The great Eurasian explosion

By Razib Khan | September 17, 2012 9:38 pm

Dr. Joseph Pickrell has updated his preprint, The genetic prehistory of southern Africa, with some more material on the Sandawe. I’ve explored the genetics of the Sandawe a bit using ADMIXTURE, so I jumped straight to the section on ROLLOFF:

…To further examine this, we turned to ROLLOFF. We used Dinka and French as representatives of the mixing populations (since date estimates are robust to improperly specified reference populations). The results are shown in Supplementary Figure S22. Both populations show a detectable curve, though the signal is much stronger in the Sandawe than in the Hadza. The implied dates are 89 generations (2500 years) ago for the Hadza and 66 generations (2000 years) ago for the Sandawe. These are qualitatively similar signals to those seen by Pagani et al. [65] in Ethiopian populations. There are two possible historical scenarios that could lead to these signals: either the Hadza and Sandawe both directly admixed with a western Eurasian population about 2,000 years ago, or they admixed with an east African population that was itself admixed with a western Eurasian population. The latter possibility would be consistent with known east African admixture into the Sandawe [16] .


Pagani et al. refers to the paper Ethiopian Genetic Diversity Reveals Linguistic Stratification and Complex Influences on the Ethiopian Gene Pool, which inferred that the origin of the West Eurasian-African hybrid population of the Horn of Africa date to ~3,000 years B.P. Let’s back up here for a moment. First, it looks like the population of the Indian subcontinent came into being in its current state in the Holocene. Second, it also seems likely that there has been major disruption to the genetic substratum of Europe since the end of the last Ice Age. The top-line result is that a substantial minority of European ancestry, with a southwest-northeast cline, may be of East Eurasian provenance. But I think just as important is the possibility that there has been successive replacement of West Eurasian groups over the course of European prehistory. The genetic signal of the latter exists, but is more difficult to tease apart because these branches of the phylogenetic tree have a more recent common ancestor and are not as differentiated. Third, in historical time we have seen a massive reorganization in the center of Eurasia, as Iranian populations have given away to Turkic ones. Fourth, there is a fair amount of circumstantial evidence for massive population replacement across Southeast Asia of Negrito groups. First by the ancestors of Austro-Asiatics, and then Austronesians. Finally, now there are these confirmations of massive genetic turnover on the fringes of East Africa, from Ethiopia, down to the south in Tanzania.

One issue which now comes to mind for me is the nature of the people of Punt, a mysterious land, perhaps in modern day Somalia, which traded with antique Egypt. From the depictions in Egyptian wall art they do not seem to exhibit a conventional Sub-Saharan African appearance, despite their likely African location. And ancient Egyptians clearly were familiar with people of Sub-Saharan African appearance, as Nubians appear early on in their wall wart. But these results from ROLLOFF and Pagani et al.’s inference as to time of admixture give us a possible explanation: the people of Punt were outriders of Southwest Asian expansion into East Africa, and they exhibited an unadmixed appearance because admixture had not yet been extensive. Ultimate we’re taking para-history here. The people of Ethiopia were on the fringes of Egypt and the Near East, but at far enough of a remove that we are not treated to any literary documents which attest to the process of ethnogenesis.

This does not even touch upon the Bantu expansion, which seems to have occurred after the first intrusion of West Eurasians into the East African landscape. It is notable that the Sandawe as an amalgam of Khoisan and West Eurasian, while their Bantu neighbors presumably have less of both of these elements. When the pyramids were rising in Egypt all of this had not occurred.

CATEGORIZED UNDER: Genetics, Genomics
  • Grey

    I’m thinking as agriculture spread to particular spots where it was most suited (at least after a bit of adaptation e.g. irrigation) and could produce the highest population densities e.g. Yellow river etc – those spots turned into massive population pumps.

  • Karl Zimmerman

    I recently took a look at East African populations in the Harappa World database, basically excising the SSA components and looking at the proportions of what was left. There was basically a three-stage breakdown.

    1. The Ari groups, Sandawe, and Massai all have Southwest Asian comprise 90%-100% of their West Eurasian component. This is interesting, considering modern Arabians have at most around 65% of this component, meaning the African segment is more “pure.” There’s only two ways to read this. One is Arabians expanded into Africa when they were still unadmixed with Near Eastern populations – which probably means pre-agriculture. The other is a large proportion of SW Asian is a stabilized admixture between Near Easterners and some now-extinct aboriginal population which existed on both sides of the Red Sea.

    2. The Somali groups showed mainly Southwest Asian ancestry (88%-80%), but also show non-trivial amounts of Caucasian and Mediterranean ancestry, in roughly equal proportions. This is unlike modern Arabians, who show far more Caucasian, and less Mediterranean, but North Africans from Libya onwards show a lot more Mediterranean than SW Asian, and Egyptians show an elevated element of this as well. This seems to be an echo of the expansion of Afro-Asiatic in general, which probably spread from Egypt during late prehistoric times.

    3. Ethiopian groups (both Semetic and non-Semetic) show even lower SW Asian ancestry (78%-68%), with the Caucasian portion elevated to almost Arabian norms, although the Mediterranean component is still elevated compared to Arabs. This element was obviously introduced quite late in history.

    I realize this is a very rough analysis. Hopefully people with more actual quantitative skills than I will be able to look at this.

