Arthur R. Jensen (2000) "biological Determinism" as an Ideological Buzz-word. Psycoloquy: 11(021) Intelligence g Factor (33)

Volume: 11 (next, prev) Issue: 021 (next, prev) Article: 33 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 11(021): "biological Determinism" as an Ideological Buzz-word

Reply to Raymond on Jensen on Intelligence-g-Factor

Arthur R. Jensen
Educational Psychology
School of Education
University of California
Berkeley, CA 94720-1670


The term "biological determinism" has become a buzzword in the IQ controversy for those who adhere to what Sandra Scarr once called "naive environmentalism" -- the notion that genetic factors play no part in individual or group differences in behavioural traits, which are assumed to result entirely from environmental inequalities imposed by the capitalist economic system and its attendant social injustices. Criticisms of "The g Factor" (Jensen, 1998; 1999) based on this implicit ideology accordingly miss the mark; most of the substantive points in Raymond's (1999) critique are either incorrect or irrelevant to my book's argument.


behavior genetics, cognitive modelling, evoked potentials, evolutionary psychology, factor analysis, g factor, heritability, individual differences, intelligence, IQ, neurometrics, psychometrics, psychophyiology, skills, Spearman, statistics
1. Raymond's (1999) abstract sets off on the wrong foot by stating that "The g Factor" (Jensen 1998, 1999) presents no evidence for the external validity of g. In psychometrics, external validity refers to a psychometrically derived score's correlation with variables outside the psychometric domain, or non-psychometric variables. In Chapter 9 ("The Practical Validity of g"), I review a great many external "real-life" correlates of IQ and of the g factor in particular. It is overwhelmingly the case that g is the chief active ingredient in the practical predictive validity of psychometric test batteries, such as the Wechsler scales, the Armed Services Vocational Aptitude Battery, The Differential Aptitude Tests, and the General Aptitude Test Battery, to name a few. The increments of predictive validity added by various psychometric factors independent of g are minute by comparison. If anyone still has any doubts that g has external validity, I urge that they read Chapter 9 of Jensen (1998) (and for more evidence, Chapter 8 of Jensen, 1980). Also, in Chapter 8 I point out that the main source of the correlations between untimed psychometric tests and the chronometric measures obtained from a variety of elementary cognitive tasks (ECTs) is almost entirely attributable to the g factor.

2. Spreading the g variance throughout the various other factors -- which can be achieved by orthogonal rotation of the factor axes (as is usually performed by the Varimax algorithm) to get what Raymond (1999, par.#2) refers to as Thurstone's "simple structure" -- has at least three disadvantages, one technical, one theoretical, and one practical.

3. (1) Technically, Thurstone's simple structure criterion can't even be approximated to a satisfactory degree when the analyzed correlation matrix harbors a large general factor. With orthogonal rotation of first-order factors, the so-called hyperspace gets filled up with the substantial loadings that consist of the bits of g that are scattered throughout the first-order factors. Because of this failure to achieve a good simple structure, Thurstone introduced oblique (i.e., correlated) factors. The correlations between these yield a general factor, either at the second or third order (or stratum, in J.B. Carroll's [1993] terminology). The oblique first-order factors, after their g is extracted and placed at the second (or third) order of the factor hierarchy are then rendered orthogonal (uncorrelated) and usually meet Thurstone's simple structure criterion. But the extracted g factor is usually the largest linear component of variance in the whole analysis, and in some batteries g is larger than that of all of the remaining orthogonal factors combined.

4. (2) The theoretical objection to ignoring the general factor, g, is that no valid or compelling theoretical argument has been put forth that the general factor of a correlation matrix in the abilities domain g doesn't deserve to be represented in the analysis of the matrix as well as any other significant factors. An orthogonalized hierarchical factor analysis cannot show a significant general factor if it is not inherent in the correlation matrix. If a general factor does exist, why should we not know its size (i.e., percent of variance it accounts for), its loadings in the observed variables, and its external validity as compared with that of any other independent factors revealed by the analysis?

5. (3) The practical disadvantage of dispersing g across all the first-order factors is that, since g is responsible for the largest share of tests' practical predictive validity, one must include every test or every extracted factor as a predictor variable in a multiple regression equation. A g factor score or just the total score on the whole test battery would serve almost as well, so little do the non-g factors usually contribute to the prediction (Thorndike, 1985; Jensen 1998, Chapter 9).

6. The g factor itself is not a theory of intelligence. It is a latent trait reflecting correlated individual differences in a wide variety of tests involving different task demands and cognitive processes.

7. At present there is no evidence that the level of g has changed in the population over the last few decades. Test scores per se have increased, at least in the industrialized world, but whether these increments represent g, other factors, or test specificity has yet to be established. One way to do this is by the method of correlated vectors (Jensen, 1998, Appendix B). If it is g that has changed, then, in a battery of tests, we should find a significant positive correlation between the tests' g loadings and the standardized increments in the mean scores on those tests obtained in separate but comparable samples tested a decade or more apart. This would be a worthy study and would help to throw some light on the well-known Flynn Effect (i.e., the secular increase in IQ over the past half-century or so).

8. The method of correlated vectors proves that g is correlated with such non-psychometric variables as heritability of tests, evoked potentials, reaction times in ECTs, as well as many types of practical predictive validity. Since IQ correlates more with g than with any other factors, it is a generally safe asumption that g is a large component of the correlation between IQ and other variables. In those cases where we could correlate some external variable with both g and IQ (derived from the same test battery), we have found that both these variables show highly similar correlations. The method of correlated vectors shows that the degree to which the various subtests in a battery are g loaded predicts their respective degrees of correlation with some external variable (e.g., as the habituation of the evoked potential) or, stated more succinctly, the vector of the subtests' g loadings is significantly correlated with the vector of the subtests' correlation with the external variable.

