[Note: This commentary was written in June 2010. For a critique of more recent claims about genomics, race and IQ, see my review of Charles Murray’s book Human Diversity.]
Few issues have created as much controversy in the history of psychology as that of the relationship between race and intelligence. The controversy has been fueled for decades by a subgroup of psychologists who have argued that some racial groups score lower than others on intelligence tests for genetic reasons. At times this pro-genetic position has held significant support among psychologists, but it has lost ground as time has gone on. And this loss of support, as I argue below, has occurred primarily because the central claim of the pro-genetic position is not credible and its reasoning is fallacious. Nevertheless, this position has been kept alive in recent years by the writings of a few proponents (particularly Arthur Jensen and J. Philippe Rushton) and, perhaps more importantly, by the fervent advocacy of a variety of persons from outside the field of psychology.
The pro-genetic position is grounded in two initial facts. The first is that people in disadvantaged groups generally score lower on IQ tests than those who are not disadvantaged. While there are a few exceptions, this is a well-documented trend that has been observed throughout the world. The second fact is that many disadvantaged groups are defined by phenotypic markers – that is, by genetically determined physical traits (skin color, facial features, etc.) that separate them from more privileged groups and that serve as insignias of marginalized status. Given these two facts, it seems natural to many people to conclude that the genetic factors that distinguish disadvantaged groups also cause them to score lower on IQ tests. While such an inferential leap may seem reasonable to many (particularly those lacking in education), it is the kind of logic that raises – or, at least, should raise – skepticism in psychologists, who know all too well the problems of jumping from correlation to causation. Specific behaviors like test performance occur in complex circumstances and should not, as a rule, be attributed, straightforwardly and unproblematically, to what seem to be “obvious” accompanying factors. Such caution is all the more appropriate in the present case when one additionally considers the fact that socially disadvantaged groups that are not defined by phenotypic markers – i.e., marginalized groups that are genetically indistinguishable from dominant groups (for example, disfavored religious and linguistic minorities) – show the same kinds of lowered IQ performance as those that are defined by such markers (see Sowell, 1994, for several examples from a variety of cultures). These trends are consistent with evidence that suggests that measured IQ can be affected by a number of environmental factors that disproportionately act on marginalized populations, including exposure to toxins, malnutrition, injuries during birth, lack of resources and support in the family, dysfunctional social systems, schools lacking in resources, and chronic poverty and discrimination (Aiken, 2000; Anastasi & Urbina, 1997; Cohen & Swerdlik, 1999; see also Wiesen, 2009). As evidence like this has accumulated, most psychologists have been increasingly reluctant to draw genetic conclusions about racial differences in IQ.
Despite the reluctance of most contemporary psychologists to make the inferential leap from correlation to causation, a number of researchers in the early days of the discipline readily embraced the genetic interpretation of racial differences in IQ. There are many examples of early psychologists coming to sweeping conclusions about the inferiority of racial minorities based on the flimsiest of data; and this was nowhere more evident than in the field of psychological testing, where the influence of white supremacist ideology was often quite unconcealed (see Tucker, 1994). As time passed, however, the dwindling pro-genetic school moderated its claims and framed its arguments, increasingly, in complex scientific terminology, invoking highly technical and sometimes arcane statistical relationships. The effect has been to gradually mask, under arrays of data and technical jargon, many of the same errors and sophistries that characterized the original pro-genetic argument and that have persisted, essentially unchanged, over the past 100 years.
The specifics of these errors and sophistries are complex and cannot be exhaustively detailed in a short piece like this. However, I will briefly sketch what I consider to be a few of the most important problems with the pro-genetic position.
To begin with, the pro-genetic argument assumes the equivalence of intelligence and IQ. In reality, however, intelligence is a much broader phenomenon. Many different kinds of intelligence have been identified by psychologists and others (for example, see Sternberg et al., 2005); and IQ tests measure only a rather narrow portion of this broader spectrum – the portion that correlates most highly with successful performance in the kinds of academic institutions associated with dominant Western cultures. There is nothing wrong with looking at a limited segment of intelligence, and it may be impossible to study intelligence without doing so. But when that segment encompasses precisely those abilities that are disproportionately promoted, rewarded, and educationally transmitted within the dominant stratum of a society, then a methodological circularity has been established that ensures that the members of that privileged group will perpetually be “discovered” to be the “most intelligent” members of that society, and of the world at large. More generally, the concept of intelligence, in the broader sense, is an extremely hazy one, and problems like this are scarcely avoidable. As Sternberg et al. note, experts have never been able to agree on a theoretical definition of intelligence, a fact which limits our ability to do meaningful research on this topic (46, 57).