  • priestcohen

    Karl Zimmerman:” The Somali groups showed mainly Southwest Asian ancestry (88%-80%), but also show non-trivial amounts of Caucasian and Mediterranean ancestry,”

    Could you explain what you mean by “Caucasian” and “Mediterranean?” For example, when you use Caucasian, are you referring to (A) people from the Caucasus, (B) West Eurasians (what people used to call Caucasoids), or (C), Europeans?


  • Razib Khan

    #2, surprised. this actually makes sense in like of pagani et al., they found ethiopian west euro was more like levantine. i was skeptical about the time.

  • Karl Zimmerman

    3 –

    I’m referring to the admixture components which are identified in the Harappa World spreadsheet. Roughly speaking, SW Asian peaks among modern-day Arabians, Caucasian peaks in the Caucuses (but is actually more dominant in the “fertile crescent” than Southwest Asian), and Mediterranean is dominant in both Southern Europe (especially Sardinia), as well as North Africa.

    Another similar spreadsheet which people often cite is Deinekes, which estimates ancestry components for a Europe-focused population. His own version of SW Asian and Caucasian are very similar, but his Mediterranean component is more strictly limited to Southern Europe, in part because he finds another group, called “Northwest African” – which actually seems to be a mixture of Mediterranean and SSA in Harappa World.

    Razib –

    Here’s some four-population breakdowns – I took Baloch out as well since most of the West Asian Balochi admixture was probably Bronze Age or more recent. In retrospect I should have taken out NE European as well. The first number is the total percentage of the genome made up by the following four components (Caucasian/NE Euro/Med/SW Asian).

    Qatari (76%) 22%/4%/3%/71%
    Saudi (88% 30%/1%/4%/65%
    Bedouin (86% ) 25%/2%/8%/65%
    Yemeni (69%) 39%/1%/8%/52%

    All Arabian populations show dominant SW Asian, secondary Caucasian, and minor Mediterranean

    North Africa:
    Tunisian Berber (79%) 0%/0%/66%/34%
    Fulani (29%) 0%/0%/64%/36%
    South Morocco (59%) 0%/0%/61%/39%
    North Morocco (77%) 2%/1%/60%/37%
    Algeria (72%): 4%/1%/54%/42%
    Mozabite (74%) 1%/1%/53%/45%
    Saharawi (73%) 0%/0%/52%/48%
    Moroccan (72%) 5%/1%/49%/45%
    Libya (77%) 14%/1%/35%/50%
    Egypt (79%) 36%/1%/21%/42%
    Egyptian (77%) 40%/0%/17%/43%

    There’s a big shift here from west to east. It would seem to that the lack of Caucasian outside of Libya/Egypt shows these were later, intrusive influences from the Near East, and the Neolithic population was mainly Mediterranean with some SW Asian.

    East African 1

    Ari Blacksmith (16%) 0%/0%/0%/100%
    Maasai (13%) 1%/1%/1%/97%
    Ari Cultivator (18%) 2%/1%/1%/96%
    Massai (13%) 3%/1%/1%/96%
    Sandawe (13%) 5%/1%/4%/90%

    SW Asian is dominant to such a large extent that I’d lean towards an indigenous origin. The Hazda aren’t shown because they show a very different pattern, with a lot of “Northeast European” admixture in their small West Eurasian ancestry.

    East African 2

    Somali (31%) 0%/0%/12%/88%
    Somali (37%) 8%/0%/6%/87%
    Ethiopian Somali (35%) 9%/0%/10%/81%
    Somali (36%) 10%/0%/9%/81%

    Minus the first Somali population (a small one from Reich) they are all highly similar.

    East African 3

    ethiopian (44%) 14%/0%/9%/78%
    ethiopian jew (46%) 14%/0%/9%/77%
    wolayta (33%) 14%/1%/9%/76%
    oromo (40%) 19%/0%/9%/72%
    afar (45%) 19%/0%/10%/71%
    amhara (47%) 21%/0%/9%/68%
    tygray (49%) 22%/0%/10%/68%

    In general, it seems like Semitic populations show more Caucasian (along with more West Eurasian overall). Everyone, both Cushites and Semites, shows around 10% Mediterranean ancestry (like Somalis), suggesting the Med ancestry is Cushitic substratum for the whole region.

  • Maju

    “From the depictions in Egyptian wall art they do not seem to exhibit a conventional Sub-Saharan African appearance”…

    Pick please a color version of that image, Razib, preferably one with the other “Puntians” (because the color may have been better preserved)? The preserved fragments of the color of the skin of the people of Punt is clearly dark brown. You can see that on the Queen of Punt herself (specially the legs) and on her husband and their assistant. Several examples with different illumination: example 1, example 2, example 3.

    Other images of Puntians (example 1, example 2, example 3) also show them brown to dark brown, very different shades as used by Egyptians to depict themselves, West Asians or Berbers.

    So why would you choose of all these images one that has no color? I’m hoping it’s an honest mistake but a simple image search exercise would have nullified your version (also the Queen of Punt has quite wide nostrils even for the modern average Ethiopian or Somalian, who actually have quite narrow noses, even for Caucasoid standards).

    Some stuff in this entry is clearly wrong. The rest highly speculative but the pretending that Puntians were unmixed Asians… is simply ridiculous.

