12 IQ myths, checked against the research

Each myth is taken in turn below, with the evidence. "Half true" means the popular version overstates something real.

Half true. Rank order is remarkably stable. When Scottish people tested at 11 in 1932 were retested at about 80 on the same test, the correlation was about .66, and a 2014 review by Ian Deary concluded that roughly half of the individual differences in intelligence are stable across most of life. But half is not all. Schooling raises scores by about 1 to 5 points a year (myth 3 has more), and individual teenagers can move by 20 points over four years (Ramsden and colleagues, 2011). See what affects IQ.^{[1, 2]}

False. Scores measured in youth predict real outcomes years later. In Tarmo Strenze's 2007 meta-analysis of long-term studies, intelligence correlated .56 with years of education and .43 with occupational level, with a weaker link to income. Those are large effects by social-science standards, though far from destiny: plenty of people out-earn and out-achieve their score. The IQ and income page covers the detail, and the g factor explains why such different tests agree.^[3]

False. Heritability describes how much of the variation in a particular population tracks genetic differences. It says nothing about whether the average can move. Swedish brothers adopted into other homes scored about 4.4 points higher than their brothers raised by the biological parents (Kendler and colleagues, 2015), and an extra year of schooling adds roughly 3 points. Height is highly heritable too, and it rose across the 20th century with better food. Our heritability page explains the difference.^[4]

False. You get better at the games. A 2016 meta-analysis of 87 publications on working-memory training found gains on the trained and closely related tasks, but no convincing improvement on intelligence tests once training was compared with an active control group (Melby-Lervåg, Redick and Hulme). Enjoy the apps; do not expect a higher IQ.

False. A 1993 study reported a short-lived boost on spatial tasks after listening to a Mozart sonata. A 2010 meta-analysis (Pietschnig, Voracek and Formann) found little evidence left for any specific effect once publication bias was corrected, and noticeably larger effects in studies from the original lab than from anyone else.^[5]

False in well-designed studies. Sala and Gobet (2020) analysed 54 studies of nearly 7,000 children. Small gains appeared only in studies that did not randomise children or compared them with doing nothing. Against an active alternative such as dance or sport, the effect on cognitive and academic skills was zero. Music is worth learning for itself.^[6]

Unknown. There is no record that Albert Einstein ever took an IQ test, and modern adult tests did not exist for most of his life. The 160 figure is a later estimate repeated until it sounded like a fact. Our Einstein profile gives the estimate as a range and explains what it is based on.

False. A 2018 meta-analysis of 18 studies with full IQ scores found the difference between handedness groups negligible, and if anything slightly in favour of right-handers, about one point (Ntolka and Papadatou-Pastou). The difference disappeared when the largest study was left out.

Tiny effect only. In three large national samples from the US, UK and Germany, measured intelligence declined slightly from first-born to later-born siblings (Rohrer, Egloff and Schmukle, 2015). The lead author called the effect reliable in large samples but "barely meaningful on the individual level". Birth order tells you almost nothing about any one sibling.^[7]

Half true. Speed and fluid reasoning peak early, from the late teens to the twenties. Vocabulary and general knowledge keep rising into middle age and beyond. A large online study (Hartshorne and Germine, 2015) found different abilities peaking at very different ages, with no single age at which you are cleverest. IQ scores themselves do not decline, because each age is compared with its own norm group. The age guide has the timeline.^[8]

Usually false. Every score has a margin of error, typically several points either side on a professional test and more on a short one. Different tests also use different norms, ages and content. A gap of 10 points between two tests a year apart is ordinary. The why IQ scores differ guide and the confidence interval calculator show how much is noise.

Half true at best. A well-built online reasoning test can give a useful estimate. None can give a clinical IQ, because it has no supervised administration and, usually, no census-matched norm group. Be most suspicious of tests that hand nearly everyone a score above 120. Our guide to online IQ test accuracy explains what to look for, including in ours.