Race and running: A reply to Carl and Sailer

Tade Souaiaia's reply, followed by Noah's comments.

Written by Tade Souaiaia.

Back in April, I and a few other scientists and athletes posted a preprint revisiting claims about race and running that were in vogue during our early running careers. While well received by other athletes and scientists, there has been a bit of push-back from some of the journalists responsible for popularizing those claims in the first place.

Jon Entine dismissed it as “scientifically silly”, while Steve Sailer fired off an impressive strawman. Quoting a line from the abstract that stated, “racial perception has greater impact on performance than racial physiology,” he curiously asserted:

The argument here is basically that if Nature can’t be proved beyond a shadow of a doubt to determine 100% of running results, then Nurture must be 100% responsible and race play no role whatsoever other than stereotype threat.

Personally, I’ve always argued for the interplay of nature and nurture, but the authors and their friends appear to be convinced that that’s cheating: you’ve got to pick one and only one: nature or nurture? Which is it?

Interesting. Some of his readers were gracious enough to provide free estimates of my IQ, excoriate me for typos (the pre-print process works!), and share with me a strange belief that counter-examples actually prove what they disprove!

But, I digress. One of the more measured critiques shared with me came from Aporia editor Noah Carl, who claimed our manuscript’s arguments aren't convincing, questioned our motives and criticised our analysis of the introductory data. While he didn’t contact me beforehand, he was gracious when I informed him that he had made a few errors. He even offered to consider a rebuttal. Well played. So in good faith, I will do this in three parts.

First, I will consider Carl’s objections to and (mis)characterizations of my work, and explain what he got wrong with respect to both track and field and statistical genetics. Then I will explain why the argument he does find convincing – let’s call it “the Sailer method” – is contradictory. Finally, because Carl didn’t really engage with the central arguments made in our manuscript, I will restate some of them and try to make them clearer.

Carl’s objections

The first thing that Carl takes issue with is our overview of the recent performance data, summarizing our argument as follows:

Souaiaia and colleagues point out that several of these records have since been broken by athletes without the relevant ancestry. For example, a Chinese athlete broke the 2004-record in the 110m hurdles. The argument here is that if non-West Africans can break records previously set by West Africans, there can’t be an inherent West African advantage.

Carl is right about this. Record-breaking performances from group B don’t make it impossible that group A has inherent advantages. But they certainly are contrary evidence!

Unlike with race, when analyzing sex differences in athletics we can appeal to a large body of compelling functional evidence that supports male advantage. However, it is in no way meaningless that male and female records do not overlap. Females breaking male records wouldn’t mean that there can’t be inherent male advantages, but it would absolutely justify that we question the model!

Most runners are aware that this has already occurred, albeit only in the relatively unpopular ultra-marathon. Here, a spate of races won by women and the general knowledge that the gaps narrow with distance has led reasonable minds to suggest that females may actually have an advantage in the ultras!

Putting my cards on the table, I am not convinced. Sample sizes are small and extrapolation can be dangerous. And while fat metabolism certainly matters more in the ultras, no one has shown that it completely mitigates the male advantage in aerobic capacity. With that said, outside of statistical genetics my expertise is confined to the track. So while I am not convinced yet, I remain open to the idea that female physiology is advantageous in the ultras. (And I wouldn’t write an essay lampooning a manuscript offering further evidence in support of the hypothesis.)

Carl disagrees, and follows his point with some interesting data analytic advice:

But this is a rather arbitrary way to analyse the data. As Steve Sailer notes, a much simpler method is to look at the top 100 all-time performances in each event, and ask what are the ancestries of the athletes who achieved them. Sailer finds that in the 100m, 97 of the top 100 times were achieved by men with West/South African ancestry. And in the marathon, 96 of the top 100 times were achieved by men with East African ancestry. In other words, West/South Africans still overwhelmingly dominate sprinting, while East Africans still overwhelmingly dominate long-distance running.

The first issue here (putting aside the fact that determining “methodological arbitrariness” is itself arbitrary) is that our aim was to revisit stereotypes based on disparities in performance. We are well aware that disparities exist, but just as disparities do not prove discrimination, they do not prove differences in inherent genetic ability either. For this reason we focused on change over time. Carl’s criticism only makes sense if one accepts a priori that conditions are unchanging.