  • Razib Khan

    just went by the wikipedia entry. can you chill with the conspiratorial aspersions? i’m getting tired of your attitude. your point seems fair enough on the substance.

  • Amun

    In Tishkoff et al. 2009 (one of the largest African population genetics studies) they didn’t seem to have detected a Eurasian or North African component in those groups:

    However, it does appear in many East African Afroasiatic groups (and to a lesser extent in some Nilo-Saharan groups admixed with Afroasiatics) but not in the Hadza or Sandawe. Quite peculiar.

  • Razib Khan

    #8, microsatellites are not the best for some things.

  • Amun

    @Karl Zimmerman,

    There’s only two ways to read this. One is Arabians expanded into Africa when they were still unadmixed with Near Eastern populations – which probably means pre-agriculture. The other is a large proportion of SW Asian is a stabilized admixture between Near Easterners and some now-extinct aboriginal population which existed on both sides of the Red Sea.

    It would be interesting to see how Soqotri samples would behave in these analyses, they may have much lower levels of ‘Northern’ Middle Eastern ancestry than most modern-day mainland Arabians. The natives of Soqotra seem phenotypically quite different from mainland Arabians as well and resemble South Indians more. This possibly suggests they may have a higher fraction of aboriginal Arabian ancestry, but that’s just speculation for now.

  • ohwilleke

    1. Pickrell’s conclusion that there was admixture with an already Eurasian admixed population is very plausible.

    2. Somali uniparental genetics show fairly high frequencies of Y-DNA and mtDNA haplogroups that have clearly back migrated from Eurasia. About 10% of Somali males are Y-DNA haplogroup T and another 5% have other Eurasian Y-DNA haplogroups. The balance is mostly E1b1b1a – found in almost all Afro-Asiatic linguistic populations and some populations geographically adjacent to them, with just 5% of Y-DNA being sub-Saharan African. Likewise, a large percentage of Somalis have mtDNA haplogroup M1 (also back migrated at some point from Eurasia and distributed in line with Afro-Asiatic populations); one study estimates that 46% of Somalis have Eurasian mtDNA lineages (with only about 20% being M1 that may have deep time depth and an Africa-wide Afro-Asiatic affinity).

    3. It is not at all clear whether the source of this Eurasian admixture arrived in Somalia by sea, or by land (presumably via the Nile).

    The population genetics of Somalia relative to Ethiopia relative to Sudan relative to Egypt makes the quite homogeneous and non-diverse genetic profile of Somalia relative to its neighbors seem less plausible than a scenario in which the predominant source is SW Asian mariners who don’t cross genetically diverse intervening territories en route and have a trip measured in days not years. I have real doubts about how “pure” a founding population of Eurasian contributors could remain via a land route trip that could easily end up being multigenerational.

    The fact that the Somalis speak a Cushitic language rather than an Ethio-Semitic language argues for a period of admixture that was separate from the arrival of proto-Ethio-Semitic languages in Ethiopia, and for a very different process of admixture. The Eurasian admixture in Ethio-Semites is male biased. The Eurasian admixture in Somalis is, if anything, female biased or gender balanced. A scenario of mutual bride exchange by SW Asian mariners and Somalis who engaged in trade relationships with each other seems more plausible given the gender balance and lack of language shift. This also fits with a recent paper that suggests that Yemini maritime trade was conducted by small extended family sized enterprises rather than a coherent powerful state (a la the East India Company or the Christopher Columbus expedition). A decentralized society is less likely to mount the kind of migration campaign that produces a population with enough impact to bring about language shift.

    The Cushitic language also argues against the hypothesis of Punt having an unadmixed Eurasian population. A Eurasian population with that much coherence would be unlikely to have adopted the local Cushitic language (a branch of Afro-Asiatic that is clearly African in origin whatever one’s thoughts about the ultimate place of origin of Afro-Asiatic languages).

    4. I suspect that the estimate of “the origin of the West Eurasian-African hybrid population of the Horn of Africa date to ~3,000 years B.P.” is probably a bit too recent due to miscalibration issues with mutation rate dates. An estimate of 5,000-6,000 years B.P. would have much to recommend it.

    For example, the older date is quite a bit more plausible as the date when Somali begins to participate in Indian Ocean maritime trade probably via Yemen and on to Mesopotamia and South Asia given, for example, the trade goods that end up in South Asia and Mesopotamia at the time and the existence of Harappan-Sumerian maritime trade in that time frame. It is hard to see how Somali couldn’t start to become somewhat admixed with SW Asians once they were conducting active martime trade with SW Asia.

    The more “pure” SW Asian population component of Eurasian ancestry that Karl Zimmerman identifies makes quite a bit of sense in a time period before the bidirectional admixture that likely accompanied the adoption of the Semitic Akkadian language by the Mesopotamian populations that previously spoke Sumerian. Until then, Semitic peoples wouldn’t have had an opportunity to admix with peoples from the West Asian highlands.

    A more “pure” SW Asian population component of Eurasian ancestry would also make sense if this component arrived predominantly by sea from Yemen, a likely ancient trade route. No part of SW Asia is more distant from potential sources of admixture, and there is some historical indication that Indian Ocean trade with Sumeria may have been mediated with a cargo handoff to new boats somewhere in the vicinity of Qatar.

