At the request of my co-blogger Mike, I’m reposting this article which originally appeared at Science 2.0 on 30 December 2008 where some authors of the paper in question respond in the comments during the run-up to the publication of their book The 10,000 Year Explosion.
Unfortunately for all of us still breathing braniacs, the title only applies to those of us who are also medieval Ashkenazi Jews, according to the authors of the 2006 paper “Natural History of Ashkenazi Intelligence”.
Discussing “race” and intelligence is always a touchy subject and definitely not politically correct; but science should not be fettered by the chains of political correctness like a mangy circus lion. It must run free across the intellectual savanna, striking down the juvenile wildebeest of ignorance. Following articles on the biology and significance of race by Michael White, Massimo Pigliucci, and moi, my attention was directed to “Natural History of Ashkenazi Intelligence”. Thanks to press attention from biological research bell-weathers like The Economist and the New York Times, as well as discussion on National Public Radio, this paper has gained Goodyear AquatredTM-esque traction on the internet.
CHH were trying to explain a long-standing dilemma: the relatively high frequency of alleles for genetic disorders in Ashkenazi populations. One hypothesis is that the disease allele frequencies were the result of one or more founder effects (changes in allele frequency by chance due to small effective population size) in Ashkenazi history. An alternative hypothesis is that selection for a character on which the disease alleles have an overdominant or heterozygote advantage effect (heterozygote has higher character value than either homozygote) explains the relatively high disease allele frequency.
Have CHH found the solution? They claim that the relatively high disease allele frequency is explained by positive, natural selection for intelligence, on which the disease alleles have an overdominant effect, in the medieval Ashkenazi. They suggest that cultural factors forced the Ashkenazi to work in certain professions (e.g., finance), which required above average intelligence and were associated with high income. Intelligence is associated with wealth, and wealth is associated with number of viable offspring, leading to natural selection for increased intelligence.
We’ll analyze their theory in three steps. First, we’ll look at the validity of their premises. Second, we’ll look at the population genetic evidence. Third, we’ll look at the plausibility of the proposed mechanism. In conclusion, I’ll follow up with some, potentially snarky, commentary.
DISCLOSURE: My analysis has been based on a pre-publication PDF available online, because the Washington University School of Medicine library does not subscribe to the Journal of Biosocial Sciences. I’ve tried to reduce the bulk of technical definitions in the text by linking out to Wikipedia pages. I cannot vouch for the completeness nor the accuracy of all referenced Wikipedia entries.
Enough foreplay, let’s take a look under the hood.
PREMISES: There are four observations that motivated the analysis of CHH.
The Ashkenazi have higher frequencies of alleles for several genetic disorders than expected
The allele frequency for a number of genetic disorders is significantly elevated in the Ashkenazi population. This observation has motivated a great deal of research into its cause.
There are more lysosomal storage diseases (LSDs) represented in the Ashkenazi disease set than expected
The apparent enrichment of LSDs in the Ashkenazi (Jorde ) has motivated selection-based explanations for these allele frequencies. It has been postulated that selection, not random processes, was the only reasonable mechanism that could produce such enrichment. Furthermore, the LSD disease alleles affect neuronal pathways, evidenced by their deleterious effects on the brain in homozygous recessive individuals, suggesting the possibile overdominant effect on intelligence in heterozygotes.
Is this enrichment special? In 2003, Risch et al. compared the frequency distribution, number of mutations, and coalescent times of LSDs and non-LSDs (NLSDs) and found no difference between LSDs and NLSDs, demonstrating that LSDs are a representative subset of the total set of genetic disorders present in the Ashkenazi. They are not special.
CHH do not counter the argument of Risch et al. (2003) directly. Instead, they argue that NLSDs are also a special caseMG. They speculate, without evidence, that a subset of NLSDs also have effects on intelligence. The reasoning is circular. This observation (enrichment of intelligence affecting disease alleles) only exists if one accepts the mechanism (NLSDs affect intelligence) that only exists to explain the observation. The mechanism is its own grandfather.
The Ashkenazi have a higher mean IQ than other populations
According to Lynn and Longley (2006), the median IQ of British Jews is 110 and American Jews is 109.5, with the global population mean defined at 100 with a standard deviation of 15.
Environmental contribution to IQ is negligible
Adult IQ has been shown to have a high narrow-sense heritability (h2~0.8). The importance of environmental effects on IQ, however, is still debated. CHH’s interpretation is that there is minimal environmental influence on IQ. Identical twin and adoption studies appear to support this interpretation, if one ignores the potential environmental influence of culture. Narrow-sense heritability reflects the ability to predict the phenotype of the offspring from those of the parents. It reflects the genetic contribution to phenotypic variation within populations, but not necessarily between populations. It has not been established that the observed, average IQ in the Ashkenazi is a genetic phenomenon. This issue would be relevant even with perfect heritability (h2=1).
