In the Scottish Mental Surveys of 1932 and 1947, all Scottish schoolchildren 11 years of age at the time were administered the Moray House Test, a standardized group test of verbal, numerical, and spatial reasoning ability. The analysis and follow-up of these results have led to a rich series of papers by Ian Deary and his colleagues, and I highly recommend using the search engines at your local university library to peruse the fruits of this project. The Scottish Mental Surveys are a terrific source of information on greater male variance, relative stability from childhood to old age, effects on health and mortality, and many other important facets of mental abilities.

I wanted to draw attention to this neat study, published in Behavior Genetics a few months ago, that estimated the heritability of IQ using the twins in both Survey cohorts. The estimates from both cohorts are consistent and indicate that 70 percent of the variance in IQ in these populations is accounted for by genetic variation. That is a tad high for populations assayed at this age, but still no surprise. The clever part of this analysis is that the investigators were able to attain these estimates without knowing the zygosities of the twins! Here is how they did it. One variable that was recorded for all twins was whether they were of the same sex or opposite sex. Obviously, twins of the opposite sex are dizygotic. On the assumption that opposite-sex dizygotic twins are no more frequent than same-sex dizygotic twins, the respective proportions of twinships that are monozygotic and dizygotic are readily obtained. Now, the respective distributions of within-pair differences for MZ and DZ twins for a heritable trait are different in theory, with the latter showing larger variance. The application of the likelihood-ratio test to the empirical distributions of within-twinship IQ differences in the Scottish Mental Surveys overwhelmingly rejected the null hypothesis of a single normal distribution in favor of the alternative hypothesis of two distributions. At this point the investigators were able to use maximum likelihood, incorporating the MZ and DZ proportions as weights, to estimate the variances of these two distributions. For an example of a similar statistical disaggregation of a composite distribution, see this piece by our friend La Griffe du Lion. These variances then yielded the respective intraclass correlations for MZ (~0.90) and DZ (~0.50) twins. Using the traditional "Falconer formula," subtracting 0.50 from 0.90 and multiplying by 2 would yield an estimated heritability of 0.80, but here the investigators used a more sophisticated model-fitting approach and somehow got their confidence interval of (0.58, 0.83).

There was no evidence for non-additive genetic effects. It is my own view, however, that dominance deviations account for a nontrivial proportion of variance in g and that studies of children obscure this because of the distorting effects of common evironment. Allow DZ twins to age, shed the effects of common environment, and become less similar in IQ (which is in fact observed in the regrettably few samples of mature or reared-apart DZ twins), and then non-additive genetic effects will become visible to analysis. Also, I recall that the subject of height and IQ has come up more than once in comment threads. Here, Deary and his colleagues found as usual a phenotypic correlation between height and IQ but no genetic correlation. This is another item of evidence against an intrinsic relationship between height and IQ.

The results of this study are by no means earth-shattering. But I thought it was still cool.