    3000 B.P. is right around the historical moment that the Bantu expansion reaches the East African coast and that the Austronesians from Borneo (at least linguistically) who settle Madagascar together with East African Bantus arrive in Madagascar. A population whose ethnogenesis is 5000-6000 BP has a much greater likelihood of having very little sub-Saharan African contributions than one with an ethnogenesis 3000 BP right around the time that Bantus reach East Africa.

  • Karl Zimmerman

    10 –

    I was originally interested in South Arabian populations as well. Not so much the Soqotri, who at least on the surface to have been admixed with horners, but the mainland populations like the Mehri. But I realized that since modern day Arab-speakers, as well as Ethiopian Semitics, show elevated Caucasus ancestry. South Arabian languages are more closely related to Ethiopian Semitic languages than Arabic. Thus, the chances are very high that South Arabs would also have high Caucasus. The first two phases of admixture I outline above are almost certainly “pre-Semetic” in nature.

    11 –

    Your point about the Akkadians set off a light bulb in my head. I was making the assumption previously that Neolithic Arabians were Afro-Asiatic, if not Semetic. However, the three populations which show the “pure” SW Asian component speak Omotic (Afro-Asiatic, but barely), Nilo-Saharan, and a language isolate. It may be the following scenario was correct.

    1. Preliterate neolithic Arabians were a non-core Afro-Asiatic population. Perhaps they spoke Omotic, perhaps Nilo-Saharan. Maybe something related to Sumerian or even Sandawe. Maybe more than one is true – to the best of my knowledge, no one has tried to link Sumerian to any African languages. They expanded from Yemen and into West Africa, ultimately even having genetic influence over Eastern Sudan, if not even further.

    2. The secondary expansion – both of Cushites into East Africa, and Berbers into North Africa, clearly has involved a population with some “Mediterranean” elements. Given its present in both Cushites and Berbers, I believed that Egypt made the most sense. But I wasn’t considering that during the early Neolithic in Africa, there was a “Wet Sahara.” The drying happened around 3,000 BC, which is roughly 2,000 years after the development of Sahelian agriculture. Modern Afro-Asiatic populations ring the Sahara – Berbers on the north, Chadaic-speakers on the South, and Egyptians and Cushites to the East. It may be their homeland was somewhere in the Central Sahara, and the drying both forced them to the margins, as well as allowed them to displace earlier populations, including conquering the Arabian-admixed population around the Horn of Africa. Perhaps the same movements caused a huge displacement of the Nilo-Saharan populations as well.

    3. Finally comes the Semitic expansion. From the Fertile Crescent into Arabia, then later first into the Ethiopian highlands, and later through Arabic into Egypt and beyond – taking with it a “Caucasian” signal over the “Southwest Asian” substrate.

    Conjectural yes, but it’s another model which seems to fit the genetic data.

  • Joshua Gatera

    The ROLLOFF results from both Pickrell et al. (The genetic prehistory of Southern Africa) and Pagani et al. (Ethiopian Genetic Diversity Reveals Linguistic Stratification and Complex Influences on the Ethiopian Gene Pool) are both accurate in respect to dating legitimate admixture events that took place in the general vicinity of NE Africa and SE Africa in the distant past. The conclusions of the respective authors and the misinterpretations by the likes of the author of this blog and Dienekes are unfortunately inaccurately interpreting the implications of these aforementioned results.

    ROLLOFF works best in scenarios where the “hybrid” population is the result of a single admixture event between two separate ancestral populations. African-American for example are a great example; because they are largely the product of racial admixture between African slaves (predominantly) and European-Americans (minor-moderate) during the past 500 years, ROLLOFF is able to accurately date the aforementioned admixture events to a particular time frame in between 1526 and the present. On the other hand that’s clearly not the case in the Horn of Africa, made evident by the fact that Somalis for instances are notably less Eurasian than northern highlanders, and in turn Semitic speakers are slightly more Eurasian than their Cushitic speaking (Agaw) counterparts. In situations likes these, ROLLOFF tends to pick up the most recent cases of foreign gene-flow into any particular population. As was the case in Pagani et al. This particular study dated the Eurasian-African admixture event to 3,000 BP, but whats clearly obvious to anybody well informed on the biological dynamics of the Horn of Africa is that Pagani had simply identified the latest case of Eurasian gene-flow into the Horn of Africa, that being the introduction of Semitic into the northern highlands. According to Kitchen et al. (using the Bayesian computational phylogenetic techniques) “contemporary Ethiosemitic languages of Africa reflect a single introduction of early Ethiosemitic from southern Arabia approximately 2800 years ago”; the results from Pagani et al. seems to support the aforementioned, but instead of realizing the importance of the study we have certain individuals hijacking these results in the effort of furthering their agendas. Tigray-Tigrinya (on both sides of the Eritrean/Ethiopian border) are about 3% more Eurasian than their Agaw speaking counterparts, i.e. the Beta Israel and the misclassified “Afar” samples from Pagani et al. who are in fact Xamtanga speakers from the Amhara region; Amharas are more heterogeneous, they tend to fluctuate between the former and latter categories, in due consideration of the fact that the Amhara linguistic-ethnic expansion involved the assimilation of indigenous Agaw speakers in the region.