At least, two of the four premises of this study are suspect. While the Ashkenazi do have higher than expected frequencies of several disease alleles, the evidence that there is an enrichment for LSDs is, at best, debatable. Although the data do support a higher than average measured IQ in the Ashkenazi, the application of IQ heritability for individuals to entire ethnic groups has not been demonstrated to be a valid assumption. The assumption that Ashkenazi IQ must represent differences in allele frequencies, not differences in culture, controlling IQ is not supported at this time.
GENETICS: CHH make two claims critical to their theory from population and quantitative genetics.
Response to selection could reasonably account for the elevated mean IQ of the Ashkenazi
The argument for selection on IQ begins with a string of correlations. CHH do not describe this string of correlations (Figure 1). High IQ is correlated with certain modern professions (r1). Those modern professions are correlated with medieval professions (r2), to which the Ashkenazi were almost exclusively restricted. Those medieval professions were correlated with high income (r3). High income is correlated with number of offspring (r4). They assume that a modern correlation of IQ with income (rP=0.4) is reasonable for Medieval Times, combining r1, r2, and r3 into one correlation, rP. If any one of these correlations is invalid, their assumption of rP will also become invalid. They assume that income and fitness are interchangeable (r4=1).
CHH use these assumptions to estimate the response of IQ to natural selection. They want to predict the response of a correlated character (IQ) to natural selection on fitness. It would be convenient if someone had come up with an equation to deal with this situation. Well, I don’t have a copy of the edition (2ND) of An Introduction to Population Genetics used by CHH, but I do happen to have an edition (4TH) updated by some chick that works on fruit flies to update (Falconer and McKay ). My copy has just such an equation.
Don’t feel bad if you missed it (it’s kind of near the back). CHH missed it too. Instead, they fall back on an equation (much closer to the front) used to predict the response to artificial selection in domestic animal breeding. Imagine taking IQ tests for the right to have sex, which sounds like eugenics (or heaven if you are a socially awkward academic with a genius IQ).
When the process is actual treated as natural selection, the expected response to selection (IQ=144) is unreasonably greater than either CHH’s estimate (IQ=120) or the claimed, observed difference (IQ=115), if one uses the appropriate equation (Box 1).
The Ashkenazi have experienced no founder effects
CHH admit that the high disease allele frequencies could be accounted for by one or more founder effect events, if those founder effect events occurred. Risch et al. (2003) and Slatkin (2004) say they did. CHH say they didn’t. CHH base their claim on the similarity of 652 polymorphisms at 251 loci between the Ashkenazi and Europeans, not including the disease alleles in question. Unsurprisingly, they find that polymorphism frequencies are most similar between geographically close populations. They argue that the Ashkenazi are essentially European, not Middle Eastern refuting any founder effect hypothesis. Their analysis, however, does show that the Ashkenazi are closer to Middle Eastern populations than are Europeans, consistent with a migration from the Middle East and a low rate of gene inflow from Europe. Because founder effects are chance changes in allele frequency, one would expect most allele frequencies to be similar between populations (especially if one throws out a set of the alleles with divergent frequencies, as CHH have done), even after a founder effect event. This leaves things a bit murky.
Fortunately, there are better markers available, such as the Y-chromosome, mitochondrial inheritance, haplotype blocks, and linkage disequilibrium. Hammer et al. (2000) found that the Ashkenazi were derived from a Middle Eastern source population, based on Y-chromosome haplotypes. Behar et al. (2006) found that half of extant Ashkenazim can be traced back to only four women, based on mitochondrial DNA. Olshen et al. (2008) found distributions of haplotype blocks and linkage disequilibrium that were consistent with a founder effect, based on the analysis of 435,632 polymorphisms.
The preponderance of quality data continues to support the founder effect hypothesis.
CHH’s conclusion that Ashkenazi IQ can be reasonably explained as a response to selection does not hold up when the appropriate analysis is applied. The data do not support CHH’s conclusion that there have been no founder effects in the Ashkenazi population history. Chance effects from founder effect events remain the most likely explanation of elevated disease allele frequencies among the Ashkenazim.
MECHANISM: CHH propose that overdominant effects on intelligence drove elevated disease allele frequencies.
Wild speculation is really all there is to say. Some of the Ashkenazi diseases affect neurons, detrimentally. Does this make it plausible that they all have overdominant effects on intelligence? Not really. Evidence would be appreciated.
Occam’s razor directs us to, when faced with two equally evidenced theories, to prefer the one that requires the fewest assumptions. In this case, the theories are not equally evidenced, but let’s set that aside for the moment. The founder effect hypothesis requires the assumption of 1-3 founder effect events. CHH’s overdominant selection hypothesis requires the assumption of 4-8 overdominant effects on intelligence by deleterious, recessive alleles. Parsimony favors the founder effect hypothesis.
