Educability-and-Group-Differences-1973-by-Arthur-Robert-Jensen

Race Differences in Intelligence 191
tenth of a normal distribution with an overall mean of 100 and a
variance of 168-75. The Negroes and whites have the same
distributions of genotypes (column means) in this model. The 12
rows represent environmental categories, which have different
distributions for whites and Negroes, based on the 12 socioeconomic
categories of the 1960 U.S. Census (see Figure 7.3). Since the
non-genetic variance is 1 —A2 = 0-25, the environmental component
of IQ variance will be 0-25 x 225 = 56-25, with an overall
mean of 100. The mean IQs of the 12 rows are obtained by
assuming a normal distribution of environmental effects on IQ
(an assumption which is supported by twin studies) and dividing
this normal distribution (with mean = 100, variance = 56-25)
into 12 unequal parts with frequencies corresponding to the
Census data. Thus, from the 10 x 12 grid’s row means and column
means, assuming additive effects of genetic and environmental
factors, one can obtain the mean IQ within each of the 120 cells.
The resultant overall mean IQs for whites and Negroes are 100
and 91-26, respectively. Since this difference falls short of the
15-points difference that needs to be accounted for, Light and
Smith introduce the assumption of 1 percent ‘interaction’ variance
into the model.24 This means that low genotypic IQs are more
depressed by poor environment than higher genotypic IQs. Light
and Smith had a computer find the optimal ‘malicious’ allocation
(as they call it) of Negro genotypes to environmental categories
in producing this 1 percent interaction effect so as to maximize
the overall Negro-white IQ difference. This brings the white and
Negro means to 100 and 86-81, respectively, which is very realistic
in terms of the empirical values of white and Negro mean IQs.
Light and Smith went further and assumed a 10 percent interaction
due to ‘malicious’ allocation of poor genotypes in poor
environments; the optimal allocation of 10 percent interaction
effects by the computer to maximize the white-Negro difference
resulted in mean IQs of 100 and 82-59. Light and Smith state:
‘We may therefore conclude that with an interaction component
of variance somewhere between 0-01 and 0-10, the black mean IQ
may be expected to be approximately 85, even though blacks are
distributed identically with whites over the genetic categories’
(1969, p. 498). Thus, the claim is made that this environmental
allocation model can accept the available heritability estimates for
IQ and, by assuming a small interaction component of between