    The two admixture events noted by both Pickrell et al. and Pagani et al. are in no way related to one another. The perceived “Western Eurasian” gene-flow into the Hadze and Sandawe and the respective admixture date is in fact accurately dating an admixture event between the Khoisan-like hunter-gatherer ancestors of the aforementioned peoples and the NE African-like pastoral ancestors of the same two groups. Various linguists, geneticists and other anthropologists have provided ample amounts of evidence that in turn supports the leading theory that the southward expansion of NE Africans (and possibly a related group, intermediate between the aforementioned and Nilotes) into SE Africa during that 2,000-3,000 BP time frame radically altered the genetic composition of the region. In which NE Africans both subsequently 1) assimilated the indigenous Khoisan-like inhabitants (explaining groups like the Hadze and Sandawe) and 2) forced out a segment of that population, i.e. the now pastoral Khoe, towards the direction of southern Africa. If Pickrell et al. had included samples from the Horn of Africa region, they would have also realized the aforesaid phenomenon.

    Karl Zimmerman is absolutely correct, the “Western Eurasian” affinity characteristic of populations native to the Horn of Africa and the vicinity clearly signifies an indigenous Western Eurasian affinity (i.e. the Ari), subsequently layered by various waves of legitimate Western Eurasian gene-flow.

    So in reality, you have a situation like the following…

    1) A group of Africans that are differentiated from other Africans by their significant Western Eurasian inclination; southern “Western Eurasians”, exemplified by the “Southern” and/or “SW Asian” clusters seem to possess alleles from this ancestral population to a significant degree.

    2) “Erythrean” or the first wave of Eurasian integration; this Eurasian component differentiates Omotics from other Afro-Asiatic speakers, including Cushites. This component in my opinion was likely incorporated (into a Northeast African population) in what is now Egypt and was subsequently introduced back into the Horn with the “back-migration” of Cushitic speakers from Egypt; parental lineages like M-728, T, M1, U6, and some other maternal lineages are likely associated with this event. The back-migration of Cushitic speakers implies two things, 1) Cushitic speakers become increasingly differentiated from northern Erythrean speakers (i.e. Egyptian, Berber-Chadic, Semitic) either through more NE African admixture in the former or more Eurasian admixture in the latter, and 2) these incoming Cushitic speakers subsequently force out Omotic speakers from the likely origin of Afroasiatic, i.e. Eastern Sudan-Ethiopian highlands, into southern Ethiopia where they there assimilate significant hunter-gatherer ancestry. Reference the Ehret section of the subgrouping tab for aforementioned linguistic references.

    3) Neolithic gene-flow from Arabia which differentiates most Eastern-Southern Cushitic speakers from speakers of the northern branches of Beja and Agaw in what is now the Eastern Sudan and northern highlands. Ethiopian Jews for example are about 10-15% more Eurasian than Somalis. J1 seems to be the primary parental lineage associated with this event, in addition are a select few of maternal markers. I’d like to emphasis the fact that the “Afar” samples from Pagani et al. are misclassified Agaw speakers, similar to the Beta Israel in nature; in addition, the Oromo samples aren’t “pure” Oromos from the region of Boran for example, but are encultured highlander samples from northernmost Oromia.

    4) The small, but non-trivial (3-4%), Eurasian component that differentiates Semitic speakers from their northern Cushitic neighbors.

    I’d like to end my post by stating that while Tishkoff et al. may have had it’s limitations, their results were the most likely and definitely realistic; northern highlanders were about 25-35% Eurasian, in comparison to groups similar to the Somali who were on average ~15% Eurasian.

  • Razib Khan

    why do we need more research? we have joshua gaterra to pronounce the truth 😉

    though more seriously, i’m skeptical that ROLLOFF would show such a strong signal from such a small ‘last-wave’ migration. though i haven’t done the simulations to check. perhaps pickrell et al. could comment.

  • Maju

    Actually I just came here from the RSS feed because what Joshua Gatera says appears to make much sense and I was hoping he had a link to a resource of his own, where I could find more interensting information/meditations in that line. It seems it’s not the case but his analysis seems on first sight at least quite valid to me.

    At least it is the kind of interpretation that makes sense with all I think I know about population genetics in the overall region, unlike some interpretations I have seen these last days, which look far fetched and pretty much forced.

  • Lank

    I think Joshua has a lot of good points. When the Ethiosemitic group is about 3000 years old, I seriously doubt the Hadza and Sandawe could descend from a Eurasian population that lived just 2000-3000 years ago. I’m guessing what we’re seeing is actually a signal of the mixing between distinct groups that the ancestors of East African hunter-gatherers took part in during the past 3000 years.

    The West Eurasian admixture signal is much stronger in the Sandawe than the Hadza, and the reason for this is evident. Sandawe have clear traces of South Cushitic ancestry (note that the South Cushitic group is probably around 3000 years old), whereas Hadza mostly do not. It seems unfathomable that these groups could descend from a hypothetical population that was “purely” Eurasian 2000-3000 years ago, and had remained so while living in Africa for thousands of years.

