What affects IQ, and by how much

Each row gives the best single estimate we found, the kind of study behind it and how much weight we put on it. Strong means large meta-analyses or natural experiments that point the same way. Moderate means good evidence with a real dispute about size or cause. Weak or none means well-designed studies find little or nothing.^{[1, 2, 3, 4, 5, 6, 7, 8]}

The strongest evidence that IQ can be raised comes from schooling. Ritchie and Tucker-Drob (2018) pooled three kinds of natural experiment: studies that control for earlier IQ, changes to the school-leaving age, and children born just either side of a school-entry cut-off. All three found that an extra year of school raised IQ, by roughly 1 to 5 points depending on the design, about 3.4 on average, and the gains were still there in later life.

What schooling raises is still debated. It may build reasoning itself or mainly the knowledge and test habits IQ tests reward. Either way, the result undercuts the idea that IQ is fixed by the end of childhood. It is also one reason average scores rose through the 20th century, the Flynn effect. Age matters too, in a different way: see at what age IQ is highest.^[1]

Iodine. The thyroid needs iodine to make the hormones that build a baby's brain. In iodine-deficient regions, children score markedly lower: about 12 points in a meta-analysis of Chinese studies (Qian and colleagues, 2005) and about 13.5 in an earlier international review (Bleichrodt and Born, 1994), with a later review of under-fives putting the gap at 7 to 10 points. These are comparisons of deficient and sufficient areas, so they describe populations, not individuals, and they do not mean extra iodine helps people who already get enough.

Lead. In a pooled analysis of seven cohorts, Lanphear and colleagues (2005) linked a rise in children's blood lead from 2.4 to 30 µg/dL with about 6.9 IQ points lower, and found the steepest losses at the low end: about 3.9 points between 2.4 and 10 µg/dL. There was no safe threshold in their data. Removing lead from petrol and paint is one of the clearest public-health gains for intelligence on record.^{[9, 2, 10]}

Kendler and colleagues (2015) compared Swedish brothers where one was adopted into another family and the other was raised by the biological parents. At conscription, at 18 to 20, the adopted brothers scored about 4.4 points higher, and every step up in the adoptive parents' education was worth more. Same genes on average, different homes, a measurable difference. That fits with heritability estimates: high heritability does not mean the environment cannot move the average.

Early-childhood programmes are the frustrating case. They reliably raise scores while they run. Protzko's meta-analysis of 39 randomised trials (2015) found the advantage then fades over the following years, because the children who received the programme lose ground, not because others catch up. That is not a case against such programmes, which can help in other ways, but a programme that promises a permanent IQ boost is over-promising.^{[11, 4]}

Breastfed children score about 3.4 points higher on average across 17 studies (Horta and colleagues, 2015), and about 2.6 points in studies that adjust for the mother's IQ. The authors read this as causal. Der, Batty and Deary (2006), working with 5,475 children in a US survey, found that almost all of a 4-point raw gap disappeared once the mother's intelligence was taken into account, leaving about half a point. Mothers with higher IQs are more likely to breastfeed, and that alone can produce most of the gap. Our reading: any direct effect is small.^{[12, 3]}

Sleep is about performance, not underlying ability. In a meta-analysis of 70 studies of short-term total sleep deprivation, Lim and Dinges (2010) found large effects on simple attention and much smaller effects on reasoning accuracy. A tired person loses focus and makes careless slips, and on a timed test that costs points. Nobody has shown that sleeping more raises IQ in a well-rested person, but testing after a bad night will understate it. The guide to online test accuracy covers other test-day factors.^[5]

Brain-training apps. Melby-Lervåg, Redick and Hulme (2016) pooled 87 publications on working-memory training. People got better at the trained tasks and similar ones, but against an active control group there was no convincing gain on intelligence tests or other "far transfer" measures.

Music lessons. Sala and Gobet (2020) analysed 54 studies of nearly 7,000 children. Small benefits appeared only in weaker studies without random allocation or with a do-nothing control; in the better studies the effect was zero. Learn an instrument for its own sake. More popular claims are checked on our IQ myths page.

The Mozart effect. The 1993 report that listening to a Mozart sonata briefly improved spatial reasoning proved hard to replicate. A 2010 meta-analysis found little evidence left for any specific effect, and noticeably larger effects in studies by the original lab.

Practice on IQ-style items does raise scores, by familiarity rather than ability. Retest effects in cognitive tests are well documented and level off after about three attempts (Scharfen, Peters and Holling, 2018). That is why a second attempt at the same test is not a fair measure of change; why IQ scores differ explains how much movement is normal.^{[6, 7, 13]}