III

Turning to race, we must begin with the fraught question of whether it even exists, or whether it is instead a social construct. The Harvard geneticist Richard Lewontin originated the idea of race as a social construct in 1972, arguing that the genetic differences across races were so trivial that no scientist working exclusively with genetic data would sort people into blacks, whites, or Asians. In his words, “racial classification is now seen to be of virtually no genetic or taxonomic significance.”25

Lewontin’s position, which quickly became a tenet of political correctness, carried with it a potential means of being falsified. If he was correct, then a statistical analysis of genetic markers would not produce clusters corresponding to common racial labels.

In the last few years, that test has become feasible, and now we know that Lewontin was wrong.26 Several analyses have confirmed the genetic reality of group identities going under the label of race or ethnicity.27 In the most recent, published this year, all but five of the 3,636 subjects fell into the cluster of genetic markers corresponding to their self-identified ethnic group.28 When a statistical procedure, blind to physical characteristics and working exclusively with genetic information, classifies 99.9 percent of the individuals in a large sample in the same way they classify themselves, it is hard to argue that race is imaginary.

Homo sapiens actually falls into many more interesting groups than the bulky ones known as “races.”29 As new findings appear almost weekly, it seems increasingly likely that we are just at the beginning of a process that will identify all sorts of genetic differences among groups, whether the groups being compared are Nigerian blacks and Kenyan blacks, lawyers and engineers, or Episcopalians and Baptists. At the moment, the differences that are obviously genetic involve diseases (Ashkenazi Jews and Tay-Sachs disease, black Africans and sickle-cell anemia, Swedes and hemochromatosis). As time goes on, we may yet come to understand better why, say, Italians are more vivacious than Scots.

Out of all the interesting and intractable differences that may eventually be identified, one in particular remains a hot button like no other: the IQ difference between blacks and whites. What is the present state of our knowledge about it?

There is no technical dispute on some of the core issues. In the aftermath of The Bell Curve, the American Psychological Association established a task force on intelligence whose report was published in early 1996.30 The task force reached the same conclusions as The Bell Curve on the size and meaningfulness of the black-white difference. Historically, it has been about one standard deviation31 in magnitude among subjects who have reached adolescence;32 cultural bias in IQ tests does not explain the difference; and the tests are about equally predictive of educational, social, and economic outcomes for blacks and whites. However controversial such assertions may still be in the eyes of the mainstream media, they are not controversial within the scientific community.

The most important change in the state of knowledge since the mid-1990’s lies in our increased understanding of what has happened to the size of the black-white difference over time. Both the task force and The Bell Curve concluded that some narrowing had occurred since the early 1970’s. With the advantage of an additional decade of data, we are now able to be more precise: (1) The black-white difference in scores on educational achievement tests has narrowed significantly. (2) The black-white convergence in scores on the most highly “g­-loaded” tests—the tests that are the best measures of cognitive ability—has been smaller, and may be unchanged, since the first tests were administered 90 years ago.



With regard to the difference in educational achievement, the narrowing of scores on major tests occurred in the 1970’s and 80’s. In the case of the SAT, the gaps in the verbal and math tests as of 1972 were 1.24 and 1.26 standard deviations respectively.33 By 1991, when the gaps were smallest (they have risen slightly since then), those numbers had dropped by .37 and .35 standard deviations.

The National Assessment of Educational Progress (NAEP), which is not limited to college-bound students, is preferable to the SAT for estimating nationally representative trends, but the story it tells is similar.34 Among students ages nine, thirteen, and seventeen, the black-white differences in math as of the first NAEP test in 1973 were 1.03, 1.29, and 1.24 standard deviations respectively. For nine-year-olds, the difference hit its all-time low of .73 standard deviations in 2004, a drop of .30 standard deviations. But almost all of that convergence had been reached by 1986, when the gap was .78 standard deviations. For thirteen-year-olds, the gap dropped by .45 standard deviations, reaching its low in 1986. For seventeen-year-olds, the gap dropped by .52 standard deviations, reaching its low in 1990.

In the reading test, the comparable gaps for ages nine, thirteen, and seventeen as of the first NAEP test in 1971 were 1.12, 1.17, and 1.25 standard deviations. Those gaps had shrunk by .38, .62, and .68 standard deviations respectively at their lowest points in 1988.35 They have since remained effectively unchanged.

