Twin Studies and IQ Heritability: How Much of Intelligence Is Genetic?

About half of the variation in IQ scores within a population tracks genetic differences, and the share rises with age. In the largest twin study of general cognitive ability, 11,000 pairs from four countries, heritability was 41% at age 9, 55% at 12 and 66% at 17. Plomin and Deary's 2015 review puts it at about 20% in infancy, 40% in adolescence, 60% in adulthood and perhaps 80% in later adulthood. Identical twins reared together correlate about .86 on IQ, fraternal twins .60, and identical twins reared apart .72 in the pooled 1981 review (only 65 pairs) and .69 to .78 in the Minnesota study. Heritability is a statistic about a population, not a share of any one person's IQ, and DNA-based estimates come out lower than twin estimates. This guide sets out the correlations, the age trend, the adoption and DNA evidence, and what none of it means.

How much of intelligence is genetic?

The answer depends on age, sample and method, and the published ranges are wide. The most cited single study is Haworth and colleagues (2010, Molecular Psychiatry), which pooled six twin studies from four countries. Its model-fitting estimates of heritability, with the share due to the shared family environment beside them:

• Age 9: heritability 41%, shared environment 33% (identical twins correlate .74, fraternal .53)
• Age 12: heritability 55%, shared environment 18% (identical .73, fraternal .46)
• Age 17: heritability 66%, shared environment 16% (identical .82, fraternal .48)

Plomin and Deary's review (2015) describes the same linear rise, from about 20% in infancy to perhaps 80% in later adulthood, falling back to about 60% after age 80. The largest all-trait meta-analysis, Polderman and colleagues (2015, Nature Genetics; 14,558,903 twin pairs, 17,804 traits), found an average heritability of 49% across all traits, and for higher-level cognitive functions twin correlations of .710 for identical and .441 for fraternal twins. Applying Falconer's formula, doubling the gap between the two, gives 0.54 for cognition; that is our arithmetic, since the paper's text does not print a figure for cognition alone. Bouchard and colleagues' 1990 Minnesota study of twins reared apart concluded that genetic factors account for approximately 70% of the variance in IQ among its mostly adult twins. Put together, a defensible summary is 40% in childhood, 50 to 60% on average and up to about 70 to 80% in adults. For a multiple-choice quiz that offers 25, 40, 75 or 90 per cent, 75 is the only option inside the published adult range, which is our judgment and not a printed answer.

Which twins have more similar IQ scores, identical or fraternal?

Identical twins, in every large data set. In the 1981 review the correlation was .86 for identical and .60 for fraternal twins reared together; in Haworth's pooled sample it was .76 against .49; in Polderman's cognition data .71 against .44. So the statement that identical and fraternal twins differ significantly in their average IQ correlations is true, and it is the basis of the heritability estimate: if genes matter for IQ, pairs who share all of their DNA should end up more alike than pairs who share about half. Identical twins develop from one fertilised egg and share close to all of their DNA; fraternal twins share, on average, about half of the segregating genes.

What do the classic family correlations show?

The foundation is the 1981 review by Thomas Bouchard and Matthew McGue in Science, which pooled 111 studies and 526 familial correlations covering 113,942 pairings. Its weighted average correlations for IQ were:

• Identical twins reared together: .86 (34 correlations, 4,672 pairs)
• Identical twins reared apart: .72 (3 correlations, only 65 pairs, with the disputed Burt data left out)
• Fraternal twins reared together: .60 (41 correlations, 5,546 pairs)
• Ordinary siblings reared together: .47 (69 correlations, 26,473 pairs)
• Siblings reared apart: .24 (203 pairs from two investigations)
• Unrelated adopted siblings reared together: .34 (6 correlations, 369 pairs)

Identical twins reared apart, at .72, are more alike than fraternal twins reared together at .60 and than unrelated adopted siblings in the same home at .34. The Minnesota Study of Twins Reared Apart, begun in 1979, then tested more than 100 sets of separated twins and triplets over a week of assessment. Its reared-apart correlation depends on the test: .69 on the WAIS (48 pairs), .78 on a Raven and Mill Hill combination (42 pairs), and .75 as the mean of its multiple tests. The figure often printed as "the" reared-apart correlation, .75, is therefore a multi-test mean and not the pooled value, and the pooled value rests on a small number of pairs.

Does the heritability of IQ change with age?

Yes, and it rises. Bouchard (2013) proposed calling the pattern "the Wilson effect" after Ronald Wilson, whose Louisville Twin Study gave the first clear evidence that the heritability of IQ increases with age. Haworth's numbers above show it, and Briley and Tucker-Drob's 2013 meta-analysis of 16 articles from 11 samples (11,500 twin and sibling pairs measured at least twice between 6 months and 18 years) found that innovative genetic influences dominate in early childhood but quickly fade, while amplified genetic influences account for the rising heritability after about age 8. The usual explanation is gene-environment correlation: as children grow up they increasingly select, change and even create their own experiences in part based on their genetic leanings. Bouchard puts the shared family environment at about .10 by young adulthood, small but not zero.

