Misrepresents the Science

Review of Russell T. Warne’s In The Know

It might be questioned whether the science of psychology can ever be completely objective. Russell Warne thinks it can, and in this book he states repeatedly that the facts of psychology are “value-neutral.” We cannot, he says, assume that these facts will conform to our wishes, because “is” does not imply “ought.” (p. 311). Accordingly, Warne makes it clear that he will tell readers the hard truths about intelligence. It turns out that he is not up to this task, partly because he adopts a perspective that seems aimed, above all, at supporting and justifying a hereditarian view of intelligence and denying many significant social and cultural influences on IQ test scores. This hereditarian perspective is so fixed and so intransigent that it involves numerous mistakes and misrepresentations about the nature and study of intelligence, particularly as Warne builds to his claims later in the book about race and IQ. These mistakes and misrepresentations include, but are not limited to, correlation/causation fallacies, circular reasoning, overestimation of heritability, conflation of heritability and genetics, and conflation of within-group and between-group heritabilities. I will give a number of examples, below, of how Warne falls into these and other traps. The sources I cite can be found on the “References” page of this website.

Warne sees himself as a debunker of “myths,” and he repeatedly inflates his claims by referencing “experts” and “mainstream” scientists. However, rather than representing the mainstream, these researchers constitute just one slice of the psychometric field; and a number of them belong to a contingent so wedded to the same hereditarian perspective as Warne, particularly regarding race and IQ, that their stance can best be described as ideological. The field of psychometrics has long included this contingent, and while its members have lost much influence over the years, they have often been disproportionately prominent because their controversial claims have brought them public attention. The current round of ideological hereditarian claims can be traced to 1969 when Arthur Jensen published an article in the Harvard Educational Review suggesting that African Americans were genetically less intelligent than whites. In this and subsequent publications, Jensen made most of the mistakes listed above, along with a variety of highly technical and highly dubious claims about the nature of intelligence. Other ideological hereditarians include Charles Murray, coauthor of the Bell Curve (Herrnstein & Murray, 1994), which shared Jensen’s perspective on race and IQ, and Linda Gottfredson, who recruited a group of likeminded authors to publish a statement in the Wall Street Journal claiming that Jensen’s and Murray’s views represented the “mainstream science on intelligence” (Gottfredson, 1994/1997). These distortions so concerned the Board of Scientific Affairs of the American Psychological Association (APA) that it organized a more representative taskforce of intelligence researchers, who published a detailed and deeply sourced counterstatement with conclusions that contradicted those of Jensen, Murray, and Gottfredson (Neisser et al., 1996).

It is notable that the two authors Warne cites more than any others are Jensen and Gottfredson; and he additionally relies heavily on Murray, J.P. Rushton, Richard Lynn, and other ideological hereditarians. It is also notable that while Warne cites the APA statement in order to make certain points, he never mentions its conclusions. To be fair, Warne also cites researchers who are not especially committed to hereditarianism. For example, he recommends respected books by Earl Hunt (2011) and N.J. Mackintosh (2011) as more thorough and comprehensive than his own (p. xvii). But he fails to note that both of these authors contradict many of his central assertions, as do many other researchers in the field.

Like many hereditarians, Warne conceptualizes intelligence as a single, quantifiable resource located within each individual. In doing so, he makes three interrelated assumptions about the nature of intelligence: (1) that IQ scores are measures intelligence “just as kilograms and pounds are measures of weight” (p. 12); (2) that the intelligence being measured is actually a general cognitive factor called “g”; and (3) that g is a mental ability located in the brain. All three of these assumptions are problematic.

The first assumption confuses psychological with physical measurement. Weight—technically rest mass—is a physical invariant that is measured on a ratio (absolute) scale; IQ, on the other hand, represents a sample of behaviors that are normalized on an interval (constructed) scale. Warne’s wish to assimilate psychology to physics is understandable because the kind of intelligence tapped by IQ tests—the kind that is emphasized in the Western educational system in which IQ tests were developed—is one that aims at rational control of the material and social world. This kind of intelligence is, of course, valuable—indeed, our very survival can depend on it—but it represents just one region of the intelligence spectrum, a region that can be sampled in discrete tasks that can be unambiguously assessed and scored. IQ tests are systematically refined to exclude other kinds of intelligence partly because those other abilities are more difficult to quantify and measure. An example is the Picture Arrangement subtest, which appears to tap social intelligence, since it requires inferences about social relationships from a series of pictures (Sipps et al., 1987). However, this subtest is difficult to administer and has lower reliabilities than desired; hence, it was recently eliminated from the Wechsler IQ tests. For similar reasons, other “multiple intelligences” (emotional, practical, artistic, etc.), which are also difficult to study, often produce less consistent results than those of standard IQ tests (Gardner, 2011, p. xiv). In an act of circular reasoning, Warne concludes that only those skills captured by standard IQ tests represent genuine intelligence (p. 61); but this confuses evidence with exactitude (the so-called “measurement fallacy”; Yorke, 2011) and ignores the long-recognized psychometric trade-off between reliability and validity (Clifton, 2020).

