Yes I think about 1% of Afrikaner ancestry is probably Khoikhoi

As a follow-up to my previous two posts on Afrikaners, I wanted to reiterate something that I implied/said earlier: yes, I think about 1% of the ancestry of modern day Afrikaners derives from Khoisan pastoralists of the Cape who were resident there when the Europeans first arrived. These people are often called Khoikoi. Unlike the more famous Bushman the Khoikhoi were not hunter-gatherers. Rather, they herded cattle. Both archaeological and genetic evidence points to the fact that pastoralists arrived in southern Africa through the expansion of East African nomads, who had some Eurasian ancestry (ergo, Khoisan peoples have differing degrees of non-Khoisan African ancestry, as well as Eurasian ancestry).

Today there are no major Khoikhoi groups in South Africa that have not been extensively influenced by other populations (in Namibia the related Nama maintain tribal cohesion and continue the cultural tradition of Khoisan pastoralism). Where did the Khoikhoi go? Many died due to disease, and the privations of slavery. But, some were certainly absorbed into other populations. The Xhosa people have substantial Khoisan ancestry for example.

The plot to the left has various populations, including Dutch, whites from Utah, white South Africans, Nigerians, African Americans, Barbadians, and Bantu populations (click the image for a larger version). As well as Khoisan groups which are a combination of Nama and San Bushmen samples.

If you click the larger image you can see that the South African Bantus are shifted toward the Khoisan. The Kenyan Bantus are skewed in the direction of Eurasians…though only mildly so (no doubt due to Cushitic admixture).

The plot to the right (click to enlarge) is a zoom in. It is clear that the South African samples are very subtly shifted out of the normal Northern European cluster. If you look at the cline from the Nigerians running toward the Northern Europeans, the South African whites look to be perturbed from it. Notably, some of them are clearly shifted in the direction of the Khoisan.

Next, I ran Admixture analysis. I set the reference populations as Esan from Nigeria, Khoisan, and Dutch whites. You can see that African Americans exhibit a cline as you’d expect. A minority of their ancestry is Northern European. But mostly they are African, with the dark blue representing the Esan Nigerian reference population. This is as it should be; most of the slaves who came to America seem to have come from the Congo up the Africa coast all the way to Senegal.

The fraction of African ancestry in the South African samples is low. But observe that many of them have just as high a fraction of the red component, which comes from the Khoisan reference population. These ten mostly white South Africans average 1.4% Khoisan and 2.3% non-Khoisan African.

Finally, I decided to run Treemix and do a three population test.

With two migration edges the results make a lot of sense. The African Americans are placed next to the Nigerians, but there is a migration edge of some significance from the Northern Europeans. The South Africans are in a clade with the Dutch samples, with Utah whites being the outgroup. But, they have a migration edge from between the Esan from Nigeria and the Khoisan.  Recall that there was more Nigerian-like ancestry in the South African whites than Khoisan-like ancestry according to Admixture. The gene flow edge seems to be closer to the Esan by some margin.

Finally, I ran a three population test, which tests gene flow by placing an admixed population as an outgroup to source populations. Negative statistics indicate “complex population history” not accounted for by the tree.

Outgroup Pop 1 Pop 2 f3 Z score
Af_American Netherlands EsanNigeria -0.0103 -89.1922
Af_American UtahWhite EsanNigeria -0.0102 -88.7189
Af_American South_Africa EsanNigeria -0.0099 -87.1784
Af_American Netherlands Khoisan_SA -0.0034 -16.6754
Af_American Khoisan_SA UtahWhite -0.0033 -16.5408
Af_American South_Africa Khoisan_SA -0.0029 -14.6855
South_Africa Netherlands EsanNigeria -0.0015 -10.9174
South_Africa Netherlands Khoisan_SA -0.0015 -10.5677
South_Africa UtahWhite EsanNigeria -0.0014 -9.0416
South_Africa Khoisan_SA UtahWhite -0.0014 -8.7962
South_Africa Netherlands Af_American -0.0011 -10.0158
South_Africa Af_American UtahWhite -0.0010 -8.1132
UtahWhite Netherlands EsanNigeria -0.0001 -1.6344
UtahWhite Netherlands Khoisan_SA -0.0001 -1.6344

The bottom two results can be ignored. What you see is that African Americans have the most negative f3 values with the highest z-scores. There is a drop-off from the Nigerians to the Khoisan as one of the source populations because the Nigerians are a much better fit. The values for South Africans are much lower, which makes sense in light of their lower admixture proportion. But observe that the f3 statistic for using Esan vs. Khoisan is not that different. This suggests neither group is necessarily a better proxy for the other.

As for the ethnographic details of where this ancestry came from, I think it was the proto-Cape Coloured population.

