Is IQ debunked? No. The IQ score is one of the best-studied measurements in psychology, and the popular claims that it is meaningless, fake or a scam do not survive the data. What has been challenged, with more success, are the stories built around it: that it is a fixed number, that it captures all of intelligence, and that a single hidden factor called g explains it. This article sorts the common criticisms into those that hold up and those that do not, with the numbers behind each.

Is IQ real? What stability and prediction show

A real measurement should be stable and should predict something. IQ passes both tests, with limits. In 1932 almost every 11-year-old in Scotland took the same test, and Ian Deary's team later retested survivors. Between age 11 and age 77 the correlation was .63 (.73 corrected for the narrow range of survivors) among 101 people retested on the same test (Deary and colleagues, 2000). Between 11 and 90 it was .54 (.67 corrected) among 106 people (Deary, Pattie and Starr, 2013). Those are strong but not perfect: stable enough to be a real trait, loose enough that the score is not fixed for life.

On prediction, a study of more than 70,000 English children (Deary, Strand, Smith and Fernandes, 2007) found that g at age 11 correlated .81 with a latent measure of exam results across 25 subjects at 16. That figure is corrected for measurement error, so raw score-to-grade correlations are lower, but g accounted for 58.6% of the variation in maths, 48% in English and 18.1% in art and design. On health, a 2011 meta-analysis of 16 studies covering 1,107,022 people and 22,453 deaths (Calvin and colleagues, International Journal of Epidemiology) found that one standard deviation higher cognitive score went with a 24% lower risk of death (95% confidence interval 23 to 25), over follow-ups of 17 to 69 years. Our piece on childhood IQ scores and later health covers the argument about why.

Is IQ pseudoscience? What the best technical critiques say

The strongest technical critiques attack the explanation, not the measurement. Cosma Shalizi's 2007 essay "g, a statistical myth" argues that positive correlations among tests do not prove a single underlying cause: a model in which each test samples many independent abilities produces the same correlations, and factor analysis cannot tell the two apart. Shalizi does not claim tests are useless; he accepts that they predict grades and salaries, less than advertised. Han van der Maas and colleagues (Psychological Review, 2006) proposed a mutualism model in which cognitive processes support each other during development, producing the pattern of positive correlations without a single g. It does not deny the correlations either; it explains them. Neither critique shows that IQ scores measure nothing. Both ask what the scores are a measure of, a debate our explainer on the g factor lays out.

What do critics get right about IQ?

• A score is an estimate, not a fixed quantity. On well-normed adult tests the standard error is commonly around 2 to 3 points, so a 95% interval runs roughly 5 to 6 points either side (see the margin of error explainer)
• Norms expire. A 2015 meta-analysis of 271 samples, about 4 million people in 31 countries from 1909 to 2013, found full-scale gains of about 2.8 points per decade (Pietschnig and Voracek), and Norwegian conscript data show scores peaking for men born around 1975 and then falling, including within families (Bratsberg and Rogeberg, 2018). Our piece on why IQ norms expire explains the mechanics
• IQ does not capture everything that matters. In the English study, g explained 58.6% of the variation in maths results and only 18.1% in art and design
• Hiring validity was overstated. Sackett and colleagues revised the validity of many selection tools downward in 2022, cognitive tests included, though cognitive ability still predicts job performance (see what a work test score predicts)
• The history of misuse is real. IQ testing was used to justify eugenic policy in the early twentieth century, a story told in our post on eugenics and the history of IQ testing

What do critics get wrong about IQ?

• "IQ predicts nothing." It correlates .81 (error-corrected) with exam results at 16 and goes with a 24% lower risk of death per standard deviation
• "IQ is only test-taking skill." Test-taking skill does not obviously explain why a score is linked to survival decades later. Adjusting for adult social class cut the mortality link by 34% and adjusting for education by 54% in the 2011 meta-analysis, but neither removed it
• "Multiple intelligences replaced IQ." A 2006 critical review by Lynn Waterhouse found that multiple intelligences, the Mozart effect and emotional intelligence lack adequate empirical support. See our posts on multiple intelligences and brain myths about intelligence

Defenders get things wrong too. A score of .54 to .63 across decades means IQ is not fixed, and no test covers everything people mean by "smart"; our post on whether intelligence is limited to IQ takes that up. One popular shortcut, that IQ is just pattern recognition, also fails when you audit the items on a real test (see is IQ just pattern recognition).

What about Nassim Taleb's "pseudoscientific swindle" essay?

The most-shared modern critique is Nassim Nicholas Taleb's 2019 essay "IQ is largely a pseudoscientific swindle," published on Medium. As reposted, it makes four main claims: that IQ mostly detects the low end, meaning learning difficulties; that it explains only a small share of performance in real tasks; that the field mishandles correlation when data have fat tails; and that the score fails some formal requirements of a measurement scale. The first overlaps with a real historical point, that intelligence testing began as a way to identify children who needed help (see our history of intelligence testing). On the second, the disagreement is about how much prediction counts as enough: the correlations above run from modest to substantial depending on the outcome, and a 24% lower risk of death per standard deviation is not trivial. The statistical objections are the ones to test against the literature, and they are part of why claims about g are more contested than the raw correlations.

So is IQ meaningless, fake or a scam? A quick scorecard

• "IQ is meaningless": does not hold. It is stable across decades and linked to exam results and survival
• "IQ is fake or a scam": does not hold. It is a standardized measurement with published norms and stated error margins
• "IQ is pseudoscience": partly. The measurement is sound; the single-cause reading of g is legitimately contested
• "IQ is fixed and complete": does not hold. Scores drift with the norm group and g explains between 18% and 59% of the variation in exam results, depending on the subject
• "One score tells you everything": does not hold. A score is an estimate within about 5 to 6 points

What should you do with an IQ score, then?

Treat it as an estimate with a range, compared against a norm group and read alongside other information. Our explainer on IQ test accuracy and the posts asking whether IQ tests are reliable measures of intelligence and whether IQ tests really work go through what a score can and cannot bear. To see where any figure falls on the population scale, the IQ percentile calculator converts it to a percentile, how IQ tests work explains the norming, and if you want a baseline of your own you can take the IQ Metrics IQ test.

The bottom line: IQ is not debunked, but the version of it that says one number is fixed, complete and caused by a single factor is. What survives is a stable, modestly to strongly predictive measurement with a margin of error and an expiry date on its norms.

Common questions