The attempt to study genetic differences in terms of “race” is even more problematic. Races are notoriously difficult to define, and the genetic constitution of such races is correspondingly difficult to specify. If we attempt to do so by phenotype (e.g. skin color) then we invariably encounter problems (for example, many Caucasians have skin of a darker shade than that of many African Americans). Shifting to a different criterion (e.g. hair type) will produce the same problem, but for different individuals; the more traits that are used, the greater the number of individuals who are ambiguous and the fuzzier the boundary between races. Definitional ambiguities, however, are not the only ones. Racial groups, however defined, are genetically open and subject to the migration of genetic variants across boundaries. For these reasons, and a variety of social ones as well, self-reported race is unreliable, often differing from one context to another, deviating significantly from DNA evidence, and frequently leading to inconsistent findings in genetic research (Shields et al., 2005, 90-91). To put it another way, the race concept is based on a typological model of human variation that does not reflect the continuous nature of genetic variation for many traits. Thus, even when differences that may actually be genetic are discovered between (socially defined) races (e.g. differential responsiveness to certain drugs) the relationships are typically so weak they are clinically insignificant (91-92), suggesting that the overlap with race is more likely a result of genetic drift than of selective pressures. Given these problems, productive scientific research is undermined by the use of the race concept as a stand-in for genetics, and this is why geneticists generally avoid using race as a biological category, relying instead on concepts like gene pools and clines (variations of traits within a species).
The concept of heritability also presents serious problems. Pro-genetic adherents rely heavily on the concept of heritability, which is usually defined as the proportion of variability of an observable trait (like performance on an IQ test) that is statistically attributable to genetic variability. This proportion (represented by the term h2) can be estimated by various procedures. Most commonly, the IQ scores of identical twins, who share all of their genes, and those of fraternal twins, who share fewer genes, are mathematically compared to each other via a formula that produces an estimate of h2 as a proportion or percentage. Thus, if identical twins are more similar in IQ than fraternal twins, a relatively high value of h2 will result – for example, a value of .6, indicating that 60% of the variability of IQ is attributable to genetic factors. Since heritability studies of IQ typically produce values in this range, pro-genetic writers interpret this as indicating that IQ differences among different racial groups are attributable to genetics.
However, there are problems with this reasoning. h2 applies to variability among individuals in a specific group (shared environment), and it cannot be used to compare differences between one group (with one shared environment) and another group (with another shared environment). A common example is height. Height seems to be highly heritable – i.e., genetic factors seem to strongly determine variability in height among individuals within a particular environment. But average height can vary considerably between groups that share different environments with regard to, say, nutrition. In fact, this seems to be why average height has been increasing throughout the world in recent decades. (Interestingly, a similar increase has been observed in IQ – the so-called Flynn Effect – which is generally interpreted as supporting environmentalist theories of IQ.) Thus, the use of heritability to make inferences about genetic variability among racially defined groups occupying significantly different environments that exert significantly different effects on cognitive development is fundamentally problematic.
The fact that genetic variability can be most clearly detected within a relatively restricted environment (as in the case of height) also raises questions about the high heritability estimates for IQ. Most such estimates have been obtained from samples of white, middle-class individuals sharing relatively similar environments. Because of the way h2 is defined – genetic variability divided by total (genetic plus environmental) variability – such a restriction on the environmental variability of a sample increases the relative influence of genetics and therefore the numerical value of h2. When heritability is calculated in controlled biological settings, such as the agricultural breeding contexts in which the concept was originally developed, this procedure of maximizing h2 by minimizing environmental variation makes some sense, but the extension of this technique to use restricted environments to make inferences about psychological traits in the larger human environment, with all its variability, has arguably led to greatly inflated estimates of human genetic influence, at least in the case of IQ. One way this might be addressed would be to study the genetic influence on psychological traits across a much broader range of individuals and environments. But this carries its own problems, for the individuals and environments encompassed by this wider range are not randomly distributed. As mentioned earlier, individuals are socially and nonrandomly “assigned” to different environments, in part, by genetically determined physical traits like skin color. The increased environmental variability of IQ will therefore correlate with some genetic factors, regardless of any actual (biological) connection or nonconnection of those factors with IQ. That is, h2 cannot distinguish between variability in IQ that is directly determined by genetic (biological) factors and that which is determined by social assignment to membership in discriminatory environments based on genetically determined physical markers that are cognitively irrelevant.
