Witty and Jenkins (1936)

Witty and Jenkins’ ancient study of 63 prodigious black children is frequently cited by environmentalists and others. It’s often asserted that this study provides strong evidence against the genetic hypothesis (e.g., Flynn, 1980; Brody, 1992; Lee, 2010; Nisbett, 2009). For example, in his critical review of Nisbett’s Intelligence and How to Get It, James Lee states:

[Nisbett] cites a study failing to find elevated European ancestry in a sample of gifted black children (Witty & Jenkins, 1936). Although this study does pose rather strong evidence for an environmental hypothesis, Nisbett does not mention a critical limitation: the investigators ascertained degree of white ancestry by parental self-report.(Lee, 2010. Review of intelligence and how to get it: Why schools and cultures count, R.E. Nisbett, Norton, New York, NY (2009). ISBN: 978039306505)

The study, which had two components, was a subpart of a larger study by Witty and Jenkins on intellectually superior black children in the Chicago Public schools. Using Terman’s methodology, Jenkins was able to identify, and then study the characteristics and demographics of 103 intellectually superior (IQ >120) children out of a population of ~8000. To address the genetic hypothesis, Witty and Jenkins looked at the relationship between genealogy and IQ for a subsample of these. Witty and Jenkins reasoned that, were the genetic hypothesis true:

In a mixed group such as we have in the United States those individuals having the largest amount of white ancestry should on the average stand higher in tests, other things being equal, than persons of total or larger amounts of Negro ancestry. (Witty & Jenkins, 1936, p. 180).

They conducted two tests of this hypothesis. In the first, they estimated the racial admixture of 63 of the children on the basis of parental report* and then compared the average amount of admixture found to that found in a supposedly nationally representative sample of blacks discussed by Herskovits (1930). Herskovits estimated racial admixture from reported parental and grandparental admixture and in some instances genealogical records going back two generations.

Table 1. Herskovits’ Ancestral data (A) and Methodology (B)

Witty and Jenkins determined that the superior children had less White ancestry and concluded that the genetic hypothesis was falsified. Unfortunately for their conclusion, as Mackenzie (1984) pointed out, Herskovits’ sample was not representative. The sample had a substantially higher than average SES, with 50% of the individuals being either Howard University students or well-to-do professionals. Worse, as discussed by Loehlin et al. (1975), Herskovits’ sample seems to have had more White admixture than the national average. If we translate Herskovits’ ordinal ancestry data into percentages (e.g., N= 100% African; NNW=66% African, 33% Caucasian, etc.), we find that his sample had a White admixture rate of 31%; this is compared to the current national estimate of 20% based on DNA markers (e.g., Zakharia et al. 2009) and to an estimate of 13% for Chicago blacks — the more relevant comparison population — again based on DNA markers (e.g., Reed, 1969). Using the same method of conversion, as above, we see that Witty and Jenkins’ sample had a 33% admixture rate. Applying this method to a sample of mostly college students reported by Meier (1949), with which Herskovits’ method of tabulating ancestry was used, we get an admixture rate of 35%. Witty and Jenkins, of course, were right that their intellectually superior sample didn’t have a higher percent of White ancestry than Herskovits’ — or Meier’s — but both samples, nonetheless, seem to have had a higher percent than found in both the national and Chicago populations as determined by DNA. (Jenkins’, Herskovits’, and Meier’s samples were more admixed, as determined on the basis on genealogy, than the national average, as determined on the basis of DNA markers, by a standardized difference of or over 0.35 SD. See table 2.) And all samples had higher social economic statuses than average; in Jenkins’, 2/3rds of the children hailed from families in which the fathers were in the “upper occupational levels”; in Herskovits’, 50% of the sample was highly selected; in Meier’s, the individuals were mostly college students.

Table 2.


