Below is a good article by the famous anthropologist Margaret Mead written in 1924 and published in 1926 in The American Journal of Sociology. (A footnote says Mead had been overseas since 1924, presumably in Samoa.) Mead, the superstar of Franz Boas’s famous school of cultural anthropology at Columbia (alongside Ruth Benedict and black novelist Zora Neale Hurston), offered sensible methodological concerns about studies trying to correlate IQ and racial admixtures.
From very early in the history of IQ testing, it was observed that blacks averaged lower scores than whites. That quickly raised questions of causality: Nature? Nurture? Both?
One way to falsify the genetic theory of the IQ racial gap would be if there proved to be no positive correlation among self-identifying African Americans between IQ and the proportion of white ancestry. Finding a positive correlation wouldn’t prove that the race gap was at least partly genetic—there are always more complicated explanations that can be devised. But not finding a positive correlation would be highly embarrassing to the Nature camp.
We now finally have huge databases of ~10,000 volunteers, such as the Philadelphia Neurodevelopment study and the Adolescent Brain Cognitive Development, that by combining cognitive testing with DNA data on genetic ancestry allow us to correlate IQ and racial admixture better than the studies that Mead criticized a century ago.
Of course, they are immense hot potatoes. Doing these studies helped get tenured professor Bryan Pesta fired from Cleveland State U. But very little news coverage has been given to Pesta’s firing, perhaps because writing about it would require describing his research, which is unthinkable.
THE METHODOLOGY OF RACIAL TESTING: ITS SIGNIFICANCE FOR SOCIOLOGY
MARGARET MEAD
Barnard College
American Journal of Sociology, March 1926
ABSTRACT: A discussion of the methodology of racial intelligence testing is both pertinent and necessary. Three problems are involved: measurement of (i) the racial admixture factor; (2) the social status factor; (3) the linguistic disability factor. The methodology of each of these phases of the problem is discussed.
… Perhaps no results of experimental psychology have been utilized so widely and so un- critically as the results of intelligence testing, and particularly of the intelligence testing of different racial and nationality groups. In the discussion of race problems, a controversy so encumbered by worn-out dogmas and hot partisanships, this quantitative type of material was particularly welcome. The first research to be generally exploited was the army testing.
Lewis Terman introduced the IQ test to America just before the Great War with his Stanford-Binet Test. The written Alpha IQ test was given to draftees in 1917. But it was quickly discovered that quite a few American young men, especially in the hillbilly regions, were too illiterate to take the test. And in the northern cities, many immigrants were not familiar with English. An oral Beta test that used diagrams and mimes was ginned up for 1918.
The Alpha and Beta tests had their problems, but they generated a huge amount of data for academics to review after the war and thus greatly sped up the maturation of intelligence testing.
But here so many opponents of the resulting generalizations came forward with destructive criticism, and so many defenders of the tests carefully tried to warn the layman against unjustified conclusions, that the mischief became too public to be dangerous. No discussions today which pretend to scientific caution quote the army tests without many explicit reservations. But the criticisms of the army tests were in great part devoted to the deficiencies of all intelligence tests, of verbal tests, or of group tests as such.
Far less attention was devoted to the special problems inherent in racial and nationality testing. Thus, while the writer on general social problems has learned that the methodology of intelligence testing is still in swaddling clothes, he is not so conscious that a methodology adequate to deal with racial and nationality testing has not even been born. And, unwarned, he draws freely and uncritically upon the findings of special studies which appear from time to time in the scientific journals. Many of these studies appear to have avoided the pitfalls into which the army tests fell. They are often individual tests instead of group tests; the question of inequalities of education is ruled out when school groups are used as subjects; the selection argument does not seem to be so readily applicable when the subjects are all taken from the same city and often from the same public school. So these results come to be utilized in far other ways than their authors intended. Paradoxically enough, one of the experiments most often quoted in heated discussions is by the author of “On the Need for Caution in Establishing Race Norms.” …
The special problems involved in this type of testing are three in number: (i) the practicability and validity of attempts to equate test score and amount of racial admixture; (2) the effect which social status has on the results of such tests-the problem here is threefold: Does social status influence test score? Is it particularly influential in the case of immigrant groups or groups which suffer from social discrimination because of their race? What methods are adequate to evaluate the social status of the children so tested? (3) What effect does linguistic disability have on the test results? …
RACIAL ADMIXTURE On the face of it, no method of determining race differences in intelligence seems more promising than this attempt to equate test scores with the amount of Negro or Indian blood, as the case may be, when it is possible, as it often is in this country, to study such a mixed group. In such an instance it seems highly probable that language and social status might be controlled, and the effect of race as such isolated and studied. But here we are confronted with two difficulties: one technological, the other theoretical …
It’s taken close to a century for the technology to mature to accurately quantify racial admixture, but we now have it.
