Genetic and environmental influences on adult intelligence and special mental abilities

Human Biology, Apr 1998 by Bouchard, Thomas J Jr

Heritability Estimate from Numerous Kinships: Texas Adoption Study. The Texas Adoption Study (Horn et al. 1982) is a longitudinal adoption study that makes use of 14 kinship pairings, some of which involve adoptive relations and some of which involve biological relations, because adoptive families often have their own biological children as well as adopted children. The Texas study provided some of the data points for the adult unrelated individuals reared together (see Figure 2). The most recent data collected used the Beta test, which also had been given to the parents 10 years earlier and to all the children in the follow-up study. The earlier analysis of data from the Texas Adoption Study, collected when the children were young and when different tests were used for both parents and children, yielded several estimates: a true score variance (corrected for unreliability) of 0.38 for additive variance (narrow heritability), a shared environmental influence of 0.19, a geneenvironment correlation of about 0.10, and a nonshared environmental effect of 0.33 (Loehlin 1979; Loehlin et al.1989). The more recent analysis (Loehlin et al. 1997) yielded a narrow heritability for true score variance (0.776), no gene by environment correlation, and virtually no shared family environmental influence.

The Texas Adoption Study as a whole shows the same decrease in shared environmental influence with age as the studies for unrelated individuals reared together and the studies for MZ and DZ twins reared together.

Sex Effects. Turner (1996) attracted considerable attention by suggesting that genes coding for intelligence evolved on the X chromosome. The evidence that there are some genes on the X chromosome that influence intelligence is indeed growing, and it is well known that the fragile X site and other specific genes (Thapar, Gottesman et al. 1994) increase the proportion of males in the lower tail of the IQ distribution. The suspicion that there are more males in the upper tail of the IQ distribution has some support as well (Halpern 1997; Hedges and Nowell 1995), but no genetic mechanism to explain it has been proposed. Nevertheless, for the ordinary range of intelligence there is virtually no difference in the magnitude of IQ correlations for samesex and opposite-sex familial correlations (Bouchard and McGue 1981), demonstrating that most of the quantitative trait loci (previously called polygenes) lie on the autosomes, not on the X chromosomes.

Maternal Environmental Effects. Recently, Devlin et al. (1997) reanalyzed a subset of the kin correlations originally compiled by Bouchard and McGue (1981). Devlin claimed that twin maternal effects account for 20% and sibling maternal effects account for 5% of the IQ variance in this database. With this maternal parameter in place, Devlin's model estimated the additive genetic variance at 34% and the nonadditive genetic variance at 15% (48% total genetic variance). This model estimated the shared environmental parameter to be 17%. Devlin's analysis challenges the view expounded here that there are strong age effects and that the heritability of IQ increases with age, with common environmental influence fading out in adulthood. Devlin et al. (1997) purported to have tested this hypothesis, and although their data were not inconsistent with the hypothesis, they claimed that their maternal hypothesis fitted the data better. Curiously, the database did not contain unrelated individuals reared together. Devlin's preferred model would predict that this group should have a correlation of 0.17 both in childhood and in adulthood, a prediction falsified by the data presented in Figure 2. The correlation for unrelated individuals reared together of 0.32 (based on childhood data) has been cited for years to support important common environmental influences on IQ. The Devlin model requires that this evidence be sacrificed in favor of maternal effects. The question is, Which is more probable?