Warne’s second assumption is based on the fact that tests that measure IQ generally correlate with each other. This is not surprising, given the systematic development of these tests described above. Since these tests intercorrelate, certain statistical techniques can be used to derive a hypothetical “factor” which theoretically embodies what all these IQ tests have in common—and this is the general factor, g. Since g is defined by purely mathematical operations, it is not clear just what it represents in the real world, if anything. Warne believes it represents an actual general intellectual ability, but other researches have pointed out that this same mathematical factor would likewise be produced by independent cognitive abilities acting partially in concert (Bartholomew et al., 2009) or strengthening each other during development (van der Maas et al., 2006), or by different environments or different cognitive strategies among individual test takers (Mackintosh, 2011, p. 164, 165). Thus, g may not represent a single entity at all, but rather a variety of different kinds of influences on IQ. Reviewing these and other problems, Hunt concludes “a pure g model is too simplistic” (2011, p. 110).

Warne’s third assumption is that g is an ability—that is, a structure/function in the brain that “transmits its impact” through neural networks to produce one’s performance on IQ tests (p. 10). Warne is very invested in this claim, and he invokes it, implicitly and explicitly, throughout the book. In order to support it, he cites numerous brain studies, concluding “it is indisputable that some characteristics of brain functioning and anatomy correlate with g” (p. 45). Indisputable, yes, but irrelevant. All of these studies involve correlations, and they prove nothing whatsoever about causation, particularly the one-way causation of g repeatedly implied by Warne’s descriptions (e.g., pp. 43-45). Since interactions with the environment affect the brain (see, e.g., Cash-Padgett et al., 2013, Langer et al., 2013), differences among individuals both in g and in the brain can be assumed to result at least partly from the influences of family, schooling, practice effects, encouragement, and other environmental factors and experiences. The studies cited by Warne do suggest that brain structures mediate IQ differences among individuals but they establish nothing at all about the origins of these differences—a fact which other researchers have repeatedly pointed out (e.g., Mackintosh, 2011; Sternberg, 2019).

Warne also slides between correlation and causation when he discusses the “heritability” of IQ. This term sounds like “heredity,” but it means something quite different. Heritability is calculated by comparing the correlations in IQ between related individuals (usually identical twins) to those of other less- or non-related individuals from the same sample. While this measures the variability in IQ associated with genetic difference (a correlation) it does not measure the variability due to genetic determination (a causal relationship) because its magnitude also depends on the variability of the environment from which the sample was drawn. Hereditarians like Warne tend to minimize these environmental influences and therefore overestimate both the size of heritability statistics and their relationships to genetic causation. For example, when Warne first introduces the concept of heritability he relates it to “genetic difference,” a valid statement. But in the very next sentence he replaces this phrase with “genetic influence” (p. 109)—an invalid slide from correlation to causation—and he continues to overestimate both the size and the meaning of heritability estimates throughout the rest of the book.

But the problems get worse. The heritability statistic is constructed mathematically under a set of assumptions called the ACE model, which drastically oversimplifies the ways in which heredity and environment actually operate. For example, it assumes that the social environments of identical twins are no different from those of fraternal twins—an assumption that circularly and disproportionately attributes similarities in traits like IQ to genes rather than the environment (Joseph. 2015). Moreover, the ACE model treats genetic and environmental effects as additive, rather than interactive, and asks how much each one contributes to specific observed behaviors. Mischel (2005) observes that this is like asking which areas of a rectangle are due to the width and which to the length (para. 17). Warne knows about Michel’s analogy, and, following Gottfredson (2009), he attempts to argue that width and length can be “decomposed” as separate influences (pp. 120-122). But this “decomposition” is still an additive process, and employing it is like trying to reduce a rectangle’s area to its perimeter. Furthermore, unlike width and length, genetic and environmental processes both change in complex ways over time, producing interactions that are so dynamic and synergistic that, for most ordinary behaviors, there may be no such thing as purely genetic or environmental effects (Turkheimer, 2011). Consistent with this, when subjected to actual empirical tests, the ACE model’s predictions of additive genetic and environmental effects have not been supported, both at the level of molecular genetics (the so-called “missing heritability problem”; Maher, 2008) and in behavioral studies (detailed in Turkheimer and Waldron, 2000).

Nevertheless, armed with his faith that IQ is largely genetic, Warne minimizes evidence for economic, social and cultural influences on IQ because when these influences are statistically divided into narrowly defined components (income, occupational status, number of books in the home, etc.), and when these components are taken one by one, they have only weak associations with IQ (p. 108). But there is strong evidence that these same influences are powerful when they act in concert—e.g., studies of deprivation and adoption (pp. 135,126; see also Nisbett, paras 9-12, in Turkheimer et al., 2017). Warne, though, is dismissive of this evidence because it is “impossible” to “isolate” the effects of each component (p. 126); and he adds that a broad program to raise the IQs of disadvantaged children would be “expensive” in time, resources and money (pp138-9). In other words, it would be too much trouble, and it’s best just to leave things as they are. This is pure “is-ought” thinking.

