The last thing I wanted to see happen with my piece on race, genes, and IQ has happened, namely that the comment thread was hijacked by an acolyte of J. Philippe Rushton. Someone who isn't familiar with the relevant literature, or (understandably) doesn't have the time to wade through a few reams of journal articles, could easily walk away with the understanding that there is some impressive body of research supporting Rushton's claims of black racial inferiority.

In fact, one of the most interesting consequences of researching the piece was my discovery that a huge proportion of the popular and academic literature of the last couple of decades supporting the hypothesis of race-linked genetic discrepancies in intelligence is either Rushton's own work or based on his work. This casts the entire discourse in a new light since --- as I hope I am going to show --- Rushton and his cadre are the 9-11 truthers of social science.

This will shortly get prolix and technical, so the rest is below the fold.

Rushton is a psychology professor at the University of Western Ontario whose entire corpus, more or less, is an effort to demonstrate a strict, biologically hard-wired hierarchy of races. His method, which he calls "aggregation," is to select hundreds of thousands of pieces of raw data of wildly variant quality, garnered from multiple, incongruous studies done in inconcruous spatial and temporal settings, and then take their simple arithmetic mean. I can't hope to improve on Douglas Wahlstein's summary, so here it is:

However, averaging does nothing to reduce bias in sampling and measurement, and such flaws are abundant in the cited literature [from Rushton]. For example, among the 38 reports on brain weight, all but two gave figures for only one group, with most cases being people living in the nation of their ancestors, such as an article on Japanese living in Japan and another on Kenyans living in Kenya. The obvious differences in environment make all of these data of dubious worth for testing hypotheses about genetic causes of group differences. The methods of obtaining the brains were also far from contemporary standards for neuroscience. A report of five black Civil War soldiers from 1865 is given the same weight as a 1934 study of over 300 dead Kenyans. One of the two studies with more than one racial group involved the unclaimed bodies of the indigent and executed criminals in the Baltimore area. Those data varied greatly in the time from death to removal of the brain and method of preserving the brains. Numerous factors can affect measures of brain size, and valid inferences about group differences can be drawn only if it is certain that members of different groups were treated the same way. In my opinion, most of the data raked into one big pile by Rushton are worthless for scientific analysis and should be excluded. Unfortunately, Rushton has not done the hard work of separating the potentially valuable data from the trash. He misleads unwary readers by claiming that averaging many studies can overcome poor research methods.

Or to put it more simply, Rushton's inferences are based on obviously and transparently fraudulent methods of data gathering and processing. So much so that his own publisher, Irving Louis Horowitz, was forced to disown him, albeit in highly circumspect language. That's not even the worst of it. Zack Cernovsky's (yes, peer-reviewed) review of Rushton's Race, Evolution, and Behavior (sorry, no link) noted that "some of Rushton's references to scientific literature with respects to racial differences in sexual characteristics turned out to be references to a nonscientific semipornographic book and to an article in the Penthouse Forum."

(The only respnse Rushton has managed to muster is a laughable claim that it means nothing for critics to expose the fraudulence of research until they themselves produce contradictory studies. Sadly for Rushton, many of them have done so.)

If Rushton is bad at statistics, he is worse at biology. Proceeding, per his form, from pre-theoretical assumptions of a strict three-tiered race hierarchy explicable along the r and K dimensions (briefly, r quantifies fertility and tendency to reproduce, K quantifies efficient resource husbandry), Rushton wields this framework as a "Procrustean bed," in psychologist David P. Barash's phrasing, torturing all available data to fit the model and ignoring whatever data remain recalcitrant. More from Barash:

In brief, he argues that `Negroids' are relatively r-selected, `Mongoloids' K-selected, and `Caucasoids' in between. All racial distinctions are then seen to derive from this grand pattern, from differences in genital anatomy, to reproductive regimes, to IQ, etc. He even points to the higher frequency of low birth weight babies among black Americans, data that are undeniably consistent with an r-selection regime, but which might also be attributed to poor nutrition and insufficient prenatal care, and which, not coincidentally, have other implications for behaviour, IQ not the least.

What Barash is getting at is a typical pattern for Rushton, namely, taking a set of data of dubious provenance to begin with, which, even if assumed to credible, is open to a wide variety of interpretations, and then simply pronouncing the data to exclude all non-genetic explanations.