Second, Carl reveals an unawareness of the historical stereotypes that we were reexamining when he writes that “West/South Africans still overwhelmingly dominate sprinting.” As we document in our manuscript, such a claim was neither widely believed nor reflected in data before the 21st century. In the words of Amby Burfoot, one of the first science communicators to popularize the narrative of racial specialization:

South African blacks are related to East African blacks through their common Bushmen ancestors. West African blacks, representing the Negroid race, stand apart.

Not being a historian of the sport, Carl may not have been aware that until quite recently South Africa was known for producing world record distance runners, rather than sprinters. And this was not just expressed by runners in Burfoot’s dated vernacular. The 1990s produced multiple scientific studies that purported to explain South African fatigue resistance! Carl does pay some mind to prevailing wisdom about East Africans by explaining away the significance of the African world record holder being East African rather than West African with this strange appeal to linguistics:

For example, the African 100m record is held by an athlete from Kenya. Yet this is unpersuasive, since the athlete in question, Ferdinand Omanyala, is ethnically Bantu. The theory isn’t that particular places have a causal effect on sprinting ability; it’s that particular ethnic groups have traits that make them better or worse at sprinting. Souaiaia and colleagues note that more of the top sprinting performances in the last five years have come from Kenyan or South African athletes than from Nigerian athletes. But this is easily explained by the fact that Kenya and South Africa have large Bantu populations.

Admittedly, I did not initially understand this argument. But I think I get it. It is true that Omanyala belongs to an ethnic group (Luhya) that speaks a Bantu language. So too does Sydney Maree, the record-setting distance runner who is frequently referenced as evidence of South African excellence in the endurance events. And so do hundreds of millions more. Because south of the Sahara, Bantu languages are dominant everywhere but West Africa - where Niger-Congo languages are spoken. Of course, it’s not just language that separates West Africans from East African speakers of Bantu languages. There are major genetic differences as well, seen in this 2023 work by Kevin Esoh and colleagues showing that Luhya separate from West Africans on the first two principal components when analyzed alongside Cameroonian Bantu speakers.

Source.

Why does Carl point out that Omanyala’s ethnic group speaks a Bantu language? Reading the Steve Sailer article that Carl cites provides a clue:

There is now one East African top man in the 100m, Ferdinand Omanyala of Kenya. Yet, note that he is of the Bantu tribe, which is quite distinct from the Nilotic Luo tribe of former president Obama. Omanyala is much wider and more muscular than the Kenyan Nilotic Kalenjin distance runners, such as marathon great Eliud Kipchoge.

It appears that Sailer decided by looking at him that he is too physically big to be a traditional East African, categorised him as part of the non-existent “Bantu tribe” and assumed that this makes him more genetically similar to a West African. Carl then repeated the error. Interestingly, this post-hoc racial designation is only ever made in one direction. Sailer isn’t perplexed by the fact that Ben Azamati (the 135lb Ghanian 100m record holder) looks more similar to Kipchoge than Omayala.

We can find more evidence that Sailer’s claim is wrong in publicly available data from the 1000 Genomes and Simons Diversity Project. Here we observe that the Luo (and the Somali, Dinka and Tswana) indeed cluster with Luhya. While Carl is right that “particular places” don’t determine genetics, in this case they approximate it far better than the eye-test.

Source: own chart.

Given the usual caveats around PCA and genetic clusters, what we see is that the East-West clinal variation in Africa doesn’t make exceptions for speed. Thus, if the turn-of-the-century racial hypotheses had been “preregistered”, it is certain that the records set by south African athletes like Letsile Tebogo and Wayde van Niekerk (who derives the majority of his recent ancestry from South Asia) would absolutely have not been predicted.

In closing this section, it’s worth mentioning the recent performances of a junior athlete from Australia by way of South Sudan. Gout Gout, who recently became the fastest sixteen year old over 200m and the second fastest over 100m (after Puripol Boonson from Thailand), hails from a Nilotic speaking ethnic group, which would require Sailer and Carl to classify him as distinctly non-West African. While performances like these might make this fine grained analysis of genetic background seem unnecessary, it is important to illustrate how confirmation bias and misclassification work together to uphold racial stereotypes.

Sailer’s contradiction

Far more important than the error discussed above is the fact that Sailer’s argument, which Carl supports, isn’t genetically coherent to begin with. To understand why, consider Carl’s response to my complaint that despite multiple studies and interested parties, no genetic variants of large effect have been discovered for speed:

Souaiaia and colleagues complain that “attempts to locate "speed genes" have largely failed”. And while this is in part because scientists naively put too much faith in the candidate gene approach, it is also because of papers like the very one Souaiaia and colleagues wrote, which casually deploy terms like “racialist” and “scientific racism” in an effort to smear those who subscribe to genetic explanations for racial differences in athletic performance.

