The Harvard Educational Review 1969 debate

Here’s the complete collection of papers published in the Harvard Educational Review (HER) in 1969 in reply to Jensen’s famous paper. For Jensen’s reflections on the events, see the introduction to his 1972 book Genetics and Education.

These articles are also available on the publisher’s website: issue 39-1, 39-2, 39-3.

Jensen’s own reply to the first batch of criticism is well worth reading (“Reducing the heredity-environment uncertainty”). Of particular note is that Jensen advocated the use of future genetic tools to estimate individual racial admixture for testing competing models:

[…] Brazziel is quite correct in noting, for example, that the Negro population of the United States, like the white, is very far from being genetically or racially homogeneous. In fact, it is doubtful that any babies of pure African descent are being born in the United States today, unless they are born to African exchange students. But Africans, too, are genetically heterogeneous. A number of studies based on the differential frequencies of various blood groups in African and Caucasian populations have shown that, on the average, persons socially classified as American Negroes now have an admixture of 20 to 30 per cent Caucasian genes (Reed, 1969). The percentage of Caucasian admixture varies greatly in various regions of the country, going from an average of below 10% in some Southern states to above 25% in some Northern states. These figures can be estimated with considerable precision in large population samples, depending on the number of different blood groups and other genetic polymorphisms one is able to take into account. With these methods individuals, too, can be categorized by proportion of Negro-Caucasian admixture on a probabilistic basis. Possibly these same genetical techniques could provide a basis for more refined and accurate tests of hypotheses concerning racial differences in ability patterns. Since skin color is but poorly correlated with the percentage of Caucasian admixture, and because it may have social-environmental consequences, it could be statistically controlled in studies of the correlation between Negro-Caucasian admixture and measures of psychological characteristics. Environmental differences would not be an obstacle, since there is a wide range of racial admixtures in any large sample from highly similar environments. In fact, where there are half-siblings, intra-family comparisons might be possible, thereby controlling a host of environmental family-background factors. Other quite different approaches are possible, or a number of methods used in combination. The finding that electroencephalographic visually- evoked potentials are related to IQ means that intelligence might be measured on a physiological level, and such a measure would come closer than anything we now have to a true culture-free test. Studies of foster children of one race or social class adopted by parents of another is one more avenue. Such are only a few of the possible suggestions. Geneticists should be able to evaluate these and come up with better ideas. Collaborative research by geneticists and behavioral scientists could surely advance our scientific knowledge of racial and social class differences. To argue to the contrary, it seems to me, is to claim the impotence of a scientific approach and of human ingenuity, an attitude which is clearly contradicted by our great advances in other fields of inquiry. If the heredity-environment uncertainty is unresolvable in the sense that, say, perpetual motion is impossible, we should at least not be satisfied until we have discovered precisely the laws of nature which make it so.

As far as I am aware, this is one of the first mentions of “admixture analysis” – using direct genetic methods to estimate the causal effect of individual ancestry.

In Hirsch et al 1967, Behavior Genetic Analysis, there is a chapter on the topic of racial differences by Spuhler & Lindzey; after extensive discussion, they conclude that the only possible methods are “common garden” adoption studies rearing members of different races in as similar environments as possible (rejected as infeasible), and analyses of simple Mendelian traits where the penetrance is sufficiently high as to make any non-genetic explanation untenable and allowing estimation of gene frequencies in each race. If Hirsch, Spuhler, or Lindzey were aware of admixture analysis, no mention is made. Thoday 1969, “Limitations to genetic comparison of populations”, concurs, stating that it is “impossible” to investigate racial differences and may well remain impossible “for the foreseeable future”, and that “only transplant experiments could solve this problem”. It’s also interesting to note that James Crow, who participated in the responses to Jensen, was interviewed in 2006 and showed no apparent awareness of Jensen’s suggestion, Shockley’s papers, or later development of admixture analysis:

6) When your commentary on Arthur Jensen’s infamous Harvard Educational Review article on the inheritance of IQ and racial differences was published in 1969, did you have any inkling that the issues raised by Jensen would remain largely unresolved over thirty-five years later? What kind of evidence do you think would decide these issues one way or the other?

I did not expect the issues to be resolved soon, for there were no new methods that promised be more informative. Of course, the structure of DNA had been discovered, but the powerful methods now available had not yet been developed. I think further identification of individual genes, usually by molecular methods, and a combination of statistical and molecular methods are pointing the way toward a solution. I don’t expect racial differences to be either entirely genetic or entirely environmental, but of course I don’t know the relative amount; it is likely to be different for different traits and different human groups. 

This suggests that if admixture analysis had already been developed (perhaps by R.A. Fisher), the human behavior geneticists were unaware.

