Response to Saletan

Saletan's claim that a recent study's data can be used to show a genetically based cognitive disadvantage for Africans is wrong

William Saletan, in three recent articles on Slate.com (1, 2, 3 ), makes the case for innate racial differences in IQ potential. A centerpiece of his argument in his last article in the series ("All God's Children") is a claim that data from a recent study linking breastfeeding and IQ scores shows, through further investigation, a genetic disadvantage for Africans. Yet Saletan's reading of the evidence is selective, his methodology is unsound, and conclusions are unfounded.

[Saletan was quickly taken to task, with bloggers leading the charge, for his supporting evidence and methodology in the series as a whole, as I summarize in another post. But the IQ-breastfeeding claim has, it seems, so far escaped notice. ]

The study itself, published in the highly respected journal PNAS, suggests a link between breastfeeding, individual genetic makeup, and cognitive development. That breastfed babies, all else being equal, tend to end up with higher IQs than those who are not breastfed is a well-established fact. This study links that IQ boost from breastfeeding to a specific gene. The gene in question, called FADS2, codes for a key element the of body’s machinery for processing fatty acids, and breast milk is high in certain fatty acids thought to be important in brain development.

The researchers examined two large sample groups (one from New Zealand and one from Britain), and found that breastfed babies ended up with higher IQs than non-breastfed babies only if they had a certain form of the gene, as did the large majority of the sample. In the small minority of babies who lacked this form of the gene, no IQ benefit from breastfeeding was seen. Among all breastfed babies, those with the IQ-boosting version of FADS2 had a four-point IQ advantage over those who lacked it — a meaningful and provocative result suggesting a very specific interaction between nature (the children's genetic makeup) and nurture (whether or not they were breastfed).

Saletan uses the study's results combined with gene frequencies available in a research database to claim that this IQ-boosting gene is less prevalent in people of African descent by a factor of two to four as compared to other population groups around the world, and by extension that Africans are "cognitively disadvantaged." In fact, this conclusion is completely unsupported by any data and is the result of unsound methodology. (Continue reading...)


The study
The researchers tested for different forms of the fatty-acid processing gene in their sample groups. These different forms were the product of differences of just one letter in the DNA code -- a "C" rather than a "G", for instance. Such places of variation are known as SNPs (single nucleotide polymorphisms, pronounced "snips"). SNPs are specific spots where the gene sequence is known to vary by one letter in the population. SNPs are relatively rare in the DNA code (because the vast majority of our DNA is identical with that of everyone else on the planet), and once identified (or "mapped") provide a convenient way to map genetic differences.

Any single gene can have many SNPs scattered throughout it. To select SNPs to examine in the study, the study's authors made use of the International HapMap Project, which has created a database and map of SNPs in the human genome. The SNP map and data are based on DNA sequences from four small samples of populations around the world: Yoruba from Ibadan, Nigeria; Utah residents of northern and western European ancestry; Japanese from Tokyo, and Han Chinese from Beijing. In HapMap's database, there are over 40 SNPs identified in the sequence of the FADS2 fatty-acid processing gene. Making educated guesses about which SNPs were important to the gene's function, the study's authors chose two of these SNPs to test in their Britain and New Zealand sample groups.

The result for the first SNP (called "rs174575") were very meaningful. At this SNP, there are two variants in the human population: C and G. The C-form is the form that seems to enable an IQ boost in breastfed children. In both study populations, breastfed babies with at least one copy of the C-form scored several IQ points higher later in life than non-breastfed babies with at least one C-copy. (Remember that we all have two copies of every gene.) Those that did not have a single C-copy (what we can call the G/G group) saw no IQ benefit associated with breastfeeding. The G-version was also the least common: the G/G group made up just 10% of the British study group and 7% of the New Zealand study group.

The results from the second SNP ("rs1535") were not so conclusive. Like the first SNP, the G-version is least common and showed no propensity to boost IQ scores in the presence of breast milk. The more common version (in this case, the A version) did boost IQ scores in breastfed children, but only in the British sample group. In the New Zealand results the data showed no link between the variations of the FADS2 gene at this SNP and IQ scores in breastfed children.