No, Afrikaners do not have British or English ancestry

Update: Some people are taking this post as a criticism of 23andMe. Really it’s not. It’s just to point out that customers sometimes overinterpret the granularity of these regional tests. There just isn’t the power to discern between British and Dutch too well (there are other ways to do this genealogically….). And, it really matters in the case of Afrikaners since their Dutch (and German and French) national origins are well known.

kan

In my post below on the non-European ancestry of Afrikaners, several readers mentioned that friends of Afrikaner background were rather chagrined to have reported British ancestry from genetic tests. The cultural reason for this is well known: many Afrikaners exhibit hostility toward British imperialism due to the deprivation and death which was the consequence of their resistance to the expansion of the Empire during the Second Boer War. This is above and beyond the antipathy which was manifestly made obvious by the fact that with the transfer of the Cape Colony to the British in the early 19th century thousands of white farmers migrated into the hinterlands to escape the new power (in part to preserve their customs, such as slavery).

By the 20th century, this anti-British aspect of Boer identity manifested itself in pro-German sentiments, as can be seen in the film The Power of One.

But the reality is that it is strange for Afrikaners to have British ancestry. Yes, they are not exclusively Dutch, with substantial German and French (Huguenot) components in their background. And there has been some recent intermarriage with English speaking whites. But presumably that’s recent enough that people would know.

Rather, I think what is happening is that genetic tests do not have the power to distinguish well between English and Dutch ancestry. In fact, the minority ancestry from Anglo-Saxons in southeast Britain would have stronger affinities with the Dutch than most of the island.

To figure out what was going on I asked people on Twitter for 23andMe profiles. I got a response from someone whose results I posted above. This individual has Boer ancestry, mostly Dutch, going back to the late 17th century on his mother’s side and late 18th century on his father’s side. And you see 17% “British” ancestry. He also provided his wife’s 23andMe output. Her ancestry dates back to the late 17th century on both paternal and maternal sides, so it is not a surprise she has more non-European ancestry:

She is 18% British. In fact, the European ancestry fractions of both these individuals are rather similar when it comes to “French-German”, British, and Scandinavian. I suspect what we’re seeing here is what the algorithm pops out quanta wise for Dutch.

I took the South African individuals who had some non-European ancestry, and ran them on Admixture and projected a PCA with British and Dutch individuals. You can make your own judgment, but I think these are definitely people who are of mostly Dutch ancestry.

Read More

The non-European ancestry of Afrikaners


A few years ago I got some South African genotypes. Some of the individuals were clearly African. A few mapped perfectly upon Northern Europeans. But many of the samples consistently were European but shifted toward non-European populations.

Based on history of the assimilation of slaves into the European population of Cape Colony in the 18th century, my assumption is that these individuals are Afrikaners.

Recently I realized that Brenna Henn had released some more Khoisan samples, so I decided to look at this question of admixture again. The two Khoisan populations are the Nama and the Khomani. I removed those with lots of Bantu and European admixture and combined them together into one population.

Running unsupervised Admixture shows how distinct the South African whites are.

The average Utah white in this sample (this population is a mix of British, German, and Scandinavian in ancestry) is 99% European modal cluster, and 1% South Asian. The average for the white South Africans in this data set is 94% European modal cluster. The residual is 1% East Asian (Dai modal), 1% Khosian, 1% non-Khoisan African, and 2% South Asian.

I ran Treemix a bunch of times, and every single plot came out like this when I ran it for three migrations:

 

The gene flow from the Utah whites to the Gujuratis is simply an artifact of the fact that the Gujurati sample is mixed caste, and some of the Brahmin or Lohannas have more “Ancestral North Indian.” The gene flow from the Europeans to the Khoisan is probably real, or, might be due to pastoralist admixture via East Africans. The last migration arrow goes from the African populations to the South African whites, with a shift toward the Khoisan.

I also ran a three population test where A is the outgroup, and B and C are a clade. A significantly negative f3-statistic indicates admixture in population A. The negative values are listed below:

A B C f3 f3-error Z-score
Gujrati Dai UtahWhite -0.00121718 0.000140141 -8.68539
South_Africa EsanNigeria UtahWhite -0.00127718 0.000147982 -8.63059
South_Africa Khoisan_SA UtahWhite -0.0012928 0.000151416 -8.53802
Gujrati South_Africa Dai -0.000778791 0.000155656 -5.00329
South_Africa Dai UtahWhite -0.000541974 0.000133262 -4.06699
South_Africa UtahWhite Gujrati -0.000103581 8.46193e-05 -1.22408

This aligns well with the Admixture results. Afrikaners have both African ancestries, and, Asian ancestry.

In James Michener’s The Covenant one of the plot lines alludes to mixed ancestry in one of the Afrikaner families. The results above suggest that mixed ancestry is very common, and perhaps ubiquitous, in this population. True, there are some Afrikaners such as Hendrik Verwoerd who migrated to South Africa from the Netherlands in the past century or so, but these are uncommon to my knowledge.