Garth and Hunter used the official records of the government schools. Subject to the accuracy of these genealogical records, this method is thorough and valid.
As a cultural anthropologist, Mead is sympathetic to Garth, because he’s using the cultural anthropology method of genealogy. Indeed, genealogy works well in some cultures where there is a lot of fascination with genealogical record keeping and extreme emphasis on female monogamy. For example, among the semi-East Asian–looking Hazara of Central Afghanistan, it’s common to find men who claim that Genghis Khan was their 34th great-grandfather. A 2003 genetic study found that they might be right.
On the other hand, genealogy is pretty woozy in other situations.
But here Garth admits that his results are quite indeterminate because the social conditions for his various groups were probably extremely varied before the children entered the government schools. Methodologically however, the procedure is sound. In his studies on the American Negro, Ferguson has made a less objective method the basis of his determination of amount of admixture. In 1916 he tested 907 school children in several Virginia cities, 421 of whom were colored. He made his estimate of amount of Negro blood on the basis of skin color graded by an eye judgment, and finds an increasingly high score with increased amount of white blood.
From which result he generalizes as follows: Such considerations indicate that it is a native ability and not an acquired capacity that differentiates the mixed from the pure negroes, and that skin color is its outward sign. [Italics mine.] They also indicate that the tests used are primarily tests of native capacity, and the consequent differences found between whites and negroes as a whole are innate differences … . The average performance of the colored people of this country in such intelligence work as that represented by the tests of higher capacity appears to be only about three-fourths as efficient as the performance of the white with the same amount of training.
This is obviously reasoning in a circle; he divides his group according to skin color, finds differences between these divisions, then assumes that this proves skin color a valid index and that, therefore, the differences so found are innate.
It seems pretty reasonable to me, although skin color would be a pretty noisy source of data on ancestry. Barack Obama, for example, is 50% white, but because his Kenyan father was jet-black, he looks more like 25% white.
Personally, I’ve always been pretty agnostic on this question of skin color and average IQ famously debated on Saturday Night Live by Julian Bond and Gareth Morris. This might be because going back to 1981 at UCLA, I knew a lot of smart jet-black Africans. Before high-quality DNA data finally became available in the 2020s, I probably leaned 80-20 toward the proposition that fairer African-Americans were in my observation somewhat smarter than darker African-Americans. But it’s hard to tell. Honest people can sincerely disagree on the question.
It was only in this century with DNA data available that I figured out why: the percentage of white admixture in African-Americans with two parents who identify as African-American tends to fall in a pretty narrow range around 20% white. This is due to the workings of the one-drop rule in the United States. America isn’t like Puerto Rico or Brazil, where there is pretty much of a bell curve running from all black to all white with a lot of people in the middle. Instead, in our culture, where people were traditionally either white or black, there was a high degree of regression toward the mean among blacks in terms of white admixture, since interracial marriage was so frowned upon.
So most self-identified blacks are somewhere around 20% white, while most whites have only a literally vanishingly small percentage of black ancestry. Thus in the U.S. the 80th percentile among blacks might be 30% white DNA and the 20th percentile 10% white DNA, which aren’t huge differences, so it’s hard to casually eyeball the question.
In 1919 Ferguson made a similar analysis of the results of the army testing of 5,425 Negroes at Camp Lee. Again on the basis of an eye judgment he divided 1,132 of these Negroes into a “darker” and a “lighter” group, and found that the median Alpha score for the “lighter” group was 51, for the “darker,” 40; on the Stenquist combination test the group classified as “lighter” made a median score of 19; those called “darker,” a median score of 17. Eight additional companies were classified by other observers and gave results, comparable to these. Ferguson has severely criticized a similar attempt, made by Morse and Strong as “rough,” yet his own methods are open to the same criticism.
There is no conclusive evidence that skin color, accurately determined, is a reliable index of racial admixture. Such a classification is even less admissible when unchecked by any such quantitative device as the color top. The methods of Garth and Hunter on one hand, and that of Ferguson and Strong on the other, are illustrations of a scientific and unscientific approach to this problem.