William Shockley appears to have independently invented admixture analysis when he proposed using it to investigate black-white IQ differences (also drawing on Reed’s work), especially using pairs of siblings in order to control for all background characteristics. His first relevant paper was in 1966, and it comes up in several of his other publications. See:

It would not be surprising if Shockley & Jensen had corresponded on this topic before or after, given the close timing of the various papers and typical publication lags, so it’s unclear which, if either, has priority. Jensen’s archives list the existence of Shockley correspondence, but no dates. (Educability mentions a 1969 letter from Shockley to Jensen, but in a different context.)

In his 1973 book Educability, on pg225 Jensen discusses it further under “Genetic Biochemical Polymorphisms and IQ”:

The use of genetic polymorphisms in the blood for researching this problem was suggested independently by Shockley (1966) [“Possible transfer of metallurgical and astrono­mical approaches to problem of environment versus ethnic heredity”], a physicist noted for the invention of the transistor, and by Heston, a psychiatric geneticist noted for his research on the genetics of schizophrenia (e.g., Heston, 1966). Essentially, Heston has proposed obtaining correlations between skin color (measured with a reflectance spectrophotometer on the underside of the upper arm) and mental test scores, on the one hand, and between percentage of Caucasian admixture (based on a dozen or more blood groups) and intelligence scores, on the other, and then testing the hypo­thesis that the non-visible index of Caucasian admixture (blood groups) correlates more highly with intelligence than the visible index of skin color. If the blood groups measure of M (proportion of Caucasian genes) correlates more highly with, say, IQ than does skin color, the hypothesis of a racial genetic difference in intelligence would be supported. Blood groups are a more reliable basis than skin color for estimating Caucasian admixture because more genes are involved and because blood groups, being non-visible, do not enter into mate selection.

Heston, with quantitative geneticist Oscar Kempthorne and statistician James Hickman, worked out a method for statistically estimating the proportion of Caucasian genes in individual Negroes, which is a more complex problem than the estimation of M, the proportion of Caucasian genes in a particular hybrid Negro population.

…Since variation in skin pigmentation, because of its social-environmental consequences, is controlled in this research design, any direct biochemical connection between degree of skin pigmentation and intelligence must be either ruled out or, if such a relationship is established, its consequences for the present design must be assessed. The possibility of a biochemical connection between skin pigmentation and intelligence is not totally unlikely in view of the biochemical relation between melanins, which are responsible for pigmentation, and some of the neural transmitter substances in the brain. The skin and the cerebral cortex both arise from the ectoderm in the development of the embryo and share some of the same biochemical processes.

If there is some correlation between amount of Caucasian ancestry of a Negro child and the cultural-environmental influences acting upon him, it could be argued that this approach does not sufficiently control or ‘read through’ environmental determinants of intelligence to allow any definitive conclusion. However, if the correlation between proportion of Caucasian genes and intelligence showed up substantially even in environmentally quite homogeneous samples, such evidence would surely strengthen a genetic hypothesis of racial intelligence differences. A further control, but one that would require the screening of very large samples for the optimal blood groups, would involve maternal half-siblings. Pairs of half-sibs would be selected to differ in their proportions of Caucasian genes, as estimated by the methods just described, and would also be measured for skin color and IQ. In other words, we would have matched controls for both prenatal and postnatal environmental effects. It is hard to imagine any reason why, on the average, the environments should favor the child with the more Caucasian genes when the independent effects of skin color and other visible characteristics are statistically controlled. It might be possible to find environments which favor the more Negroid characteristics, so that finding a positive correlation between Caucasian genes and IQ in such circumstances would be even more compelling.

Since studies of this kind have not yet been done, there is no good basis for speculating about their probable outcome when and if they are carried out.^2 Shockley (1971b)^3 [“Hardy-Weinberg law generalized to estimate hybrid variance for Negro populations and reduce racial aspects of the environment-heredity uncertainty”] has noted that California Negroes have twice as high a percentage of their genes from Caucasian ancestors as do Georgia Negroes and that the IQ difference between Negroes in California and in Georgia (estimated from army pre-induction test results) is about 10 points. But this observation can carry little conviction, since differences in the cultural and educational conditions of Negroes in Georgia and in California are completely confounded with differences in Caucasian gene frequencies as possible causes of the IQ difference.^4

… Footnotes:

2. The feasibility of this kind of study at the present time is not universally unquestioned among geneticists. In a personal communication, geneticist Peter L. Workman writes:

“Since American Negroes do not comprise an equilibrium population, morphological characters constituting an African appearance segregate together with African genes. Holding constant external appearance (skin color, lip breadth, etc.) might also partial out most of the relevant information. Further, Heston’s method doesn’t account for the non-equilibrium structure. Thus, although the experimental idea is a good one, and MacLean and I will present the appropriate methods in print shortly, I am very skeptical that it could be done at this time. We need more African data [on blood group frequencies].”