Saletan's claim and why it's wrong
The study provided strong evidence that breastfeeding boosts IQ only if you have a particular version of the fatty-acid processing gene. Specifically, the C-version of the first SNP (rs174575) provides the IQ boost, while the G-version does not. The HapMap database provides the frequencies of the two different versions for each of its populations. So Saletan looked up what percentage of each HapMap population is G/G.

rs174575: Percentage of HapMap populations who are G/G

Utah, U.S.A. (European heritage)	5%
Beijing, China (Han)	2%
Tokyo, Japan	2%
Ibadan, Nigeria (Yoruba)	10%

Saletan takes this to be an salient difference in racial IQ capability:

"The Africans are twice as likely as the Americans, and four times as likely as the Asians, to start life with a four-point IQ deficit out of sheer genetic misfortune. Don't tell me those Nigerian babies aren't cognitively disadvantaged. Don't tell me it isn't genetic. Don't tell me it's God's will. And in the age of genetic modification, don't tell me we can't do anything about it....No, we are not all created equal."

In fact the data shows nothing of the sort. Nigerian babies aren't cognitively disadvantaged. Saletan's reasoning is wrong on at least three distinct counts, any of which alone is enough to debunk any claim about racial IQ differences that can be made based on this study's data or any extension of it. (To be clear, Saletan's reasoning is not represented in any way in the actual study, which is not about race).

First, the frequency of the G/G genotype in the Yoruba cannot be taken as representative for Africa as a whole (and by extension, for people of African descent around the world). The size of the HapMap samples from which the map was made is small: each was less than 60 unrelated people. HapMap is representative in the sense that it contains most single-site human genetic variations in the world (it was constructed from four pretty spread-out groups in order to achieve this), but this doesn’t mean that the gene frequencies in these small populations are representative of large, continental populations for any given variation. HapMap's own guidelines for using their data make this clear: "These samples should not be described merely as 'African,' 'Sub-Saharan African,' 'West African,' or 'Nigerian,' since each of those designators encompasses many populations with many different ancestral geographies", with similar statements for the other populations sampled.

Furthermore, the percentage of study subjects who were G/G -- all of them whites from Britain and New Zealand -- is as high as in the Yoruba sample.

rs174575: Percentage of study participants who are G/G

British sample group	7%
New Zealand sample group	10%

(Note: these numbers are not directly presented in the study but are easily calculated.)
As you can see, the percentage of the New Zealand study group who are G/G is just as high as the HapMap Yoruba. These numbers illustrate the mistake of extending the HapMap frequencies to larger populations, as well as showing that even if we were to take the Yoruba sample to be representative for some larger group (a dubious proposition, for reasons stated above), white populations do not, in reality, have a more "favorable" genetic distribution in this case. (Both the New Zealand and the British sample groups in the study excluded nonwhites, as stated in the Methods section of the paper, and both sample groups contained over 1,000 people).

Second, the results found for the other SNP (rs1535) shouldn't be ignored. If you look at the British group only, the results are almost identical to those for the first SNP: people with only the less common variety (coincidentally also the "G" version) showed no IQ boost from breastfeeding, while those with at least one copy of the other version (A in this case) got the IQ boost. But this result wasn't replicated in the New Zealand group, where there was no association between the SNP variants and IQ. (Here's the study's raw data, reproduced). Nevertheless, as the British result for rs1535 was just as solid as the rs174575 results, it certainly would be fair to say that such a link is possible. Here's are HapMap gene frequencies for the "bad" G/G version of rs1535:

rs1535: Percentage of study participants who are G/G

Utah, U.S.A. (European heritage)	13%
Beijing, China (Han)	16%
Tokyo, Japan	11%
Ibadan, Nigeria (Yoruba)	2%

The proportions are decisively reversed in "favor" of the Yoruba.

Third, both the British and New Zealand sample cohorts excluded nonwhite participants. If you wish to show that genetic differences across different races lead to different IQ potentials, you need actual data on each of the groups you are trying to make claims about. In other words, since the study was conducted entirely on white sample groups, the results of this study can't be used to make claims about differences between white and nonwhite populations.

In his article, Saletan presents his conclusions as if they are as solid as the legitimate study he is piggybacking on, when in fact they are of a different kind altogether. Rather, his conclusions are the mixed-up result of some hasty research in an unfamiliar field, seemingly in pursuit of evidence to support his broader thesis.

Resources

• The study (only the abstract is free): Moderation of breastfeeding effects on the IQ by genetic variation in fatty acid metabolism (Caspi et al, PNAS 2007)
• Study data (group IQ scores and genotype frequencies).
  
• To view the HapMap Project map for FADS2, go here and type "FADS2" into the search box. Or use this direct link.
  
• HapMap population frequencies for rs174575 and rs1535
  

Background

Posted above is a response I wrote to refute one specific assertion made by William Saletan in his series of three articles (1,2,3) which appeared on Slate.com between Nov. 18 and Nov. 20. Saletan's claim that a recent study linking breastfeeding, genes, and IQ can be used to show an innate genetic disadvantage for people of African descent is wrong.