Mead is dubious about the use of the physical anthropologist’s method of eyeballing skin color.
If the genealogical method could be subjected to extensive verification and supplemented by some technique for holding the other factors constant it might be productive of valuable results.
SOCIAL STATUS Several careful attempts have been made to ascertain the influence which social status has in this type of testing. Very often the conclusions of these experimenters have to be rejected because they have concentrated upon this one problem and neglected other complicating conditions, but they are none the less valuable as illustrations of the particular methodological point. In 1912 Phillips made an investigation of the intelligence of white and Negro children in the Philadelphia public schools, using the Binet scale. His treatment of the social status factor was a specific attempt to hold it constant. The homes of the children were visited and rated on a four-division scale. Only those children, twenty-nine from each race, whose homes were rated “good” were used in the final comparisons. This method of elimination of all cases of incomparable social status is open to two objections. In the first place, strict comparability is exceedingly difficult to determine, and the final use of only one status group prevents the utilization of other status groups as checks. And in the second place, it is enormously unpractical in that it pares down the number of cases, and the more exact the classification, the more cases there are which will have to be eliminated.
Miss Arlitt has studied the effect of social status upon race norms more carefully than any other investigator. In most reported work social status has been a side issue, a check upon the main body of the results, not the central problem. But Miss Arlitt undertook to study it specifically. She describes two investigations. The first one was made on 304 children in the primary grades, 169 of native-born white parentage, 68 of Italian parentage, and 67 Negroes. They were divided according to social status and the following results were obtained when the total group was considered:
Median I. Q.
Very superior …. …. 126.9
Superior … ….. 118.7
Average … ….. 106.5
Inferior …….. 87
Very inferior …. 83Ninety per cent of the native-born whites came from families ranked as superior or very superior; 88.2 per cent of the Italians and Negroes came from families of inferior or very inferior social status. When the Italians were compared with native-born whites of the same social status, the difference in intelligence quotient was only 7.7 instead of 22.2, the difference when only nationality groups, without regard to status, were compared. In the case of the Negroes, consideration of the social status factor reduced the difference from 23.3 to 8.6 points. In her second study” Miss Arlitt again studied native-born whites, Italians, and Negroes, making the social status ratings on the basis of Taussig’s occupational grouping. She found a difference of 33.9 between the two extreme occupational groups in the same grade in school. When Italians and Negroes were compared with their own occupational group among the native-born whites of native white parentage a difference similar to the one quoted above was found. From which investigation Miss Arlitt concludes that: “Race norms which do not take the social status factor into account are apt to be to that extent invalid.”
Although many investigators, notably Terman and Yerkes, have endeavored to study the effect of social status upon tests scores, these are the only thorough attempts to analyze its influence when racial and nationality groups are being studied. A less careful attempt along these same lines was made by Strong in 1913 to correct the extreme differences found between the scores of Negro and white children by comparing the Negroes with the poor mill whites; they state that such a comparison was much less unfavorable to the Negro. Another body of data on the question of social status is found in the army results where the effect of northern or southern residence is appreciably indicated in the Negro scores.
Fuckunda in a study of Japanese children,
I gather that at some point over the last century the method of spelling Japanese names in American English was changed so that this honorable scholar’s name would be spelled “Fukunda.”
tried to equate occupation of parent and score of child. However, he found a correlation be- tween social status and intelligence quotient of only .19.
Note that the correlation coefficient had been discovered/invented by Professor Boas’s archrival Sir Francis Galton only 36 years before Mead’s article.
But he had only forty-three cases, and he fails to record the number in each of his three groups, for which his average I.Q.s are given as 103, 82 and 92, respectively. These experiments are suggestive methodologically, although in almost every instance the results are vitiated by overemphasis of this one factor and a failure to deal with the other aspects of the case. Once the importance of the social status factor is established, it will be necessary to further refine the methods by which it is estimated. …
LANGUAGE DISABILITY Perhaps no complicating factor in appraising the results of this kind of intelligence testing has been so neglected as the question of language disability. With the exception of the work of two investigators, the whole matter has usually been dismissed with statements that the child was under no language handicap beyond the first grade if he had gone to an American kindergarten, or that “all the Italians spoke English without difficulty, “or that the children were selected by the principal of the school as having no language handicap.”