What do adoption studies add?

Adoption separates genes from rearing in another way. The Texas Adoption Project, reported by Horn in 1983 for 300 adoptive families, found that adopted children resemble their biological mothers more than the adoptive parents who reared them; a small subset of the oldest adopted children did not, and the author treated the idea of declining gene influence with caution. A 2021 reanalysis by Loehlin and colleagues of the Texas and Colorado projects gives the standardised link between an adopted child's IQ and the birth mother's: 0.25 at the Texas start and 0.34 at follow-up, and 0.18 at age 4 rising to 0.27 at age 16 in Colorado, though not smoothly. The link to the adoptive home's socioeconomic status stayed near zero in both. Our blog guide to adoption and IQ covers the broader evidence.

What do DNA studies say, and why are they lower than twin studies?

Molecular studies explain less than twin studies do, which researchers call the missing-heritability problem. A genome-wide study of intelligence by Savage and colleagues (2018, Nature Genetics) of 269,867 people found 205 associated genomic regions, and the education study of Lee and colleagues (2018, about 1.1 million people) built polygenic scores that explain 7 to 10% of the variance in cognitive performance, against twin estimates of 50% and more. Part of the gap is technical, since DNA arrays capture only common variants. Part of it may be that population studies mix a person's own genes with family effects. Howe and colleagues (2022, Nature Genetics) compared siblings within families and found the estimated effect shrank by 22% for cognitive ability and 47% for educational attainment, and that SNP heritability of cognitive ability fell from 0.24 to 0.14. Plomin and Deary give one example of the gap: a DNA-based estimate of 0.35 against a twin estimate of 0.47 in the same study.

Does family income change how heritable IQ is?

One famous result says yes. Turkheimer and colleagues (2003, Psychological Science) studied 7-year-old twins and found that in impoverished families about 60% of IQ variance was due to the shared environment and the contribution of genes was close to zero, while in affluent families the result was almost exactly the reverse. A 2016 meta-analysis by Tucker-Drob and Bates of 14 studies and 24,926 pairs found the effect in the United States (an interaction of +.074) but not in Western Europe and Australia, where it was zero or reversed (-.027). The Texas and Colorado adoption data of Loehlin and colleagues showed no substantial interaction. Our guide to socioeconomic status and IQ covers the environmental side.

What heritability does not mean

Heritability is not a measure of how genetic one person's IQ is. An individual score does not split into a genetic portion and an environmental portion any more than the area of a rectangle splits into a length portion and a width portion. It is a population statistic: the proportion of variation in scores across a specific group, in the environments that group actually experienced, that statistically tracks genetic differences. Change the population or the range of environments and the number can change with it, as the age and family-income findings show. It also says nothing about differences between separate populations, because a statistic built from within-group comparisons cannot explain a between-group gap; the same limit undermines attempts to explain cross-national score comparisons with single-country twin samples.

Do twin studies rest on a shaky assumption?

The method assumes that identical and fraternal twins experience similarly similar environments, the equal environments assumption, so that extra similarity between identical twins can be credited to genes. It has been tested directly. Conley and colleagues (2013) studied twins whose zygosity was misclassified and found that results largely supported the assumption; Derks, Dolan and Boomsma (2006) found no detectable violation for spatial ability; and Felson (2014) concluded that the assumption is probably not strictly valid for most outcomes but that the resulting bias is likely modest. The Minnesota design avoids the problem by studying twins raised apart. A newer paper, published online on 12 October 2026, estimates the heritability of IQ with a nuclear twin family design from 660,365 twin, parent-child and spousal pairs in more than 15 countries; we could read only its abstract, so we do not report its estimates.

What is this research useful for?

The honest use is narrow. Within the populations studied, genetic relatedness tracks IQ similarity more closely than a shared household does, which rules out the simplest environmental story, that upbringing alone explains why family members' scores resemble each other. It does not tell you what caused any one person's score, and it leaves real room for everything that was not genetically shared: shared environment explains a third of the variance at age 9 in Haworth's sample. Related questions get their own pages: our blog guide to whether IQ is genetic and to inheritance from mother or father, and our notes on whether twins have lower IQ than singletons, birth order and C-section and IQ. To see where a score falls on the population curve, use the IQ bell curve guide.

The number that survives all the caveats is the comparison, not a percentage: identical twins end up more alike than fraternal twins, reared-apart identical twins end up more alike than reared-together unrelated siblings, and heritability grows with age. To see your own result on the standard scale, take the IQ Metrics IQ test.

Common questions