This is perhaps the most salient point of all to keep in mind in the face of the insistence by hereditarians --- many of whom, remember, have no research other than Rushton's on which to ground their claims --- that the relevant data exclude non-genetic explanations of the black-white IQ gap (never mind studies like this one that demonstrate no black-white IQ gap once prior skills are controlled for). The claim that non-genetic explanations of data are to be excluded is not something the data themselves proclaim, but rather, of course, an interpretation, and a highly suspect one at that. Rushton, for his part, routinely pronounces non-genetic explanations of racial disparities to be excluded even when an objective survey of the case clearly points to non-genetic explanations. Wahlstein points to a particularly flagrant example:

Such ardent partisanship also leads Rushton to proceed with genetic arguments on the basis of data that are obviously confounded with the environment. He claims that Africans have very low average IQ scores, even lower than American blacks. His evidence includes IQ test scores of black children in the Republic of South Africa prior to 1990 who were attending "typical primary schools'' there, schools widely known to be substantially inferior to those of the ruling white minority. These numbers tell nothing about the role of genes, yet that is the way they are interpreted in this book.

Is that clear? Rushton infers that disparities in academic performance between blacks and whites in apartheid-era South Africa cannot be explained except genetically.

As it happens, we actually do have non-genetic bases for explaining the correlations Rushton upholds as so impressive --- and his conclusions go far, far beyond the narrow issues of IQ test performance that folks like Saletan are interested in --- and indeed, simply glancing at these alternatives demonstrates how fundamentally shoddy Rushton's procedures and inferences are. In a comment to a study he co-authored in Current Anthropology in 2001, Jonathan Marks puts the matter in a convenient nutshell:

The bane of such quackery is the rigorous use of scientific controls, and the better the controls, the weaker Rushton's arguments about race, biology, and intelligence are empirically. Two recent studies demonstrate this nicely. David and Collins (1997) studied the relationship between birth weight and race, in which black Americans are at higher risk for having low birth-weight babies even when the data are controlled for socioeconomic variables. Here is a feature both evidently racial and biological. Yet when they introduced a significant control, namely, African immigrants to the United States, the racial pattern vanished; the African-born immigrants clustered with American whites rather than with American blacks. The low-birth-weight phenomenon appears to be not an endowment of the black gene pool but a consequence of the experience of growing up black in America. The obvious implication is that this experience is sufficiently different from the experience of growing up white in America as to render gross comparisons of diverse adult phenotypes entirely unrepresentative of underlying genetic patterns. This is not surprising to an anthropological audience...

Now, of course, there is no reason to think that such a correlation [between brain size and IQ] would be impossible. If factors such as diet and the circumstances of life affect both brain size and IQ, then they could be correlated without being causally related. Thus, Rushton's brandishing of correlations would have little scientific merit. And indeed, another recent study examines the relationship between brain volume and IQ (Schoenemann et al. 2000) but partitions the variation in a significant way. With three relevant variables (IQ, brain size, and conditions of life), these researchers control for the conditions of life by contrasting the relationship between IQ and brain size within families (where the conditions of life vary little) and between families (where the conditions of life vary more substantially). They find a correlation between IQ and brain size only across families, where both the conditions of life and the volume of the brain vary. Within families where brain volume differs but the conditions of life differ much less, there is no correlation between brain volume and IQ. To the extent, then, that there may be an empirical relationship between brain size and IQ, it is far more likely to represent a spurious statistical consequence of common life circumstances than it is to represent a deterministic nexus linking size of brain and size of thought.

Note the clause in bold. It is a token of a very basic form of illogic that pervades Rushton's work. And this gets us to the bottom line. As I hope the foregoing examples demonstrate, Rushton's entire corpus is shot-through with limpid and demonstrable abuses of biological theory, research procedures, and statistical analysis, all such abuses marshalled in order to create the illusion that there is a scientific case that blacks are inferior in all measurable respects to whites. Rushton's studies do not meet even the minimal standards of academic responsibility or defensibility, and the recourse hereditarians make to his research should be interpreted as a concession of the vacuousness of their case.

Naturally, therefore, the secondary moral of this story is that peer-review, while a necessary institution, is not perfect, and the fact that a particular study passes peer-review is not a guarantee of its credibility. (Recall the Sokal hoax.) If the best Rushton's defenders can say is that his work has been peer-reviewed, and cannot address the myriad flaws in his research, that is thin gruel indeed. Carleton Coon's journal articles continued to pass peer-review decade after decade, as he would casuistically revise and contort his imploding theories of racial hierarchy to match as closely as possible an increasingly overwhelming and undeniable preponderance of data invalidating his life's work. So it is with Rushton.

Basta.

[Full disclosure: When I was a boy, my father edited and published some of Jonathan Marks' work. I have never met Marks, have no relationship with him at all, and happened to find his material on my own.]