9. Gould (1981) himself is most responsible for the actual "error and bias" (borrowing Raymond's words) in reporting the relationship of IQ and of racial differences to head size (i.e., measures of cranial capacity, direct measurements of autopsied brains, and in vivo MRI measurements of the brain.) (See the review of the revised edition of Gould's book by Rushton [1997].) The preponderance of the evidence leaves little doubt that brain size is correlated with IQ and with racial differences. So these count as items of evidence, although not in themselves conclusive, that are at best consistent with a biological theory of racial differences in g. What is harder to demonstrate, and is still uncertain, is whether the racial differences in brain size explain (i.e., are causally related to) the racial differences in g.

10. The fact that midgets, with small head-size on a par with that of normal 2 to 3 year old children, show no differences from the general population in IQ indicates that large brain size per se, though correlated about +.40 with IQ in the general population, is neither necessary nor sufficient for low, average, or high IQ. The best evidence we have of a functional relationship between brain size and ability comes from animal studies, in which animals that have been selectively bred for a "cognitive" ability (e.g., maze learning) also show an increase in brain size, although brain size was not the variable under selection. (It would be nice to have a study in which genetic selection is fo brain size itself and learning rate is the dependent variable.) In humans, probably the best we have to offer is strong evidence of a within-family (full-siblings) correlation between brain-size and IQ, suggesting a pleiotropic (and hence a biological, and plausibly causal) connection between these variables (Jensen & Johnson, 1994).

11. Mutations and genetic drift contribute to genetic variance (i.e., different allelic frequencies) within and between populations, but their effects with respect to group differences contribute less between-groups variance than does selection. There is a connection between selection for a genetic trait, dominance variance, and inbreeding depression. Inbreeding depression, which results from an increase in the homozygosity of unfavorable recessive genes, does not occur in the absence of dominant genes; consequently some proportion of the total genetic variance is attributable to dominance deviations.

12. Selection, natural or artificial, acts more rapidly on changes in the frequency of dominant alleles, eliminating the less favorable and preserving the more favorable, while recessive alleles respond much more slowly to selection, because their effects are covered in the phenotypes by the favorable effects of the dominant alleles at the same gene loci. Hence, in the course of selection, there is a more rapid increase in favorable alleles, which are usually dominant, than in unfavorable alleles, which are usually recessive. Balanced polymorphisms in highly polygenic traits such as height and intelligence necessarily involve the preservation of both dominant and recessive genes. The presence of genetic dominance variance indicated by the magnitude of inbreeding depression for that trait is therefore an indicator of past selection for the trait. The theory of these relationships was originally formulated within a systematic and general paradigm by R.A. Fisher (1930).

13. It is indeed the case that the correlation between the mother's education and child's IQ is largely a genetic correlation (in childhood) and is an entirely genetic one when the offspring reaches late adolescence. When the genetic connection between mother and child is absent, as in adoptive parents and their adopted children, there is zero correlation between the adoptive parent's education and the adoptee's IQ in adolescence. I won't attempt to recapitulate here all of the studies on racial differences in g and IQ, the Flynn Effect, maternal or prenatal, perinatal, and early childhood microenvironmental factors that affect the biological basis of mental development; careful readers will find answer all of Raymond's comments therein.

14. Raymond (1999, par.10) notes a passage on p. 568 that he thinks "is not an impression that rings true." It refers to the statement that, since about 1980, American Blacks in many higher level occupations have been recruited from a somewhat lower range of IQ than were Whites in the same occupational category. This is well substantiated by studies cited in Jensen (1998, pp. 565-569). A recent study (Nyborg & Jensen, in press), based on a large data set collected by the United States Centre for Disease Control, shows the complement of this phenomenon, that is, Blacks who are above an IQ of approximately 100 and are statistically matched with Whites on g factor scores (derived from a battery of 19 diverse tests) have increasingly higher annual income than Whites at every higher level of g scores, and Blacks show increasingly higher job status than g-matched whites throughout the full range of g scores measured in this study (about IQ 80 and above). I wouldZZ agree with Raymond that this is a surprising finding and is indeed the opposite of the impression the general public may have.


Carroll, J.B. (1993). Human cognitive abilities: A survey of factor-analytic studies. Cambridge: Cambridge University Press.

Fisher, R. A. (1930). The genetical theory of natural selection. Oxford: Clarendon Press.

Gould, S.J. (1981). The mismeasure of man. New York: Norton.

Jensen, A.R. (1980). Bias in mental testing. New York: Free Press.

Jensen, A.R. (1998). The g factor: The science of mental ability. Westport, CT: Praeger.

Jensen, A.R. (1999). Precis of: "The g Factor: The Science of Mental Ability" PSYCOLOQUY 10(23). psyc.99.10.023.intelligence-g-factor.1.jensen

Jensen, A.R., & Johnson, F.W. (1994). Race and sex differences in head size and IQ. Intelligence, 18, 309-333.

Nyborg, H., & Jensen, A.R. (in press). Black-white differences on various psychometric tests: Spearman's hypothesis tested on American armed services veterans. Personality and Individual Differences, 28, 593-599.

Raymond, B. (1999). Biological determinism unwarranted. PSYCOLOQUY 10(57). psyc.99.10.057.intelligence-g-factor.7.raymond

Rushton, J.P. (1997). Race, intelligence, and the brain: The errors and omissions of the "revised" edition of S.J. Gould's "The mismeasure of man." [1996}. Personality and Individual Differences, 23, 169-180.

Thorndike, R.L. (1985). The central role of general ability in prediction. Multivariate Behavioral Research, 20, 241-254.

Volume: 11 (next, prev) Issue: 021 (next, prev) Article: 33 (next prev first) Alternate versions: ASCII Summary