Witty and Jenkins’ results for their first test, thus, seem to support the genetic hypothesis. Higher IQ African-American children were found to have a higher percent of White ancestry than both the national population and the population from which they were drawn. The comparison data set which Witty and Jenkins rely on, likewise, supports the genetic hypothesis. Herskovits’ sample of African Americans were found to have both a higher SES, an IQ correlate, than average and a higher percent of white admixture. The same holds with Meier’s sample. The literature on “colorism” corroborates the finding of a correlation between SES and admixture and admixture and IQ. As I noted elsewhere:

From 1850 to the early ‘1900s, US census takers were instructed to classify African Americans as Black or Mullato. They were given the following directions: “in all cases where the person is white, leave the space blank; in all cases where the person is black, insert the letter B; if mulatto, insert M” and “Be particularly careful in reporting the class Mulatto. The word is here generic, and includes quadroons, octoroons,and all persons having any perceptible trace of African blood” (Snip, 2003).

Hill (2000) found that those African-Americans classified as Mullato had a higher SES (judged by profession — e.g., white collar workers versus domestic workers) than those classified as Black and that this difference remained after controlling for social origins. Hill (2000) rejected a genetic interpretation, arguing that “[e]xplanations for a cultural or genetic origin can not be supported. Research has failed to uncover any association between white ancestry and intellectual ability among African Americans” and citing Scarr et al.; yet, as we noted above, those studies were inconclusive.

The case could reasonably be made, of course, that it’s invalid to compare admixture rates based on geneological information with those based on DNA markers because rates based on genealogy are much less accurate. If we grant this, though, we are left with no standard with which to compare Jenkins’ prodigious youth. Not only were Herskovits’ and Meier’s samples unrepresentative but, more problematic, Herskovits and Meier used a different methodology than Jenkins in calculating admixture. Given the differences in methodology, the samples can only be compared on the assumption that they accurately capture admixture. Jenkins asked both parents to estimate their own racial admixture. From this he estimated the children’s. Alternatively, Herskovits and Meier calculated admixture based on the reported genealogy. An example of the latter method can be seen in table 1. As Loehlin et al. (1975) noted, this method lends itself to overestimation. To quote:

“The figure in both Jenkins’ and Herskovits’ sample … suggest a somewhat higher proportion of caucasian ancestry (approximately 30 percent) than one might expect from Reed’s data based on blood group genes (Reed, 1969). But it is quite possible that this discrepancy is due in part to a bias in the method of classification used — for example a person reporting all four grandparents as “mixed” would be classified as “about equally negro as White” (Herskovits, 1930 p. 14) even though the odds are that such a person would have more black than white, since more “mixed” blacks in the current generation were “more negro than white” then were “more white than negro”.

Loehlin et al.’s point applies only to Herskovits’ (and Meier’s) data. Loehlin et al. didn’t realize that Jenkins used a different method, which did not suffer from this bias — but undoubtably did from others. So either we grant that Jenkins’, Herskovits’, and Meier’s results were accurate, in which case it’s valid to compare them with results based on genetically informed methods or we don’t, in which case we have no standard against which to compare Jenkins’ results. Either we have found support for a genetic hypothesis or we have found no admissible evidence against it.

In the second test, Witty and Jenkins took a gifted (IQ > 140) subset of the 63 children and compared the average ancestry of the subset to that of the larger group (IQ = 125-140). They found no average difference in ancestry and concluded, again, that the genetic hypothesis was falsified.

On problem with their methodology was that they compared the gifted subset (>140) with the larger group (>125) instead of, more properly given the small sample size, with the non-gifted subset (125 to 140). When the proper comparison is made there is a slight, but nonetheless, non-significant difference as shown in the figure below.


Witty and Jenkins’ results for their second test, thus, seem to more support the environmental hypothesis. How strongly, though? To put this question otherwise: what difference in admixture would a genetic hypothesis have predicted — 2%, 5%, 10%, 20% — given an approximately 1 Standard deviation difference in IQ? And how large of a sample size would have been needed to detect a statistically significant difference (or the absence of one)? It’s not at all clear. To answer this, one would need to know the predicted correlation between IQ and individual ancestry, in addition to the means and variance of admixture, in this population and that’s unknown. Whatever the case, there can be no doubt that this is a much weaker test than the former. Here we had 28 kids drawn out of an already IQ selected sample of 63, with an approximately 1 standard deviation difference between the groups. In the former test, we had 63 kids drawn from an unselected sample of 8000, with an approximately 3 standard deviation difference between groups. While the first test, the results of which seem to support a genetic hypothesis, probably had the power to reject the null, this second certainly did not.

