Heritability Analyses of IQ Scores: Science or Numerology?

See allHide authors and affiliations

Science  29 Mar 1974:
Vol. 183, Issue 4131, pp. 1259-1266
DOI: 10.1126/science.183.4131.1259


Estimates of IQ heritability are subject to a variety of systematic errors.

The IQ scores themselves contain uncontrollable, systematic errors of unknown magnitude. These arise because IQ scores, unlike conventional physical and biological measurements, have a purely instrumental definition. The effects of these errors are apparent in the very large discrepancies among IQ correlations measured by different investigators.

Genotype-environment correlations, whose effects can sometimes be minimized, if not wholly eliminated, in experiments with plants and animals, are nearly always important in human populations. The absence of significant effects arising from genotype-environment correlations is a necessary condition for the applicability of conventional heritability analysis to phenotypically plastic traits. When this condition fails, no quantitative inferences about heritability can be drawn from measured phenotypic variances and covariances, except under special conditions that are unlikely to be satisfied by phenotypically plastic traits in human populations.

Inadequate understanding of the precise environmental factors relevant to the development of specific behavioral traits is an important source of systematic errors, as is the inability to allow adequately for the effects of assortative mating and gene-gene interaction.

Systematic cultural differences and differences in psychological environment among races and among sociocco-nomic groups vitiate any attempt to draw from IQ data meaningful inferences about genetic differences.

Estimates based on phenotypic correlations between separated monozygotic twins—usually considered to be the most reliable kind of estimates—are vitiated by systematic errors inherent in IQ tests, by the presence of genotype-environment correlation, and by the lack of detailed understanding of environmental factors relevant to the development of behavioral traits. Other kinds of estimates are beset, in addition, by systematic errors arising from incomplete allowance for the effects of assortative mating and from gene-gene interactions. The only potentially useful data are phenotypic correlations between unrelated foster children reared together, which could, in principle, yield lower limits for e2. Available data indicate that, for unrelated foster children reared together, the broad heritability (h2) may lie between 0.0 and 0.5. This estimate does not apply to populations composed of children reared by their biological parents or by near relatives. For such populations the heritability of IQ remains undefined.

The only data that might yield meaningful estimates ot narrow heritability are phenotypic correlations between half-sibs reared in statistically independent environments. No useful data of this kind are available.

Intervention studies like Heber's Milwaukee Project afford an alternative and comparatively direct way of studying the plasticity of cognitive and other behavioral traits in human populations. Results obtained so far strongly suggest that the development of cognitive skills is highly sensitive to variations in environmental factors.

These conclusions have three obvious implications for the broader issues mentioned at the beginning of this article.

1) Published analyses of IQ data provide no support whatever for Jensen's thesis that inequalities in cognitive performance are due largely to genetic differences. As Lewontin (8) has clearly shown, the value of the broad heritability of IQ is in any case only marginally relevant to this question. I have argued that conventional estimates of the broad heritability of IQ are invalid and that the only data on which potentially valid estimates might be based are consistent with a broad heritability of less than 0.5. On the other hand, intervention studies, if their findings prove to be replicable, would directly establish that, under suitable conditions, the offspring of parents whose cognitive skills are so poorly developed as to exclude them from all but the most menial occupations can achieve what are regarded as distinctly high levels of cognitive performance. Thus, despite the fact that children differ suibstantially in cognitive aptitudes and appetites, and despite the very high probability that these differences have a substantial genetic component, available scientific evidence strongly suggests that environmental factors are responsible for the failure of children not suffering from specific neurological disorders to achieve adequate levels of cognitive performance.

2) Under prevailing social conditions, no valid inferences can be drawn from IQ data concerning systematic genetic differences among races or socioeconomic groups. Research along present lines directed toward this end—whatever its ethical status—is scientifically worthless.

3) Since there are no suitable data for estimating the narrow heritability of IQ, it seems pointless to speculate about the prospects for a hereditary meritocracy based on IQ.