Jensen (1998) does not present any evidence for the external validity of g other than through biological correlations with IQ. Hence it is impossible to assess the validity of g in comparison with competing theories from this book. The author's interpretation of heritability estimates of IQ and of racial differences in IQ adoption studies ignores any potential maternal effects on IQ. His insistence that the primary cause of variation in IQ within and between races is genetic does not seem warranted by the evidence presented.
2. It is difficult to extract any concise definition of g itself from the book. The best was "g is created from the correlations among second order factors, whose residual variance consists of whatever variance they do not have in common with each other". My understanding of g and factor analysis is that it is an attempt to reduce the multidimensionality of a battery of mental test results to one dimension which best represents the common variance shared between tests. As such, it is a mathematical abstraction, and it is not possible on theoretical grounds to dismiss other theories of intelligence (e.g., Thurstone's simple structure) which represent an equal amount of information from the data on more than one dimension. Spearman's original method of tetrad differences for testing for g was abandoned precisely because it often demonstrated that it was difficult to represent mental tests on one dimension (see p. 75). A hierarchical analysis which correlates these first order factors (e.g. spatial, verbal, and memory first order factors), avoids this problem only because these factors also tend to be correlated.
3. Some inconsistencies also arise when g is used as a fundamental theory of intelligence; for example, the increase in IQ (a heavily g loaded parameter) over the last 50 years has been matched by a decrease in scholastic ability (another highly g loaded indicator; p. 322). Has g both increased and decreased simultaneously? Jensen blames this inconsistency on the use of different 'vehicles', yet there is no direct means of measuring g other than through 'vehicles' of one form or another. Jensen also indicates that g 'typically' accounts 'for a larger proportion of the total variance than any other factor and often accounts for more of the variance than all of the other factors combined' (p. 79). Exactly what 'typically' or 'often' represents is not discussed further, despite being fundamental to the justification of the method. Furthermore, in one cited African study on elementary cognitive tasks (p. 392) the largest common factor among tests is described as being possibly something other than g. How is it possible to make this distinction, unless the author is assuming an existence of g external to the test results?
4. An assertion that g has validity beyond a mathematical abstraction requires that it be connected to biological reality, and moreover associated with biology in such a way that it has increased explanatory power relative to competing theories. This connection to basic biology is based on correlations with "stature, head size, brain size, frequency of alpha brain waves, latency of evoked potentials, rate of brain glucose metabolism, and general health" (p. 137). First, correlations themselves prove no causal link between g and biology, and the separation (p. 130-143) of correlations into intrinsic (causal) and extrinsic (spurious) is simply not valid. Correlation with body size is dismissed by Jensen himself as not causal (p. 146). EEG is correlated with IQ; but IQ, even if it is strongly correlated with g, is not g, and we cannot assess whether alternative theories of intelligence describe this relationship better (e.g., whether higher g-loaded measures of mental ability are better correlated with EEG results). Cerebral glucose metabolism, head size, and nerve conduction are also only correlated with IQ. Thus, Jensen presents no evidence of causal links, nor any evidence that g is better correlated with basic biology than multi-dimensional theories of intelligence.
5. The study of brain size in relation to race has historically been controversial and riddled with error and bias (Gould, 1981). Figure 12.4 does nothing to relieve my admittedly pre-conceived scepticism concerning this issue. The high correlation coefficient is cited but not the significance value (p = 0.037, which is just significant). Moreover, the plotted values of cranial capacity are not those corrected for body size; the correlation is not significant (p > 0.05) when the corrected values are used. The method of deriving IQ/race medians from means is obscure, and no reason for preferring means over medians is given. The black IQ of 80 is surprisingly low given the American mean of 85. There may well be evidence for racial difference in brain size, but the best study of adults cited is not described as controlling for obvious but potentially confounding environmental effects such as alcohol use (Ho et al., 1980). There are well established differences in brain size between males and females, without significant differences in IQ, and this is accounted for by differences in neuronal packing density (p. 438). Since this factor has not been studied in relation to racial differences, sweeping causal conclusions seem premature.
6. Jensen's chapter on the heritability of g fails to address a recent meta-analysis of 212 IQ on this topic. That paper estimates that 20% of the covariance in twins IQ can be explained by maternal effects and that broad heritability is 48% (Devlin et al., 1997). Thus, the shared environment effects on twins reared apart are underestimated by Jensen to give inflated values of heritability. Furthermore, the meta-analysis supported a maternal effects model better than models which assume heritability increases with age. The types of maternal effects associated with IQ changes described in the literature include: dietary supplements (Harell et al., 1955) and PCBs (Jacobson & Jacobson, 1996). Moderate alcohol consumption of mothers can affect psychomotor development (Larroque et al., 1995); and polycyclic aromatic hydrocarbons are associated with reduced head circumference of newborns (Perera et al., 1999). It is not unreasonable to hypothesize that a more socially disadvantaged group will be subject to poorer diet, increased levels of environmental pollutants, as well as increased drug and alcohol use.