An analysis by Larry Hedges and Amy Nowell uses a third set of data, examining the trends for high-school seniors by comparing six large data bases from different time periods from 1965 to 1992. The black-white difference on a combined measure of math, vocabulary, and reading fell from 1.18 to .82 standard deviations in that time, a reduction of .36 standard deviations.36

So black and white academic achievement converged significantly in the 1970’s and 1980’s, typically by more than a third of a standard deviation, and since then has stayed about the same.37 What about convergence in tests explicitly designed to measure IQ rather than academic achievement?38 The ambiguities in the data leave two defensible positions. The first is that the IQ difference is about one standard deviation, effectively unchanged since the first black-white comparisons 90 years ago. The second is that harbingers of a narrowing difference are starting to emerge. I cannot settle the argument here, but I can convey some sense of the uncertainty.



The case for an unchanged black-white IQ difference is straightforward. If you take all the black-white differences on IQ tests from the first ones in World War I up to the present, there is no statistically significant downward trend. Of course the results vary, because tests vary in the precision with which they measure the general mental factor (g) and samples vary in their size and representativeness. But results continue to center around a black-white difference of about 1.0 to 1.1 standard deviations through the most recent data.39

The case for a reduction has two important recent results to work with. The first is from the 1997 re-norming of the Armed Forces Qualification Test (AFQT), which showed a black-white difference of .97 standard deviations.40 Since the typical difference on paper-and-pencil IQ tests like the AFQT has been about 1.10 standard deviations, the 1997 results represent noticeable improvement.41 The second positive result comes from the 2003 standardization sample for the Wechsler Intelligence Scale for Children (WISC-IV), which showed a difference of .78 standard deviations, as against the 1.0 difference that has been typical for individually administered IQ tests.42

One cannot draw strong conclusions from two data points. Those who interpret them as part of an unchanging overall pattern can cite another recent result, from the 2001 standardization of the Woodcock-Johnson intelligence test. In line with the conventional gap, it showed an overall black-white difference of 1.05 standard deviations and, for youths aged six to eighteen, a difference of .99 standard deviations.43

There is more to be said on both sides of this issue, but nothing conclusive.44 Until new data become available, you may take your choice. If you are a pessimist, the gap has been unchanged at about one standard deviation. If you are an optimist, the IQ gap has decreased by a few points, but it is still close to one standard deviation. The clear and substantial convergence that occurred in academic tests has at best been but dimly reflected in IQ scores, and at worst not reflected at all.



Whether we are talking about academic achievement or about IQ, are the causes of the black-white difference environmental or genetic? Everyone agrees that environment plays a part. The controversy is about whether biology is also involved.

It has been known for many years that the obvious environmental factors such as income, parental occupation, and schools explain only part of the absolute black-white difference and none of the relative difference. Black and white students from affluent neighborhoods are separated by as large a proportional gap as are blacks and whites from poor neighborhoods.45 Thus the most interesting recent studies of environmental causes have worked with cultural explanations instead of socioeconomic status.46

One example is Black American Students in an Affluent Suburb: A Study of Academic Disengagement (2003) by the Berkeley anthropologist John Ogbu, who went to Shaker Heights, Ohio, to explore why black students in an affluent suburb should lag behind their white peers.47 Another is Black Rednecks and White Liberals (2005) by Thomas Sowell, who makes the case that what we think of as the dysfunctional aspects of urban black culture are a legacy not of slavery but of Southern and rural white “cracker” culture.48 Both Ogbu and Sowell describe ingrained parental behaviors and student attitudes that must impede black academic performance. These cultural influences often cut across social classes.

From a theoretical standpoint, the cultural explanations offer fresh ways of looking at the black-white difference at a time when the standard socioeconomic explanations have reached a dead end. From a practical standpoint, however, the cultural explanations point to a cause of the black-white difference that is as impervious to manipulation by social policy as causes rooted in biology. If there is to be a rapid improvement, some form of mass movement with powerful behavioral consequences would have to occur within the black community. Absent that, the best we can hope for is gradual cultural change that is likely to be measured in decades.