    In Ethiopia, what is obvious is that each migration of Afroasiatics (Omotics, Cushites, Semites – in that order) contributed to an increase in West Eurasian admixture. If there was a population on the coast of the Red Sea that was more Eurasian than its modern inhabitants, these Afroasiatics (or, later, the early Ethiosemites) would probably be it. However, the admixture pattern among modern populations does not support such a scenario. Beja, the oldest group of Cushites (much, much older than 3000 years), live in this area today. What you’ll notice if you compare them to Ethiopian and Somali samples is that they are very similar, although Somalis are slightly more African. Interestingly, Ethiopians have a noticeable ‘Omotic’ element that Somalis seem to lack. This supports earlier linguistic models (see, e.g., Ehret’s work) that Cushites spread into the Horn from the Red Sea hills of Sudan, absorbing the earlier inhabitants of the Ethiopian highlands (Omotics). Somalis may have bypassed this area by following a lowland route through the northern, eastern and southern Horn of Africa.

    So I don’t see how a much more Eurasian group of Afroasiatics could have lived in Punt (which recent research pinpoints to Eritrea and nearby areas, in accordance with Punt-like artefacts found earlier in this region). There may have been groups from the other side of the Red Sea who migrated across, but I don’t see how the core population could have been fully ‘West Eurasian’, as they should have been proto-Cushitic, or at least related, and the evidence indicates that the West Eurasian ancestry of Cushites is certainly older than 3000 years. Of course, ROLLOFF tells a different story, but that software produced a similar admixture date for Ethiopian Semites, Cushites and even Omotics. So I think some skepticism is warranted.

  • Karl Zimmerman

    13 –

    Consider me schooled. It looks like my first scenario was more right than the second I posited.

    I do have a few thoughts though.

    1. While I agree that a migration of Cushitic speakers from Egypt into Ethiopia makes the most sense, there is the issue of Nilo-Saharan being in the way. The most divergent languages are along the Ethiopia-Sudan border, and it appears that at least some of the lower portions of the Nile have been always outside of the Afro-Asiatic sphere. Thus traveling up the Blue Nile into Ethiopia seems unlikely, meaning a migration along the arid coastal route. Not impossible of course, but migrating through the Afar Depression seems a lonely route to take.

    2. I’m unsure if the third and fourth point need to be viewed as separate migrations. It may be that the entire area was influenced by a Semetic influx, but the population in the areas which didn’t see language shifts had slightly less Semites, so instead of the conquered taking on a new language the Semite “lords” acclimated. In addition, mixing between all ethnic groups (and entire communities shifting linguistic affiliations) in the Ethiopian highlands has been fairly common in recently documented history, which could have helped smooth out initially more dramatic differences between populations. The difference between the Tigray and the others looks to be due to “a bit more” rather than “a bit more and something different” after all.

  • Razib Khan

    the problem here is that

    1) we have people with deep thick knowledge of many facts


    2) we have people with deep knowledge of very powerful analytic methods

    i don’t count myself in either camp, but i really wish that some people in #1 would chill on the arrogance. yes, we don’t know shit. so do you just want us to not talk about this, period?

  • Onur

    1) we have people with deep thick knowledge of many facts


    2) we have people with deep knowledge of very powerful analytic methods

    Statements of the latter are more valuable for me than statements of the former. I prefer one single genetic meaurement to lots of informed speculations. The hardest thing is the interpretation of the measurements (=conversion of the digital measurement values to the analog real world); it is where most of the dispute occurs.

  • Razib Khan

    #19, genetic summaries can be deceptive and misinform. e.g., PCA. they are easier to understand. but that doesn’t mean they are more right.

  • Maju

    #18 – But, Razib, you’re appealing to authority (the weakest of all arguments after “I think so”), the relevant paper is not even peer-reviewed yet (what may not mean much but still…) and the paper does not reach to the extreme conclusions that Dienekes or you want to extract: it just suggests that they might have detected (with relatively new and only so strongly tested methodology and not in all tests) that the Sandawe and Hadza show some sort of affinity signal with French and Basques (not Arabians).

    So you’re blowing these two blue ink marks out of proportion. Totally so. I beg caution.

  • Razib Khan

    #21, they do have affinities with arabians. specifically, the yemenite jews in behar et al. i ran the analysis separately with ADMIXTURE last year. joe pickrell referred to arabians in his haldane’s sieve post i recall. and don’t confuse speculative blog posts with conclusions. i don’t care much about the final outcome of this debate, so contrary to your implication i don’t have any ax to grind.

  • Karl Zimmerman

    Hrrm…from the length of Mr. Gatera’s response, and Razib’s reply, I presumed he was an academic in the field. But I am thinking now Razib was being sarcastic, and I see there’s not much of anything besides posts on this blog when I google him.

  • Razib Khan

    #23, i think he has interesting things to say. i wish he’d be less ex cathedra. if genes are always wrong when they conflict with his priors than there’s no point in following genetics, is there? i don’t have strong final opinions on these sorts of topics, so i feel free to speculate. sometimes speculation is worthless or misinformed. OTOH, it isn’t as if those who are historically informed aren’t missing something either. for example, i work with microsats and SNPs every day, and they are useful for different things. but for this sort of phylogenetics, go with SNPs if you have thick marker sets (and they do). they perform better (microsats are better for forensic differentiation between populations, but i don’t always trust the phylogenies they generate because they mutate so fast).