7. This section also includes some fundamental errors in evolutionary understanding. The assertion that "traits that show genetic dominance provide evidence that they have been subjected to natural selection" (p. 170) is not correct. Recessive traits, and purely additive traits can also be subject to natural selection, and dominant alleles may be selectively neutral. The fact that a trait is heritable and variable means that it is potentially subject to natural selection, but even this does not mean we can infer selection; other evolutionary forces such as mutation and genetic drift can be responsible for changes in gene frequency. In this light, the discussion relating dominance to hybrid vigour and inbreeding depression is misguided (p. 189-197).
8. Jensen's "default hypothesis" is that differences in white and black IQ scores are made up of genetic and environmental effects and that environmental effects are often small relative to genetic effects (e.g. pp. 177-175, 475, 476, 489). These conclusions seem best supported by estimates of heritability, discussed above, and from the Minnesota transracial adoption study. This adoption study has several flaws. Maternal effects were not considered as a serious possibility for the lower IQ of blacks relative to whites. The IQ of parents of adopted children was unknown. The mean age of adopted infants also differed between treatment groups. And most fundamentally, the parents of children were not selected randomly from the population. Infants of parents of varying socioeconomic status and race might give children up for adoption, or have them taken into care, for very different reasons. In contrast, Jensen dismisses data from a racial admixture study in Germany which found no significant or consistent differences in IQ with race, because parental IQ was unknown, and because white and black fathers were not randomly sampled, two faults shared with the Minnesota adoption study. The absence of differences in the German study is also consistent with a maternal effects hypothesis, since the mothers were all white.
9. Within the "g factor" there is abundant evidence for various environmental correlates of IQ, and for effects which may be specific to the black population. On p. 385 a study that controlled for socio-economic status reduced the g factor difference between American blacks and whites by 12 IQ points. On p. 513, "stereotype threat", a form of test anxiety, could account for a 5 IQ point difference between black and white college students. 12 plus 5 is 17, and the mean difference between blacks and whites in the US is only 15 IQ points. There are also two strong patterns which are highly suggestive of environmental effects, namely, the steady increase in IQ scores of 3 IQ points per decade in the last half century (p. 307), and the strong geographical gradient in IQ among blacks from south to north. For example, a group of black schoolchildren in rural Georgia has a mean IQ of 71, and Jensen's comments are "we would be hard put to find a more socially disadvantaged black community... anywhere in the United States". Whereas the black mean in Minnesota is given as 90. The most poorly founded conclusion in the entire book also concerns environmental effects, namely that 'mother's education alone account[s] for 13% of the childrens' IQ variance, but this is most likely a genetic effect' (p. 502). Given these potentially numerous and powerful effects, Jensen's insistence on the primacy of genes is surprising.
10. Overall, the book leaves an impression of biological determinism; the correlation between IQ and socioeconomic status is consistently presented as resulting from the causal effects of IQ. Jensen paints a picture of America as a country with perfect social mobility, where an absence of racial prejudice is effectively demonstrated by the overrepresentation of blacks in jobs of high socioeconomic status relative to their distribution of IQ (p. 568), despite the much lower correlation between IQ and socioeconomic status for blacks relative to whites (p. 358). This is not an impression that rings true.
Devlin, D., Daniels, M., & Roeder K., 1997. The heritability of IQ. Nature 388: 468-471.
Gould, S.J., 1981. The Mismeasure of Man. Penguin, London.
Harrell, R.F., Woodyard, E., & Gates, A.I., 1955. The effects of mothers diet on the intelligence of offspring. Teacher's College, New York.
Ho, K-c., Roessman, U., Stramfjord, J.V., & Monroe, G., 1980. Analysis of brain weight: II. Adult brain weight in relation to body height, weight , and surface area. Archives of Pathology and Laboratory Medicine 104:640-645.
Jacobson, J.L., & Jacobson, S.W., 1996. Intellectual impairment in children exposed to polychlorinated biphenyls in utero. New England Journal of Medicine 335:783-789.
Jensen, A. (1998) The g Factor: The Science of Mental Ability. Praeger
Jensen, A. (1999) Precis of: "The g Factor: The Science of Mental Ability" PSYCOLOQUY 10 (23). ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1999.volume.10/ psyc.99.10.023.intelligence-g-factor.1.jensen http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?10.023
Larroque, B., Kaminski, M., Dehaene, P., Subtil, P., Delfosse, M.J., & Querleu, D., 1995. Moderate prenatal alcohol exposure and psychomotor development at preschool age. American Journal of Public Health 85: 1654-1661.
Perera, F.P., Jedrychowski, W., Rauh, V., & Whyatt, R.M., 1999. Molecular epidemiological research on the effects of environmental pollutants on the fetus. Environmental Health Perspectives 107: 451-460.