Keeping the above problems in mind, the main features of the pro-genetic argument can now be briefly summarized, as follows: Intelligence, dubiously measured as IQ, is repeatedly found to be correlated with membership in racial groups, dubiously treated as genetic types, a relationship which is inferred to be causal with the help of mathematical tools like h2, which seriously confound hereditary and environmental factors (typically overestimating the former). The result is a set of claims about the relationships between race, genetics, and intelligence which, when examined more closely, turn out to be little more than a series of free and easy leaps from correlation to causation, but wrapped up in complex mathematical justifications that most people, including many psychologists, have difficulty fully untangling. When pro-genetic psychologists are challenged by their peers about the conflation of correlation and causation (e.g., Kamin, 1995; Suzuki, 2005), they typically respond with such strategies as pointing to data indicating that when specific environmental factors are held constant, race still predicts differences in IQ. But this kind of response avoids the real issue, for what the response treats as a genetic factor (race) is, once again, also a marker of membership in a group that suffers extensive environmental disadvantage. Holding specific environmental factors constant, one at a time, merely avoids and distracts from the more important fact that the remaining complex of environmental factors remains active, in each case, and remains inextricably correlated with the genetic markers of race. Thus, what appears, superficially, to be the marshalling of a great deal of evidence is, in reality, a constantly shifting argument in which the same fallacy is repeated, again and again, but always dressed up in different technical and statistical variants.
It should be noted that one cannot prove that some underlying genetic factor does not contribute to racial differences in IQ, a fact which pro-genetic psychologists frequently emphasize. But the preponderance of evidence indicates that a genetic explanation is neither necessary nor credible. It is tantamount to arguing that because height has a strong genetic component, people who are nutritionally deprived must be shorter for genetic reasons. We do not need a hypothesized genetic factor to explain a psychological phenomenon that arises at the center of a wide array of environmental forces, all systematically converging toward the likely production of that psychological phenomenon. A number of critics of the pro-genetic school have pointed out these problems. For example, Richards (1997) has surveyed many of the conceptual and methodological fallacies in arguments by the pro-genetics school and has concluded that these arguments are no longer deserving of serious scientific attention. At the same time, Richards notes that the controversy goes on, a fact which he attributes to the continued infusions of funding and other kinds of support for pro-genetic authors from politically sympathetic sources outside the field of psychology. Richards points out that some of the most generous support for research by pro-genetic psychologists has come from groups with histories of ties to white supremacist and racist ideologies (280-282; see also Tucker, 1994).
But supporters of pro-genetic research are not all racists. There appear to be two other motives that play a role in driving the dedication of many strongly committed pro-genetic advocates. The first of these is the ideal of scientific integrity, coupled with a belief that pro-genetic findings have been unfairly suppressed by politically correct opponents To a certain extent this has been a self-fulfilling prophecy, as the pro-genetic program of attempting to establish the inherent inferiority of minorities who have already been unjustly treated has provoked some people to understandable anger and occasional excessive responses – responses which then allow pro-genetic authors to play the role of scientific martyrs and heroes. I believe that this has rarely carried over into actual suppression of academic discourse; and if this does occur it needs to be repudiated. The pro-genetic researchers are entitled to make their arguments and to get a fair hearing. This does not, however, entail that their work has any merit or that it carries any claim to attention, interest, or support by the serious scientific community.
The second motive other than racism that appears to drive advocacy for the pro-genetic position is a certain naive individualism that has deep roots in U.S. culture. This view holds that individuals do best when they are free from all societal influence, that interventions that attempt to offset social disadvantages (even for children) are misguided and destructive, and that material success is simply an accurate index and just reward for effort and ability. Of course, there is much worth in the individualist core of American ideology. But as a sole ideological principle, unaccompanied by any recognition of the limits of its underlying assumptions, it can easily harden into a fatuous and self-congratulatory attitude of entitlement by those who have inherited not only a reasonable share of actual abilities but also, and more importantly, a disproportionate share of social privilege, of which they remain emphatically and conveniently unaware. For such a person, the temptation to believe in the identity of material success and genetic superiority is likely to be irresistible, and the genetic theory of racial differences in intelligence is an ideal premise for supporting and reinforcing such a world view. There is no reason to doubt the sincerity of those who hold these beliefs, but sincerity is not a scientific argument.
In the final analysis, then, the pro-genetic argument for racial difference in IQ is an ever-changing pastiche of creatively orchestrated and fallacious interpretations of complicated data masking a pervasively flawed line of reasoning. Its status as a scientific theory is fundamentally problematic and would likely collapse altogether were it not for the continuing infusion of research funds from a segment of the public that is often abundantly resourced and highly invested in securing the dominance of the pro-genetic position. It is a classic example of how research is driven by politics, and not always in a direction that moves us closer to the truth.