While I am flattered to be blamed for the field’s failure to find such variants, it seems that Carl knows enough about genetics to know this is an empty charge. Right before, he correctly states that “rather than having one single trait or gene, these athletes likely have a combination of traits that give them an edge over their competitors.”

This suggests that he knows that there are no large-effect variants for most complex traits, and that it’s probably because they don’t exist rather than because they are kept secret due to politics. As an example, consider that despite human height being both politically benign and extensively studied, biobank studies can’t identify even a single variant capable of explaining more than a trivial share of the trait variance.

What does this mean in layman’s terms? It is true on average that the Dutch are taller than the Portuguese and it is also true that this is partially due to genetics, but this is not because there exists a single height-conferring gene that is more prevalent among Dutch people. Nor is it due to two or three genes, but rather thousands of variants – or according to recent population scale studies over twelve thousand!

This means that polygenic traits like height and skin color are not inherited from your mother or your father but from your mother and your father. For example, somebody who inherits extremely dark skin almost surely does so from two dark skinned parents rather than one. These types of traits can be contrasted with monogenic traits like Marfan syndrome or Duchenne Muscular Dystrophy (DMD), which can be inherited from either parent.

The reason why it matters that no variants with large effects have been identified with respect to running speed isn’t that genes have no role whatsoever. It’s that any role genetics plays occurs through a polygenic aetiology.

So where would we expect to find the top 0.000001% of sprinters? Would we expect to find them, as Sailer suggests, among the minority of men who have at least one parent of primarily Sub-Saharan ancestry? Absolutely not. Just like we would not expect the darkest skinned people in the world to have at least one dark skinned parent. But this is exactly what Sailer proposes and what Carl says is correct.

The data we have is arguably more consistent with a cultural transmission model. Consider how rare it is to find Americans capable of speaking Chinese before adulthood – unless they have at least one Chinese parent. Anybody who knows track and field is aware that almost all the “Sub-Saharan” sprinters in England and Canada, a large fraction of American sprinters, have one or both parents from the Caribbean. As mentioned in the manuscript, even the West African 100m record is held by an athlete with Nigerian father and a Jamaican mother!

What is the point?

So far I have dealt with what Carl and Sailer got wrong. Perhaps I should start this section on where we agree. Carl spends some time describing the morphological characteristics of East Africans and arguing that this is where their advantage in long distance running derives from:

What’s more, we actually have a decent idea of why East Africans do so well at long-distance running. Like all Sub-Saharans, they have a narrow pelvis and long limbs for their height, giving them a larger ratio of surface area to total bodymass (see below). Their body proportions appear to be a function of the equatorial climate in which they evolved, as predicted by Allen’s rule. East Africans also have low bodyfat and an ectomorphic (slender) physique …

As the authors of a 2017 paper note, “East African runners appear to have a very high level of [running economy] most likely associated, at least partly, with anthropometric characteristics rather than with any specific metabolic property of the working muscle”

Here, we are not in total disagreement! (Although note that while East Africans living in East Africa indeed have low body-fat, this phenotype is not maintained when they move to Western countries.) Let’s consider two statements where we almost surely agree:

1. There are genetically mediated morphological advantages such as relatively longer legs that are likely advantageous for many running events.

2. These characteristics vary globally. The “tropical body type” that predominates in many warm weather environments is characterised by longer limbs relative to height.

While I would urge caution with respect to the first statement when it comes to short distances, and outliers exist (even in the hurdles), there is functional evidence that relatively long legs are advantageous in the endurance events.

Given that the second statement is incontrovertible, is it possible that African countries have a larger pool of potential athletes relative to their population size? Certainly! My position is not that all morphological traits are equally distributed between groups but that: (1) “race” doesn’t impart a hidden essential characteristic that determines performance, (2) recent data calls into question turn-of-the-century claims about race and running, and (3) the disparities in performances over the events that we analyzed (100m and 1500m) are better explained by a psychocultural model.

Focusing on (3), what evidence do we have? For starters, the 1500m was dominated by Northern Europeans during the 1980s, by African athletes for the three following decades, and is today a diverse event again led by Northern Europeans. (Since 2020, every global champion is Northern European and 62 of the top 100 performances are by such athletes). As we explain in the manuscript, scientific and popular interest during these three decades of African dominance was so focused on genetic factors that the substantial slowing of European 1500m performances was hardly noticed!