3. Shockley (1970b) has also suggested dividing a large Negro group, such as the total enrolment in an all-Negro school or college, into two halves on the basis of some assessment of intellectual ability (e.g., above or below the median in IQ, college entrance tests, gradepoint average, etc.) and then determining the proportion of the lower and higher groups showing the Duffy Fy^a ‘Caucasian gene’. This test must assume no correlation between Fy^a and socially visible features which could affect IQ and no correlation between Fy^a and IQ in the white population. The feasibility of this proposal has been questioned because the American Negro population probably has not yet reached genetic (Hardy-Weinberg) equilibrium, so there would probably be a great deal of genetic linkage of visible African morphological features and blood polymorphisms. Controlling or partialling out the visible racial characteristics would therefore also partial out some of the IQ variance associated with the blood groups used as an index of the degree of African-Caucasian admixture. If it could be argued that the socially visible African features did not themselves constitute an ‘environmental’ disadvantage that might adversely affect mental development or performance on intelligence tests, the proposal would have merit despite genetic disequilibrium in American Negro racial hybrids.
4. If a 10 percent admixture of Caucasian genes raises the IQ of Negro-white hybrids by 10 points, one would have to assume a great deal of genetic interaction or some kind of hybrid vigor to explain why a 100 percent admixture of Caucasian genes would raise the IQ only about 15 points. If an admixture of Caucasian genes had the large effect suggested by Shockley, one should expect there to be much greater genetic variance among Negroes, and there is no evidence of this. If anything, the evidence is for less genetic variance in the Negro population. The 10-points IQ difference between Georgia and California Negroes would therefore most reasonably be attributed mainly to selective migration and environmental differences.

In footnote 2, Workman implies that based on Heston’s ideas, he and MacLean had been working out admixture analysis. The work in question appears to have been published as a series of 3 papers in 1973:

  1. “Genetic studies on hybrid populations I. Individual estimates of ancestry and their relation to quantitative traits”
  2. “Genetic studies on hybrid populations II. Estimation of the distribution of ancestry”
  3. “Genetic studies on hybrid populations. III. Blood pressure in an American black community”
    1. A statistically significant regression of diastolic blood pressure on the percentage of African admixture was found in a sample of persons from an American black community after removal of variation due to sex, age, and obesity. It is suggested that this regression is due to genetic racial factors of African origin, but confounding of admixture with environmental factors cannot be excluded.

None of them mention Heston, Shockley, or Jensen in connection with the idea, despite apparently chronologically preceding Workman & MacLean by several years and Workman directly corresponding with Jensen about the idea and apparently crediting it to Heston (calling it “Heston’s method”). Jensen, presumably, agrees that it is Heston’s method, as he doesn’t correct the quotation and presumably was supplying the description in the first place to Workman.

In other sources, Reed 1973, discussing number of genes necessary for the design, also appears to credit it to Workman & MacLean (with an acknowledgement to Heston for helpful comments and no mention of Jensen/Shockley), although Scarr et al 1977, in an early small application of admixture analysis to the race-IQ question (Scarr, S., Pakstis, A. J., Katz, S. H., & Barker, W. B. (1977). “Absence of a relationship between degree of white ancestry and intellectual skills within a black population”. Human Genetics, 39(1), 69–86. doi:10.1007/bf00273154), does trace the suggestion to Jensen’s Educability. Scarr 1971, “Unknowns in the IQ Equation”, also sketches the logic of the admixture analysis, but doesn’t mention a source; given the 1971 date, it could not have been Jensen’s 1973 Educability nor Workman & MacLean, but it could have been any of Jensen or Shockley’s article or however Heston was distributing his idea – Heston is again thanked in the acknowledges of Scarr 1971.

So, then, who invented admixture analysis on the individual level for inferring causality of racial genetic differences? Was it Jensen, Shockley, Heston, or someone else entirely?

The simplest narrative consistent with the dates & quotations is that Laurence Heston invented it sometime around 1960-1966, as the logical extrapolation of Reed and others’ work on estimating admixture fraction of population from a few markers/traits (a very active field of inquiry at the time), extending it down to the individual level by using many markers, but he’s too afraid to publish on it given how explosive most of the applications are; he tells Jensen by personal communication, and likewise when preprints get circulated to him, tells Scarr and others about it. Jensen communicates it further, attributing it to Heston. MacLean and Workman take the idea from Jensen and run with it after his HER response sketched the idea, plagiarizing it by refusing to cite Jensen/Shockley/Heston as the source. Shockley reinvents it around 1966, but being Shockley, no one pays him the least bit of attention aside from Jensen, who mentions Shockley’s short paper as “independently” inventing the idea from Heston, implying Heston has priority.