  • ohwilleke

    #18 re: the Foxes v. Hedgehogs comparison.

    The big flaw I see in a lot of archaeogenetic work is the failure to appreciate the importance of corroboration of an interpetations from multiple independent lines of evidence (e.g. archaeological pots and bones as well as genetics). If the two lines of evidence can’t be harmonized, then something in the model is broken.

  • Ed

    Fourth, there is a fair amount of circumstantial evidence for massive population replacement across Southeast Asia of Negrito groups. First by the ancestors of Austro-Asiatics, and then Austronesians. Finally, now there are these confirmations of massive genetic turnover on the fringes of East Africa, from Ethiopia, down to the south in Tanzania.

    Can I get quick links to this evidence? It’s not that I don’t believe you (It seems the most likely explanation to me actually). There’s certain people on a different site that I’ve been discussing this with that have trouble believing it.

  • Joshua Gatera

    lol @ Karl Zimmerman, I will take that as a compliment. No I’m not an expert in the field, but like other people in the blogsphere, like yourself, I do my best to formulate reasonable explanations in respect to studies like Pickrell et al. and Pagani et al. by using various lines of evidence (genetic, linguistic, archeological, etc) in support of my positions on a variety of different topics.

    My apologies Razib! I certainly don’t see myself as an “authority”, I do humbly consider myself knowledgeable on the subject, but I’m justifiably quick to criticize individuals who intentionally hijack studies like these to further their agendas. I have another blogger in mind who unfortunately by name dropping your name in a recent post on this particular study made you guilty by association.

    @ Maju, I don’t have any particular source or study which outlines the various points made in my aforementioned post, but my position on the subject is supported by an ample of evidence from a variety of different anthropological fields , from genetics to linguistics to archeology.

  • Maju

    #23 It is not so much the genes but what we (some people) deduce from the genes using statistical inference methods, whose results can only be taken as solid once they have been replicated from different angles: one statistcal “finding” can well be a fluke or even a methodological error, specially if the method is relatively novel and not too well tested as some of the methods that Pickrell uses, a dozen coincident statistical findings using different well-tested methods are “a fact”.

    Just because someone “find” something doesn’t make it true, specially if it seems inconsistent with previous evidence (exceptional claims require exceptional evidence). It is important to address each problem critically from different angles and to test the “findings” by means of replication. I think that in Pickrell’s own paper the results are contradictory in the matter at hand, so nothing conclusive, certainly not “genes” being right nor wrong.

    #24 “i don’t have any ax to grind”

    Glad to read that and my apologies if I understood otherwise.

    Notice that I do think that there is a West Eurasian inflow in Tropical Africa via the Red Sea and the Nile. There’s clear evidence in this regard but the degree of admixture rapidly decreases to near zero as we pass the Maasai or the Chadic peoples, so I feel that the Hadza/Sandawe “find” is quite suspect because these are very ancient rather isolate peoples.

    Pickrell would have been of more use if he used other East Africans like Maasais, Ethiopians, Luhya, etc. for contrast instead of just the Dinka. Maybe it is that the Dinka are not the right comparison point?

  • Onur

    the degree of admixture rapidly decreases to near zero as we pass the Maasai or the Chadic peoples

    What is your evidence for the claim that the degree of West Eurasian admixture decreases to near zero beyond the Maasai and Chadic peoples? ADMIXTURE and STRUCTURE analyses do not constitute evidence in this regard, as they are not good at detecting ancient admixtures.

  • Maju

    #29 Excuse my “ignorance”, Onur but Admixture and the like are a methodology I feel much more confident with, after all these years than some newer methodologies that are probably in need of refinement and that often produce very strange results, sugesting that they still need a lot of work in the basics.

    As for ancient admixture, we still have haploid DNA. In fact I would have not mentioned the Chadic peoples would not be for this one. In any case here the claim being made is of RECENT admixture, as recent as the Pyramids… go figure!

  • Karl Zimmerman

    Looking at the chart that Poster #8 linked too, West Eurasian admixture seems to exist in…

    West/Central Africa: Admixture seems to be limited to Dogon, Baggara, Fulani, Kanembu, and possibly the Hausa. In all cases (except the Dogon, which is just screwy) the reasons for admixture can be easily understandable. The Baggra are Afro Arabs, the Fulani mixed with Berpers, and the Kanembu and Hausa were part of “Sudanic” Islamic states which engaged in trade with north Africa.

    East Africa: Every single population shows a West Eurasian component, minus the Hazda and Sawande, as previously noted. Niger-Congo show the least, since these populations are either Bantu, or (like the Luo) heavily influenced by Bantu. Nilo-Saharans show more, and Afro-Asiatics show even more yet. And this is an underestimate, as the Sandawe “brown” component is part West Eurasian as well. What I would conclude here is both the elevated AA levels and the depressed NC levels are due to intrusive elements. The population substrate prior was “Nilo Saharan” in genes if not language throughout East Africa, and contained a small admixture of a West-Eurasian-like component.

    The “red” component is also interesting to me, as it’s mainly found in Nilo-Saharan populations, as well as Chadic-speaking ones. I can’t help but wonder if this is another compound of West Eurasian with something else (probably East African), as it would explain how the Chadic populations ended up Afro-Asiatic speaking without any immediately detectable West Eurasian admixture.