This brings us to the state of knowledge about genetic explanations. “There is not much direct evidence on this point,” said the American Psychological Association’s task force dismissively, “but what little there is fails to support the genetic hypothesis.”49 Actually, there is no direct evidence at all, just a wide variety of indirect evidence, almost all of which the task force chose to ignore.50

As it happens, a comprehensive survey of that evidence, and of the objections to it, appeared this past June in the journal Psychology, Public Policy, and Law. There, J. Philippe Rushton and Arthur Jensen co-authored a 60-page article entitled “Thirty Years of Research on Race Differences in Cognitive Ability.”51 It incorporates studies of East Asians as well as blacks and whites and concludes that the source of the black-white-Asian difference is 50- to 80-percent genetic. The same issue of the journal includes four commentaries, three of them written by prominent scholars who oppose the idea that any part of the black-white difference is genetic.52 Thus, in one place, you can examine the strongest arguments that each side in the debate can bring to bear.

Rushton and Jensen base their conclusion on ten categories of evidence that are consistent with a model in which both environment and genes cause the black-white difference and inconsistent with a model that requires no genetic contribution.53 I will not try to review their argument here, or the critiques of it. All of the contributions can be found on the Internet, and can be understood by readers with a grasp of basic statistical concepts.54

For those who consider it important to know what percentage of the IQ difference is genetic, a methodology that would do the job is now available. In the United States, few people classified as black are actually of 100-percent African descent (the average American black is thought to be about 20-percent white).55 To the extent that genes play a role, IQ will vary by racial admixture. In the past, studies that have attempted to test this hypothesis have had no accurate way to measure the degree of admixture, and the results have been accordingly muddy.56 The recent advances in using genetic markers solves that problem. Take a large sample of racially diverse people, give them a good IQ test, and then use genetic markers to create a variable that no longer classifies people as “white” or “black,” but along a continuum. Analyze the variation in IQ scores according to that continuum. The results would be close to dispositive.57



None of this is important for social policy, however, where the issue is not the source of the difference but its intractability. Much of the evidence reviewed by Rushton and Jensen bears on what we can expect about future changes in the black-white IQ difference. My own thinking on this issue is shaped by the relationship of the difference to a factor I have already mentioned—“g”—and to the developing evidence for g’s biological basis.

When you compare black and white mean scores on a battery of subtests, you do not find a uniform set of differences; nor do you find a random assortment. The size of the difference varies systematically by type of subtest. Asked to predict which subtests show the largest difference, most people will think first of ones that have the most cultural content and are the most sensitive to good schooling. But this natural expectation is wrong. Some of the largest differences are found on subtests that have little or no cultural content, such as ones based on abstract designs.

As long ago as 1927, Charles Spearman, the pioneer psychometrician who discovered g, proposed a hypothesis to explain the pattern: the size of the black-white difference would be “most marked in just those [subtests] which are known to be saturated with g.”58 In other words, Spearman conjectured that the black-white difference would be greatest on tests that were the purest measures of intelligence, as opposed to tests of knowledge or memory.

A concrete example illustrates how Spearman’s hypothesis works. Two items in the Wechsler and Stanford-Binet IQ tests are known as “forward digit span” and “backward digit span.” In the forward version, the subject repeats a random sequence of one-digit numbers given by the examiner, starting with two digits and adding another with each iteration. The subject’s score is the number of digits that he can repeat without error on two consecutive trials. Digits-backward works exactly the same way except that the digits must be repeated in the opposite order.

Digits-backward is much more g-loaded than digits-forward. Try it yourself and you will see why. Digits-forward is a straightforward matter of short-term memory. Digits-backward makes your brain work much harder.59

The black-white difference in digits-backward is about twice as large as the difference in digits-forward.60 It is a clean example of an effect that resists cultural explanation. It cannot be explained by differential educational attainment, income, or any other socioeconomic factor. Parenting style is irrelevant. Reluctance to “act white” is irrelevant. Motivation is irrelevant. There is no way that any of these variables could systematically encourage black performance in digits-forward while depressing it in digits-backward in the same test at the same time with the same examiner in the same setting.61