Although these disease alleles have elevated frequencies in the Ashkenazi population, they are still rare (minor allele frequency < 0.05). In order to be responsible for an average Ashkenazi IQ 15 points higher than other populations, these alleles would have to have quite significant effects. CHH estimate that Gaucher, Niemann-Pick, and Tay-Sachs heterozygotes should have an IQ 5 points above average, which is likely underestimated (Box 2). The effects on IQ of heterozygosity from any one of these disease alleles based on either CHH’s or my own calculations should be easily detectable by standard IQ testing. This is the hypothesis that should have been tested, but is not.
There is no evidence to support CHH’s overdominance hypothesis. The hypothesis is not parsimonious. Finally, the required effect sizes of the alleles on IQ should be readily detectable in heterozygotes by standard testing. While we can conclude nothing from the absence of data, this hypothesis should be testable and should have been tested before being promulgated based only on speculation.
DISCUSSION: This is where I get to be snarky.
The claim that these recessive alleles have an overdominant effect on IQ means that CHH are suggesting that they have identified high probability loci controlling quantitative variation in human intelligence. People have been looking for these loci and have found squat. This should have been huge, HUGE, the news of 2006. And the response from the scientific community . . .
[Insert sound of crickets chirping melodiously]
Uh, science, maybe you didn’t you hear me, HUGE; or, it would have been, if there was, you know, evidence. There is none. Zero. Zilch. Nada. What do they have? Well, they can show that recessive alleles that negatively affect the brains of the afflicted, well, affect brain cells. The steps from lethal, recessive effects to beneficial, IQ-enhancing, over-dominance effects is probably too obvious to you for me to bother explaining.
What’s really missing? Demonstrating that heterozygote carriers for any of the recessive diseases they address have higher IQs than wild-type homozygotes. This hypothesis is so testable it hurts. Everything else in this paper should have been a build up to data comparing the IQs of homozygotes and heterozygotes. Instead, we just get rapant speculation.
Selectionism: n; the tendency to describe all evolutionary phenomena as a result of selection
Now for a word about style: don’t pop your collar and don’t propose a mechanism before demonstrating that the mechanism is needed. CHH spend their effort establishing the plausibility of selection for intelligence (over 70% of the paper), before addressing whether selection is even needed (30%). Could this illogical structure betray selectionist tendencies?
One of the authors, Gregory Cochran, has provided outside confirmation that he is indeed a selectionist. The opening paragraph of Cochran’s “An Evolutionary Look at Human Homosexuality” reveals the way this author views evolution:
The first key idea is that evolution optimizes function. It makes things work. If there is a change in a gene that helps the organism, that change increases over the generations, becomes more common. If it causes trouble, as most changes do that make any difference at all, the change decreases over the generations, becomes rare. So although changes (mutations) happen, they don’t happen very often, and natural selection tends to keep them rare. You could think of it as a filter, constantly removing changes that don’t work. It magnifies the rare changes that improve things.
The existence of deleterious alleles in human populations is troubling to this worldview. Selection should have eliminated deleterious alleles. If alleles are beneficial, they increase in frequency. If they are deleterious, they decrease in frequency. In the selectionist worldview, evolution is simple and deterministic. Evolution is constantly moving us onward and upward (see here for a discussion, on a different topic, of how drift can allow deleterious variants to increase in frequency).
Selectionists find the solution to the existence of deleterious alleles in human populations in the sickle-cell anemia story. Sickle-cell anemia is caused by a recessive mutation in beta-hemoglobin. Homozygous recessive individuals get sickle-cell anemia. Heterozygotes are resistant to the malaria parasite (Plasmodium falciparum). Due to strong selection for malaria resistance in Sub-Saharan Africa, up to one-third of Sub-Saharan Africans carry the mutant allele.
Take the sickle-cell anemia story, apply it to everything, and see the world through the eyes of a selectionist. To a selectionist, deleterious alleles can only persist if there is a strong, countering selection for heterozygotes. As a result, a selectionist tends to see a pathogen wherever fitness-reducing variants are perceived. Except that the pathogen theory does not work for the Ashkenazi. Medieval Ashkenazim were physically intermixed with other Europeans. Something had to take the place of the pathogen. Enter IQ.
CHH appear to have been committed to a selection-based explanation for disease allele frequencies in the Ashkenazi. They were not interested in testing their hypothesis (e.g., measuring the IQs of heterozygotes). They were interested in showing that their hypothesis is plausibly and creating a speculative story to support a selection-based explanation. They fail at both.
Thanks Neil Patrick Harris and Doogie Howser, MD for first introducing me to Tay-Sachs and the concept of population genetic history influencing disease risk. This is certainly the most parentheses that I have used in a document (ever).
MG: moving goalposts logical fallacy