Rudolph Pintner has emphasized the importance of this language factor: “The question of prognosis value for school purposes must not be confused with the question of absolute intelligence of different racial groups. It seems to the writer that non- verbal tests alone are adequate for this purpose. It is inconceivable that children living in an English-speaking environment, hearing, speaking, and reading nothing but English, should not have a distinct advantage in tests requiring the finding of opposites, the hunting for an analogy, the filling in of an uncompleted sentence, and the like, as compared with children who hear a foreign language at home and, in many cases, are required to communicate in a foreign language to some people in their environment. Such contrasting groups are very far from having had equal previous practice on the elements which go to make up the usual verbal test.”
Colvin and Allen sought to isolate the effect of language on score by comparing the performance of fifty American and fifty Italian children of the same school grade and attainment on a group, the National, and an individual, the Binet, test. They found a difference of thirteen points when the written group test was used, and a difference of only one point when the oral individual test was used.
From these various attempts to elaborate a methodology of racial and nationality testing it is clearly evident that test scores are affected, to a degree not yet determined, by social status and by language disabilities. No attempt has yet been made to analyze the effects of that more subtle and less measurable aspect of environment which may determine the attitude of the subject toward the tests and profoundly affect his score. The method of equating test score and amount of racial admixture is subject to modification in terms of these other complicating factors, and also to the inherent weaknesses of the method in the present state of ignorance concerning the laws regulating the inheritance of mental traits. All these considerations should suggest extreme caution in any attempt to draw conclusions concerning the relative intelligence of different racial or nationality groups on the basis of tests, unless a careful consideration is given the factors of language, education, and social status, and a further allowance is made for an unknown amount of influence which may be logically attributed to different attitudes and different habits of thought.
In this connection I would like to refer to an unpublished investigation of my own. I gave the Otis intelligence test to 276 Italian children and compared the scores with the amount of English spoken in the children’s homes, by their parents. It is to be noted that all these children were above the fourth grade in school and all spoke English themselves. The comparison showed a steadily rising median from an index of brightness of 64.16 for the group whose parents spoke all Italian, 69.3 for those whose parents spoke some English, 74 for those of homes in which English was chiefly or entirely spoken.
One observation is how much more intelligent and civil were the blank slaters a century ago. While Mead occasionally slips into the kind of strident “there is no evidence” rhetoric popular today (but note that she says instead “there is no conclusive evidence that skin color, accurately determined, is a reliable index of racial admixture”), in general, her article is polite and helpful. Many of her criticisms are on the money.
While the Boas School is often denigrated by the Galton School, I believe it was a beneficial corrective to Galtonian groupthink. Here’s one entrance to the University of Chicago social sciences building, completed in 1929, in which Galton, representing Nature on the left and Boas, representing Nurture on the right (although their positions should be reversed), glare at each other in perpetuity.
I like that. That seems reasonable to me.
Of course, once Boas had triumphed (the 1929 stock market crash was a crucial turning point, when rich businessmen ran out of money to fund academics), Boasian groupthink became at least equally annoying. But as we see here, back when neither held the whip hand, they were actually good for each other.
The recent studies of racial admixture and IQ in the Philadelphia Neurodevelopment and ABCD databases don’t totally prove that Nature plays a role in the racial gap in IQ. It could be that the genetic ancestry drives looks, which in turn drives discrimination, which in turn drives IQ. But that seems more like Occam’s Butterknife than Occam’s Razor.
Charles Murray has been suggesting since the late 1990s that the ultimate acid test would be to look at African-Americans who are more or less whiter looking than their actual genetic ancestry would imply. The Nurture-Only theory would suggest that IQ correlates only with looks, not with ancestry in these cases.
My impression is that white Americans don’t pay all that much attention to degree of white admixture in the phenotypes of African-Americans (Who looks whiter, Samuel L. Jackson or Morgan Freeman? Jesse Jackson or Al Sharpton? Beats me). But African-Americans pay more attention.
We may be only a few years away from doing Murray’s acid test. Combining the ABCD and Philadelphia databases would provide a sample size in the low few thousands of African-Americans. That might be a large enough sample size to tabulate enough individuals whose phenotypes differ from their genotypes.
On the other hand, will free-thinking scientists ever get access to future databases to replicate upcoming results?
Science denialism is a rapidly growing state cult in America. We may soon look back upon the 2010s as a golden age for freedom of science before the Clampdown of the 2020s:
