Box 4 | The heritability of IQ controversy

From the following article:

Heritability in the genomics era — concepts and misconceptions

Peter M. Visscher, William G. Hill & Naomi R. Wray

Nature Reviews Genetics 9, 255-266 (April 2008)

doi:10.1038/nrg2322

Nowhere has the debate about nature and nurture been so controversial as in the study of mental ability in humans5, 90, 91. Controversies about the concept and use of intelligence quotient (IQ), a phenotypic measurement of relative performance on a series of mental ability tests, are manifold. They include: its definition ('intelligence is what intelligence tests measure'90); documented historical abuse relating to eugenics; inference about the cause of observed differences between ethnic groups (see Box 2); incorrect statistical inference from observational studies90; and disputed implications of IQ differences between individuals and groups on social and economic interventions92, 93. We will not discuss the uses and abuses of measures of cognitive ability, but we will point out that there is abundant empirical evidence that shows that IQ is a good predictor of outcomes in life, including educational attainment, income and health94. Controversy about IQ is by and large because of social, not scientific, reasons. Here, we focus on one point of controversy about IQ: its heritability.

Twins have been used in the majority of studies to estimate the heritability of IQ. These studies include twins that were separated at birth (adoption studies), monozygotic (MZ) and dizygotic (DZ) twins that were raised together, and a combination of these designs with additional siblings. There have also been numerous studies involving other relatives94. The empirical results are clear: MZ twins are substantially more similar in IQ than DZ twins, whether they are raised together or apart. Reported estimates of heritability for IQ from twin studies are remarkably consistent in the range of 0.5–0.8, across many age groups. The reported estimate of heritability for young (preschool) children is lower95 and estimates of heritability at old age are approximately 0.6–0.8 (Ref. 96).

A recurring criticism of estimates of heritability for IQ is that it is too high. A large heritability for IQ can be controversial because of the perceived implications that a person's or group's 'intelligence' cannot be changed by intervention strategies (this is an incorrect perception, see Box 2). Nevertheless, it is valid to question the correctness of all those twin and family studies. The issue is one of statistical inference. The reported resemblance between relatives for IQ, measured, for example, by correlation coefficients, is unequivocal. For example, across many studies, the average MZ and DZ correlation was 0.86 and 0.60, respectively, based on 4,672 MZ and 5,546 DZ twin pairs94. To estimate a narrow-sense or broad-sense heritability from these correlations (or directly from the raw data), assumptions have to be made about models of genetic and non-genetic causes of family resemblance. Some of these assumptions can be tested empirically, others cannot. For example, there is likely to be a correlation between genes and environments for IQ so that estimates of heritability might be overestimated5, 90. Another example is that if there are strong maternal (for example, in utero) effects on the IQ of twins, and these are larger for MZ twins than for DZ twins or siblings, then the estimate of heritability would be biased upwards because it is assumed (in common practice) that the resemblance due to common environmental factors are equal97. In one meta-analysis of a number of twin studies the modelling of maternal effects implied a narrow-sense heritability of only 0.3 and an estimate of broad-sense heritability of 0.5 (Ref. 98). Although much lower than 0.8, these estimates are still moderate to large when compared with measures of behavioural and other phenotypes in livestock species and natural populations. Therefore, we can conclude from the wealth of empirical data currently available that the resemblance between relatives is large and consistent with the hypothesis that a large proportion of the variation in IQ between individuals within a population is associated with additive genetic factors.