Heritability of Intelligence: Twin Studies and Beyond

The heritability of intelligence is one of the most replicated findings in psychology — and one of the most consistently misunderstood. Six decades of twin and adoption studies, recently joined by genome-wide association studies on samples in the millions, converge on the conclusion that genetic differences explain a substantial fraction of why two adults have different IQ scores. The harder questions are how that fraction is estimated, why it changes across the lifespan, and what it does (and does not) mean for any individual.

What does “heritability” actually mean?

Heritability is a population statistic, not a personal one. It expresses the proportion of variation in a trait attributable to genetic differences in the population studied. A heritability of 0.60 for adult IQ means about 60% of the variance in IQ scores in that population reflects genetic differences between people; it does not mean 60% of any individual’s IQ comes from genes, and it does not imply that intelligence is fixed at birth.

Heritability is also context-dependent. Where inequality is high, environmental variance is large and heritability estimates fall; where everyone has adequate nutrition, education, and stimulation, environmental variance shrinks and heritability rises. The same trait in two different populations can have different heritability values without anything genetic having changed.

How twin and adoption studies estimate it

The classical method partitions trait variance into three components: additive genetic (A), shared environment (C), and non-shared environment plus measurement error (E). Monozygotic (MZ) twins share approximately 100% of their genome; dizygotic (DZ) twins share approximately 50%. If a trait is more similar in MZ than DZ twin pairs, the difference is attributed to genetic influence.

The most rigorous test is twins reared apart from infancy. Bouchard, Lykken, McGue, Segal, and Tellegen’s (1990) Minnesota Study of Twins Reared Apart followed identical twins separated at birth and reunited in adulthood. The IQ correlation between separated identical twins was approximately 0.75 — only modestly lower than the 0.86 typically found for identical twins reared together — while non-twin siblings reared together showed correlations of about 0.47. The pattern is exactly what additive genetic transmission predicts.

Polderman, Benyamin, de Leeuw, and colleagues’ (2015) meta-analysis of fifty years of twin studies — covering 17,804 traits across 14.5 million twin pairs — produced the most comprehensive estimate to date. For cognitive ability the heritability estimate sits at approximately 0.50 across all ages, with the additional finding that for the great majority of behavioural traits the additive genetic model fits the data adequately, without need for non-additive or epistatic terms.

The Wilson effect: heritability rises with age

One of the more counterintuitive findings in behavioural genetics is that the heritability of intelligence increases across the lifespan. Haworth and colleagues’ (2010) analysis of six twin studies (11,000 pairs across four countries) reported heritability of general cognitive ability rising linearly from approximately 41% at age 9, to 55% at age 12, to 66% at age 17. Subsequent work has placed adult heritability estimates at 70–80%.

This pattern, known as the Wilson effect, is the opposite of what most people expect. The naive intuition is that as life accumulates, environment matters more and genetic influence becomes diluted. The data say the reverse. The leading explanation is gene-environment correlation: as children gain autonomy, they increasingly select environments (peer groups, educational paths, hobbies, careers) that match and amplify their genetic predispositions. A child with strong verbal ability reads more, which improves verbal ability further, which leads to deeper engagement with verbal material. Genes shape environment, environment shapes phenotype, and the cumulative effect over decades is that adult IQ scores reflect genetic differences more cleanly than child IQ scores do.

The new genetics: polygenic scores and GWAS

Twin studies tell us how much genetic differences collectively matter. Genome-wide association studies (GWAS) attempt to identify the specific variants involved. The picture that has emerged is what Plomin and von Stumm (2018) termed “the new genetics of intelligence”: cognitive ability is influenced not by a handful of large-effect genes but by thousands of common variants each with a tiny individual effect.

The largest published intelligence GWAS to date have identified hundreds of genome-wide significant loci. Polygenic scores constructed from these GWAS results currently predict roughly 10–15% of variance in cognitive performance — a fraction of the 50–80% heritability estimate from twin studies, but a substantial advance over earlier candidate-gene work, which produced almost nothing replicable. The gap between twin-derived heritability and GWAS-derived prediction is partly explained by rare variants, gene-gene interactions, and population stratification, all of which polygenic scores do not yet capture well.

Polygenic scores already predict educational attainment, processing speed, and reasoning performance well enough to be useful as a research tool. They also illustrate the additive picture: each individual variant contributes only a small fraction of a percent to a person’s score, and no single “gene for intelligence” has been or will be discovered.

Genetic and environmental influences operate at different cognitive levels

Jiang, Sun, Yuan, Jiang, and Wan’s (2024) twin study in Cell Reports added a structural finding to the heritability picture: genetic and environmental influences do not operate uniformly across the cognitive hierarchy. They distinguished first-order abilities (basic perceptual and processing skills) from second-order abilities (higher-level reasoning and metacognition), and reported that first-order abilities showed strong genetic influence while second-order abilities were more sensitive to shared environmental factors.

This complements an older finding: general intelligence (g) is more heritable than specific abilities. Verbal fluency, spatial reasoning, and processing speed are each somewhat heritable individually, but the highest heritability sits at the level of the latent g-factor that loads on all of them. The practical implication is that environmental interventions can shift specific abilities (vocabulary, knowledge, particular skills) more readily than they shift the underlying processing capacity that g indexes — consistent with the schooling-effect literature, where additional years of education raise crystallized abilities more than fluid reasoning.

What heritability does not mean

Several inferences that look natural from “intelligence is 70% heritable” are wrong, and the misinterpretations fuel both genetic determinism and overcorrecting social-constructionist denial.

Heritability does not mean immutability. Phenylketonuria (PKU) is a fully genetic disorder that historically would have been “heritable” for cognitive impairment — until the dietary intervention that prevents it became standard, after which the heritability of PKU-related cognitive deficits collapsed. Heritability tells you about the current population, not what is possible under different conditions.

Heritability does not apply between groups. A within-population variance estimate cannot be extrapolated to differences in mean IQ between populations — that is a separate empirical question with different and well-known confounders.

Heritability does not exclude environmental intervention. The Flynn effect — population-level IQ gains of roughly 3 points per decade across most of the 20th century — happened in populations where heritability estimates were already high. Heritability and malleability coexist.

Gene-environment interplay rather than additivity

The cleanest framing of the current evidence is not nature versus nurture but nature through nurture. Genes shape which environments people are exposed to (through parental behaviour, peer selection, educational choices, occupational sorting) and how they respond to environments once exposed. The same enriched environment will produce different gains in two children with different genetic predispositions; the same genetic profile will produce different outcomes in different environments. Cogn-IQ’s genes-and-environment overview walks through this interplay accessibly, including how twin and adoption study designs partition variance under realistic assumptions.

Epigenetic mechanisms — modifications that change gene expression without altering DNA sequence — provide one molecular pathway through which environment shapes the readout of the same genome. The science is young but the framework it offers is exactly the gene-environment interplay the heritability literature implies.

What this means in practice

For individuals, heritability is largely irrelevant: it does not tell you what your IQ is, what it could become, or how to think about whether and how IQ can be increased. For policy and interpretation, the heritability literature delivers a clear message: cognitive ability has a substantial genetic component that is unlikely to be argued away, and environmental conditions remain a powerful modifier of what that genetic potential expresses as. Genetic determinism and environmental determinism are both wrong; the correct answer is that genes and environment are deeply entangled, and treating either as the “real” cause of intelligence misses what the data actually show.

Frequently asked questions

How heritable is intelligence?

In adults, the heritability of general cognitive ability is approximately 50–80% across multiple meta-analyses (Polderman et al., 2015; Plomin & Deary, 2015). The estimate rises across the lifespan, from about 41% at age 9 to 66–80% in adulthood (Haworth et al., 2010). The range reflects different populations, age groups, and cognitive tests; the overall picture is robust.

Does that mean intelligence is determined by genes?

No. Heritability describes population variance, not individual destiny. Highly heritable traits can be entirely environmentally caused under different conditions (the PKU example), and high heritability is fully compatible with environmental interventions producing measurable gains, as the Flynn effect and the schooling-effect literature both demonstrate.

What is the Wilson effect?

The finding that the heritability of intelligence rises across childhood, adolescence, and into adulthood. The leading explanation is gene-environment correlation: people increasingly select and shape environments that match their genetic predispositions, amplifying the effect of those predispositions on the adult phenotype.

Are there specific “intelligence genes”?

Not really. Modern genome-wide association studies (Plomin & von Stumm, 2018) show that intelligence is influenced by thousands of common variants, each with a tiny individual effect. Polygenic scores aggregating these variants currently predict roughly 10–15% of variance in cognitive performance — substantial as a research tool, but a long way from twin-derived heritability and a long way from “the gene for IQ.”

Can heritability differ between groups?

Yes, because heritability is population-specific. A trait can be highly heritable in one population (where environments are roughly equal) and only modestly heritable in another (where environmental conditions vary widely). This is also why heritability estimates within a population do not extrapolate to differences in mean trait values between populations — that is a different empirical question with different and well-known methodological pitfalls.