It might be worthwhile to explore the results in some more detail to show how consistent they are with a genetic hypothesis. To do this properly, we would need an estimate of the mean and variance of admixture in the 1930 Chicago Black population, which we obviously don’t have. We do have estimates for the 1990 to 2000 national population, which we can use as a substitute. According to Zakharia, et al. (2009):

“Numerous studies have estimated the rate of European admixture in African Americans; these studies have documented average admixture rates in the range of 10% to 20%, with some regional variation, but also with substantial variation among individuals [1]. For example, the largest study of African Americans to date, based on autosomal short tandem repeat (STR) markers, found an average of 14% European ancestry with a standard deviation of approximately 10%, and a range of near 0 to 65% [1], whereas another study based on ancestry informative markers (AIMs) found an average of 17.7% European ancestry with a standard deviation of 15.0% [2].…
…These results were confirmed in the estimation of IA by using the program frappe (also in Figure 1). The amount of European ancestry shows considerable variation, with an average (± SD) of 21.9% ± 12.2%, and a range of 0 to 72% (Table 1).”

If, based on this, we assume a 1930 Chicago admixture of 20% with a standard deviation of about 15%, we can infer a predicted IQ-ancestry correlation, given a genetic hypothesis which proposes that 75% of the 1 SD Black-White difference is genetic. From this we can calculate how much more admixed we would have expected Jenkins youth to be.

One interpretation of a correlation coefficient is: amount of change in x, change y or, in this case, the amount of change in admixture per change in genetically conditioned test score. 
In this case the genetically conditioned difference between Blacks and White would be 0.75 SD, since we are proposing that 75% of the gap is genetic; the ancestry difference would be 5.3 SD, which is the number of SDs separating Blacks who are 20% White and Whites, given that 1 SD of admixture equals 15% Whiteness ((100-20)/15=5.3). The correlation between test scores and genotypic ancestry, in this population, would then be 0.75/5.3 or 0.14. This means that Blacks, in this population, who were selected 1 SD for intelligence would be selected 0.14 SD for white ancestry or that they would be 2% more admixed. This is a little more than what was seen in Jenkins’ 2nd test results but not significantly so. It’s worth noting, at this point, that other studies of admixture and IQ in the African-American population have show a correlation between genealogy and cognitive ability (e.g., Tanser (1939); Tanser (1941)). These are, of course, ignored by proponents of radical environmentalism.

In the case of Jenkins’ first test, which was the more powerful one, the difference between the selected and unselect children was about 3 SD, so the children should have been 3 X 0.14 SD more admixed or 6.3% more admixed than the reference population. If we compare these results with those found from genetic analysis, we will see that in no way do they contradict a genetic hypothesis – rather they are quite consistent with it.

Notes

*Jenkins (1934) tells us: “The racial composition of sixty-three subjects of 125 IQ and above was determined from genealogical data provided by parents…The following procedure was utilized in determining the racial composition of the children. Parents were asked to state to the best of their ability, their racial composition i.e., the approximate proportion of Negro, white, Indian, or other racial Ancestry. The racial composition of each child was then computed from that of his parents. The subjects were divided into four groups: 1) N (those having no white ancestry), 2 (NNW (those having more negro ancestry than white ancestry), 3) NW (those having about an equal amount of Negro and white ancestry), 4) NWW (those having more white ancestry than Negro.”

References

Jenkins, 1934. A Socio – Psychological Study of Negro Children of Superior

Herskovits, 1930. The anthropometry of the American Negro

Loehlin et al., 1975. Race Differences in Intelligence

Mackenzie, 1984. Explaining race differences: in IQ The Logic, the Methodology, and the Evidence

Meier, 1949. A study of the racial ancestry of the Mississippi college Negro

Reed, 1969. Caucasian Genes in American Negroes

Witty and Jenkins, 1936. Intra-Race Testing and Negro Intelligence

Zakharia, et al., 2009. Characterizing the admixed African ancestry of African Americans