Evidence for a low upper limit of heritability of mental test performance in a national sample of twins

Abstract

THE use of kinship correlations is a long established approach to the estimation of heritabilities (h²) and component variances of human phenotypes. Among the best known studies are those using monozygotic and dizygotic twin correlations, data from which are commonly interpreted to represent a substantial heritability (that is, about 0.8) of IQ and other mental test performances (see Jensen¹ for review, and Jinks and Fulker² for the empirical requirements of this approach). Some doubt has been cast, however, over the empirical sufficiency of the most quoted estimates. Among the doubts are those concerning biographies of subjects, testing procedures, and problems emerging from the use of varied and inappropriate tests and from the subjects being of widely different ages³. More recent estimates using a version of the identical–fraternal twin comparisons (but derived from a sample in which the zygosities were unknown) have arrived at a value for h² not significantly different from zero^4–6. The National Child Development Study (for details, see ref. 7), a longitudinal study of all the children in England, Scotland and Wales born in one week of March 1958, includes a nationally representative sample of twins about whom have been gathered considerable biographic, biometric, social and psychometric data (unpublished results). We have used this sample to investigate the broad heritability of performance on a general mental ability test.