Looking through the history of sports science, every analysis of disparate performances has assumed that disparities in previous generations were driven by lack of access/interest or other sociocultural factors and that current disparities are final and reflect inherent stable advantages.

As athletes and coaches ourselves, we are well aware of how small the gaps are between elite athletes, and just how much individual psychology matters. Hence we set out to quantify the advantage that speedy competitors impart on their rivals in both the 100m and 1500m races. We identified a quantifiable benefit to the “rival effect”, one that is roughly equal to a meter per second tailwind over 100m (though it was not present for female sprinters). The benefit to a speedy rival was not only present over 1500m but varied by decade depending on the continental relationship of the competitors. In summary, we found that during the period of African dominance, European athletes benefited significantly less from “rival effects” when the rival was African, which is no longer true today!

Unfortunately neither Carl nor Sailer engaged with this portion of the manuscript, and focused instead on the disparities across other events like the marathon. As stated previously, it is not so much that we reject the possibility that these disparities reflect genetic differences between geographic groups but that they alone constitute proof such differences are a key factor.

We remain open to descriptions of athletic performance disparities that integrate both functional genetics and psychocultural effects, with the caveat that we remain unconvinced that all major social shifts are a thing of the past.

Tade Souaiaia is professor of biology at SUNY Downstate Health Sciences University and a former athlete.

---

Our comments

Written by Noah Carl.

I am grateful to Tade Souaiaia for responding in good faith to my article criticising his and his colleagues’ arguments about race and running. While he makes a number of valid points, he hasn’t persuaded me there aren’t “racial” differences in sprinting ability. And confusingly, he seems to lend credence to hypotheses he previously characterised in pejorative terms.

To begin with, I have to concede that attributing Ferdinand Omanyala’s success to him being “ethnically Bantu” was erroneous—for the reason Souaiaia gives. However, I don’t think this error substantially affects my argument. The issue is whether particular ethnic groups have traits that make them better or worse at sprinting. It may be that the Luhya have such traits even though they are native to East Africa. (Omanyala does not hail from one of the East African groups that excels at long-distance running.)

When it comes to sprinting, Souaiaia insists that “the data we have is arguably more consistent with a cultural transmission model”. To me, this claim just doesn’t seem justified. His point that we’d expect the best sprinters to have two black parents, rather than at least one, is well-taken. Yet as I noted in my original article, genetics isn’t the only relevant factor. Countries like the US and to some extent also Jamaica (whose living standards are twice as high as Nigeria’s) have systems of talent identification and infrastructure for training that give their athletes a competitive edge. Skin colour isn’t a close analogy because you don’t have to train to get extremely dark skin.

And even if the very fastest men’s 100m times are dominated by athletes from the US and Jamaica, West-African countries (where European admixture is basically nil) are still well-represented in the top 4,047 times. There are 189 entries for Nigeria, 61 for Ghana, and 42 for Ivory Coast. By contrast, there isn’t a single entry for India, Pakistan or Bangladesh (countries with a combined population of almost 2 billion people). Does Souaiaia believe the absence of these countries is entirely due to culture, or would he concede that South Asians have traits that make them worse at sprinting than West Africans?

Language isn’t a close analogy either. While learning Chinese requires a huge investment of time and effort, everyone knows how to sprint. Yes, you have to train hard to be the best. But the barriers to entry into sprinting are among the lowest for any sport. Which raises the question: why hasn’t the sprinting culture that supposedly benefits black people across Jamaica, the US, Japan and elsewhere rubbed off on the non-black populations of those places?

When it comes to long-distance running, Souaiaia says “it is possible that African countries have a larger pool of potential athletes”. So the two of us agree, although I would say “East African countries” (there isn’t a single entry for Nigeria, Ghana or Ivory Coast in the top 5,408 men’s marathon times).

What’s more, Souaiaia says he doesn’t reject the possibility that “disparities reflect genetic differences between geographic groups”. Well then why did he and his colleagues refer to “racialist explanations of athletic success”? He now seems to be saying the “racialist explanation” for East African success in long-distance running could be right after all.

There is certainly less disagreement between Souaiaia and myself than I had initially assumed. But I think I can be forgiven for my assumption given some of the language in his and his colleagues’ paper.

Noah Carl is an Editor of Aporia.

---

Support Aporia with a paid subscription:

You can also follow us on Twitter.