Heritability fallacies become supercharged when the discussion moves from differences between individuals to differences between groups. This is because the groups that are studied (social, cultural, economic, etc.) typically share environments that differ significantly from those of other groups, a fact that further and flagrantly undermines the potential of the ACE model to establish anything about genetic determination. Yet differences in group heritabilities lie at the heart of Warne’s reasoning about race and IQ.

First, a word about race. On pp 248-50, Warne attempts to establish a “biological basis” for race by defining it in terms of geographical/migrational ancestral groups. He cites forensic and genetic studies indicating that such groups correlate with ordinary visible markers of race (skin color, hair type, etc.). But this amounts to nothing more than saying “People identified with a common ancestry show signs of a common ancestry.” In order to truly designate these groups as “races” in the biological sense that Warne intends, he would have to show not only that some visible traits are correlated with ancestral populations, but also, among other things, (1) that naturally selected traits other than visible markers are identified with the same ancestral populations; (2) that these other traits tend to follow the same geographical/migrational distributions as those of the visibly identified “races”; and (3) that these other traits include at least some behavioral ones (e.g. intelligence) that result from genetic rather than environmental differences among the groups. Regarding number (1), there is sporadic evidence for a few such traits (e.g., the sickle cell variant); but number (2) is false (see Brace, 2005, for several counterexamples, including the sickle cell variant); and there is no support for number (3)—except for consistently bogus arguments by Warne and other ideological hereditarians (see below). This is why the “race” concept is considered biologically meaningless by most researchers who study human variation (e.g., physical anthropologists).

Ignoring the above and embracing his definition of race, Warne begins his argument for racial differences in IQ by attacking Lewontin (1970), who first revealed the between-group fallacy in Jensen’s original (1969) article. But Warne misrepresents Lewontin’s critique (or perhaps he doesn’t understand it) because he states that Lewontin claims that heritability between racial groups is zero (p. 251). In reality, Lewontin makes no such claim—indeed, he explicitly rejects it (1970, p.7), as does anyone who truly understands the environmentalist critique here, because the inadequacies of the ACE (and other) quantification models preclude measuring genetic and environmental influences in any definitive way. In fact, the question of “how much” influence is genetic and “how much” is environmental may well be meaningless, given the interactivity described above, the unraveling of simple models of the genome (Charney, 2012), and the likelihood that neither human genomes nor human environments can ever be exhaustively decomposed, measured, and experimentally manipulated. While there are strong reasons to conclude that environmental factors alone account for racial difference in IQ (Nisbett, 2005; 2009), neither this, nor any other claim favoring either genes or environment, can be unequivocally proven, much less given exact quantitative specification (see Nisbett, para. 3, in Turkheimer et. al, 2017).

Warne, however, persists in misrepresenting not only Lewontin but all “environmentalists” as insisting that the heritability of race differences is measurably “zero,” and he even underlines his claim by portraying the alleged insistence as a mathematical formula (!). Not only does this distort the facts and create a strawman; it also lets Warne portray himself as the reasonable one in the debate—the only one who is not making a definite claim—and from this faux-reasonable position, he presents five arguments for a genetic black-white IQ gap. Three of these arguments are trivial and can be quickly put to rest. One is a variation of “Spearman’s hypothesis” that differences in g among races proves white superiority, and it entails all the fallacies about g discussed above; another is a technical argument about the validity of IQ tests across races, and it is relevant only to the naive claim that IQ differences reflect nothing but test factors; and the third concerns IQ patterns in racially admixed populations—patterns which imply nothing because genetic and environmental influences are hopelessly entangled in them.

Warne’s other two arguments contend (1) that environmental differences between races are too small to account for IQ differences; and (2) that data from molecular genetic studies support his claims. Taking the latter argument first, Warne cites several studies in which specific genes are statistically associated with IQ scores in specific populations. Just as in the admixture studies, however, genetic and environmental factors are inextricably intertwined here, since populations that are genetically distinct are typically culturally distinct as well. Warne emphasizes the former and ignores the latter. Since these data are entirely correlational, no valid inference about causation is possible. Hamer and Sirota (2000) have pointed out that one could use exactly the same line of reasoning to conclude that Asians have specific genes for using chopsticks. Warne knows he is on thin ice here, and he acknowledges a number of limitations of this research, but he assures the reader that more studies will soon bring clear results. Yet these kinds of studies have been accumulating for over two decades, have produced weak, inconsistent, and dubious results, and have established essentially nothing (Torrey, 2019; Turkheimer, 2019).

Warne’s other argument involves calculating how different the environments of whites and blacks would have to be, if all the independent causal components were discovered and added together, to produce a between-group heritability of “zero” for IQ. Given the problems of reduction and quantification described above, his “answer” (1.414 standard deviations) represents nothing more than a pseudo-exact concatenation of assumptions that are, in reality, speculative, unproven, and—in all likelihood—unprovable. And his conclusion that this result “seems unlikely” to disprove hereditarianism represents nothing more than an intuitive belief on Warne’s part that the accumulated legacies of slavery, dispossession, disenfranchisement, segregation, discrimination, and marginalization can’t really have been all that bad.

There are many other problems with this book, but space limitations preclude a complete inventory. Identifying them, therefore, will have to be left as an “exercise for the reader.”