In 1980, Arthur Jensen began a research program for testing Spearman’s hypothesis. In his book The g Factor (1998), he summarized the results from seventeen independent sets of data, derived from 149 psychometric tests. They consistently supported Spearman’s hypothesis.62 Subsequent work has added still more evidence.63 Debate continues about what the correlation between g-loadings and the size of the black-white difference means, but the core of Spearman’s original conjecture, that a sizable correlation would be found to exist, has been confirmed.64

During the same years that Jensen was investigating Spearman’s hypothesis, progress was also being made in understanding g. For decades, psychometricians had tried to make g go away. Confident that intelligence must be more complicated than a single factor, they strove to replace g with measures of uncorrelated mental skills. They thereby made valuable contributions to our understanding of intelligence, which really does manifest itself in different ways and with different profiles, but getting rid of g proved impossible. No matter how the data were analyzed, a single factor kept dominating the results.65

By the 1980’s, the robustness and value of g as an explanatory construct were broadly accepted among pyschometricians, but little was known about its physiological basis.66 As of 2005, we know much more. It is now established that g is by far the most heritable component of IQ.67 A variety of studies have found correlations between g and physiological phenomena such as brain-evoked potentials, brain pH levels, brain glucose metabolism, nerve-conduction velocity, and reaction time.68 Most recently, it has been determined that a highly significant relationship exists between g and the volume of gray matter in specific areas of the frontal cortex, and that the magnitude of the volume is under tight genetic control.69 In short, we now know that g captures something in the biology of the brain.



So Spearman’s basic conjecture was correct—the size of the black-white difference and g-loadings are correlated—and g represents a biologically grounded and highly heritable cognitive resource. When those two observations are put together, a number of characteristics of the black-white difference become predictable, correspond with phenomena we have observed in data, and give us reason to think that not much will change in the years to come.70

One implication is that black-white convergence on test scores will be greatest on tests that are least g-loaded. Literacy is the obvious example: people with a wide range of IQ’s can be taught to read competently, and it is the reading test of the NAEP in which convergence has reached its closest point (.55 standard deviations in the 1988 test). More broadly, the confirmation of Spearman’s hypothesis explains why the convergence that has occurred on academic achievement tests has not been matched on IQ tests.

A related implication is that the source of the black-white difference lies in skills that are hardest to change. Being able to repeat many digits backward has no value in itself. It points to a valuable underlying mental ability, in the same way that percentage of fast-twitch muscle fibers points to an underlying athletic ability. If you were to practice reciting digits backward for a few days, you could increase your score somewhat, just as training can improve your running speed somewhat. But in neither case will you have improved the underlying ability.71 As far as anyone knows, g itself cannot be coached.

The third implication is that the “Flynn effect” will not close the black-white difference. I am referring here to the secular increase in IQ scores over time, brought to public attention by James Flynn.72 The Flynn effect has been taken as a reason for thinking that the black-white difference is temporary: if IQ scores are so malleable that they can rise steadily for several decades, why should not the black-white difference be malleable as well?73

But as the Flynn effect has been studied over the last decade, the evidence has grown, and now seems persuasive, that the increases in IQ scores do not represent significant increases in g.74 What the increases do represent—whether increases in specific mental skills or merely increased test sophistication—is still being debated. But if the black-white difference is concentrated in g and if the Flynn effect does not consist of increases in g, the Flynn effect will not do much to close the gap. A 2004 study by Dutch scholars tested this question directly. Examining five large databases, the authors concluded that “the nature of the Flynn effect is qualitatively different from the nature of black-white differences in the United States,” and that “the implications of the Flynn effect for black-white differences appear small.”75

These observations represent my reading of a body of evidence that is incomplete, and they will surely have to be modified as we learn more. But taking the story of the black-white IQ difference as a whole, I submit that we know two facts beyond much doubt. First, the conventional environmental explanation of the black-white difference is inadequate. Poverty, bad schools, and racism, which seem such obvious culprits, do not explain it. Insofar as the environment is the cause, it is not the sort of environment we know how to change, and we have tried every practical remedy that anyone has been able to think of. Second, regardless of one’s reading of the competing arguments, we are left with an IQ difference that has, at best, narrowed by only a few points over the last century. I can find nothing in the history of this difference, or in what we have learned about its causes over the last ten years, to suggest that any faster change is in our future.