Estimation of the limits of heritability of traits by comparison of monozygotic and dizygotic twins.

This paper has three aims: (1) to present a new formula for extracting heritability estimates from twin data; (2) to show the results of the application of the formula to data from past studies of the heritability of intelligence, scholastic achievement, personality traits, and physical characteristics; and (3) to urge that heritability estimates be obtained in all large-scale educational testing programs, in the standardization of intelligence, aptitude, and educational achievement tests, and in the Selective Service and Armed Forces qualification tests. Previous Estimates of Heritability.-Although the twin method in itself does not provide sufficient information for testing detailed genetic models, it provides both the most efficient and the least ambiguous basis for an over-all estimate of heritability of quantitative traits.1 Heritability (h2) is defined here as the proportion of phenotypic variance attributable to genotypic variance, i.e., h2 = aG2/ap2. The comparison of monozygotic (MZ) twins reared together and dizygotic (DZ) twins reared together is much more feasible and has been a much more common practice than the study of MZ twins reared apart. MZ twins reared apart are rare and difficult to find.2' 3 Estimating heritability from MZ twins reared apart has the one advantage that it presents little theoretical difficulty, provided one can assume zero correlation between the relevant environmental effects acting on the separated twins, in which case h2 = rMz, the intraclass correlation between MZ twins. The prevailing method of estimating heritability from MZ and DZ twins has been by means of the H index devised by Holzinger.14 That Holzinger's H index is not a satisfactory estimate of h2 is now generally recognized in behavior genetics, but the precise nature of the inadequacy of the H index and the problem of estimating h2 from MZ and DZ twin data have remained conceptually obscure.' 4-6 Nichols7 proposed an improvement on the H index, called the HR index, but it, too, is unsatisfactory as an index of h2. One serious criticism ofH and HR is that one is not a monotonic function of the other, and neither is a monotonic function of h2. Vandenberg1 has proposed using F (the variance ratio) as a test of the significance of uWZ2/ awMz2 (DZ within-pair variance/MZ within-pair variance), but this is as faulty as an index of heritability as the H index itself, since F is a linear function of H.8 Determining the variance ratio F, however, is an essential step prior to computing h2; if F is not statistically significant, h2 cannot be presumed to differ significantly from zero.'9 10