Saletan's articles dealt with the subject of race and intelligence, and led to Saletan concluding that there is an innate intelligence gap between blacks and whites. Saletan's supporting evidence and methodology were quickly taken to task by the blogosphere as, at best, highly flawed and selective (some links collected below). The New York Times even picked up on the story here. So far though, no one (as far as I can tell) has addressed the IQ-breastfeeding claim. (Continue Reading...)

The blogs were right. In the end it was Stephen Metcalf at Slate who wrote the best point-by-point refutation of Saletan's arguments. The highlights include:

--Saletan heavily realized on the research of James Rushton, the current director of the Pioneer Fund, which was founded in 1937 to promote the principles of Nazi eugenics (really: see here and here), and has been steadily promoting racist "research" ever since--which did cause Saletan to write an apology once this was pointed out to him.

--Saletan implies that because intelligence is hereditable from person to person (to some degree), and that because there is an IQ difference across groups, then therefore there's an inherited difference between the groups. This does not follow at all: for example, in a malnourished population, height is still strongly hereditable (taller parents have taller children, shorter parents have shorter children), but once you feed everyone, the average height goes up. (Yes, I am stealing this example from Stephen Jay Gould).

--Saletan discounts studies that show no IQ difference between blacks and whites for irrelevant reasons. For example, he wrongly discounts a study showing no difference in the IQs of children fathered by black and white American soldiers in Germany after WWII.

--Saletan ignores or underplays evidence that the IQ gap in the U.S. between blacks and whites is closing. (IQ scores in the developed world have been rising steadily since testing began).

However, in all this, one assertion of Saletan's has remained unchallenged, so far as I can tell. It's his claim that data from a study linking the well-known boost in IQ scores cause by breastfeeding to a particular gene can be used to show that blacks tend to have this form of the gene at lesser frequencies, and are thus cognitively disadvantaged. Here's the relevant passage from his third article:

As my colleague Emily Bazelon explains, a new study shows that while most babies gain an average of seven IQ points from breast-feeding, some babies gain nothing from it and end up at a four-point disadvantage because they lack a crucial gene.

The study's authors claim it "shows that genes may work via the environment to shape the IQ, helping to close the nature versus nurture debate." That's true if you have the gene. But if you don't, nurture can't help you. And guess what? According to the International Hapmap Project, 2.2 percent of the project's Chinese-Japanese population samples, 5 percent of its European-American samples, and 10 percent of its Nigerian samples lack the gene. The Africans are twice as likely as the Americans, and four times as likely as the Asians, to start life with a four-point IQ deficit out of sheer genetic misfortune.

Don't tell me those Nigerian babies aren't cognitively disadvantaged. Don't tell me it isn't genetic.

I think that this specific claim has not received attention for a combination of several reasons: it was at the end of the last article in the series, and Saletan's arguments were mostly laid out in the first two articles; it is not the result of work by obvious racists or funded by hateful organizations, like many of his other claims, but is instead based on legitimate, interesting work by respected scientists; it is in an unfamiliar field (genetics); and it is hard to figure out where Saletan is getting his number from. Yet it is just as wrong as his other arguments, as I hope I have shown. And, I believe, it is no less important to show that it is wrong. Genetics, sadly, will probably be the next frontier for Pioneer-funded research and its like, and it's important to be aware that claims from genetics can not simply accepted at face value.


Some Blog Reactions To Saletan's Articles:
Brad Delong, Three-Toed Sloth: 1, 2, Talking Points Memo, Matthew Yglesias: 1, 2

Data

Expand this post to see the data from the article


The graphs for rs174575 are reproduced from similar graphs in the PNAS article. The rs1535 graphs are produced from data in the article, as are the tables at the bottom. The rs174575 graphs both show that you get an IQ boost from breastfeeding only if you have at least one copy of the "C" version of that SNP. The rs1535 British graph shows that you get an IQ boost from breastfeeding only if you have at least one copy of the "A" version of that SNP, but the New Zealand group's rs1535 data just shows a general IQ boost from breastfeeding.

About the Blog

Currently a management consultant in New York, I was a science major in college and am interested in how science is used in public discourse. I was motivated to write in response to William Saletan's article by what I saw as the take-it-at-face-value presentation of invalid conclusions about race built on publicly available genetic data used out of context.

A blog seemed the easiest way to make available what I had written.