  • Onur

    #29 Excuse my “ignorance”, Onur but Admixture and the like are a methodology I feel much more confident with, after all these years than some newer methodologies that are probably in need of refinement and that often produce very strange results, sugesting that they still need a lot of work in the basics.

    ADMIXTURE and the like are insufficient when it comes to detecting admixtures from ancient periods. We are in an age in which new methodologies are being introduced every year, but you are still advocating the outdated admixture estimation methodologies of ADMIXTURE and the like. The newer methodologies are the results of years of refinement on the previous methodologies; they did not come out of nowhere. They are certainly much better at detecting ancient admixtures and, as a result, at measuring amounts of admixtures than ADMIXTURE and the like.

    As for ancient admixture, we still have haploid DNA. In fact I would have not mentioned the Chadic peoples would not be for this one. In any case here the claim being made is of RECENT admixture, as recent as the Pyramids… go figure!

    Dating of admixtures is still problematic. But I think in the near future much more accurate and problem-free admixture dating metodologies will be developed.

  • Joshua Gatera

    @ Karl Zimmerman

    The link # 8 posted is from Tiskoff et al. 2009…

    The “Western Eurasian” admixture detected in the 9 Dogon samples from the aforementioned study is a likely error, on the part of the researchers; since Tishkoff, we’ve obtained other Dogon samples, and as expected they’re genetically similar to other closely related West African groups like the Mandenka and lack the Eurasian component mistakenly detected by Tishkoff. The Hausa too, don’t possess any legitimate Eurasian admixture, they’re largely West African, in addition to a minor Nilo-Saharan component. The Baggara and Kanembou both, according to the aforementioned study, are admixed populations (but they tend to lack blatant Eurasian admixture to any significant degree); the aforementioned two groups possess West African, Nilotic, and Cushitic admixture.

    The genetic profile of the Henn et al. Fulanis are representative of Eastern Fulanis, in particular the small nomadic Wodaabe; and so while Western Eurasian via North African admixture has been therefore confirmed in this particular Fulani subgroup, we (as in the scientific community) haven’t had the opportunity to analyze the genetic affinities of the Western core Fulani groups (the Fulani only relatively recently expanded eastwards from their place of origin in the Senegambian region). If these Western Fulanis also possess similar amounts of North African admixture, that would suggest that the Fulani assimilated the Zenaga at an early stage of their ethnogenesis, but if they don’t, the most likely scenario would be Tuareg admixture. Regardless, the Western Fulani, in particular the Wodaabe, are an exception in West/Central Africa, minus of course the Maurs and Tuareg.

    @ Razib! Your next project could be retrieving a Western Fulani sample from the likes of Guinea or Senegal.

    In respect to East Africa, the Western Eurasian component peaked in the Beja (i.e. Beni-Amer and Hadandawa) from SE Egypt, Eastern Sudan, and Eritrea, who are genetically similar to the Tigray and Amhara of Ethiopia, at a frequency of 32-33%. It subsequently drops to a frequency of 27% among the Beta-Israel, and then the Rendille (the Rendille are very closely related to the Somali and their language is part of the Macro-Somali linguistic phylum) at a frequency of about 10%, the Oromo (Borana, Gabre, and Watta) at about 5%, the Konso and Burji (from southern Ethiopia) are also about 5% or less Western Eurasian; but from there the Western Eurasian component plummets to the likelihood of noise. Nilo-Saharans, Bantu speakers, and admixed (with Afroasiatic) SE Africans like the Maasai and Sandawe lack the Western European at any significant levels; the Sandawe component is partially NE African, not “Western Eurasian”.

    I disagree, SE Africa was likely primarily inhabited by Khoisan-like hunter-gatherer populations who were subsequently assimilated by waves of Nilo-Saharans and Afroasiatics from Southern Sudan and the Horn of Africa. These populations were then themselves assimilated by Bantu-speaking Central Africans by way of the Congo. Among SE African Bantus, there seems to be some genetic division between Great Lakes Bantus like the Luhya and Luo who are about 1/4 Nilo-Saharan and Savanna Bantus who are significantly Afroasiatic (the Kikuyu for example are 1/3 Afroasiatic). The aforementioned correlates with geography, since the Great Lakes are immediately south of South Sudan and the savannas are just south of southern Ethiopia and Somalia.

    The “red” component is Nilotic, the “burgundy” component is Chadic, and the “light brown?” component is Central-Sudanic; together they collectively signify gene-flow from South Sudan and are associated with the Nilo-Saharan phylum. It’s not Western Eurasian admixed at all, regardless of being closer to Western Eurasia than they’re West African or hunter-gatherer counterparts.

  • Joshua Gatera


    “Regardless, the Western Fulani, in particular the Wodaabe, are an exception in West/Central Africa, minus of course the Moors and Tuareg.”

    I meant to say Eastern Fulani.

    @ Maju

    “Other images of Puntians (example 1, example 2, example 3) also show them brown to dark brown, very different shades as used by Egyptians to depict themselves, West Asians or Berbers.”

    Interestingly enough, while the Ancient Egyptian artistic depictions of themselves, in terms of phenotypic expressions, varied, to an extent, throughout the dynastic period, the standard mold closely…

    …resembles Egyptian depictions of the Ancient Puntites (this has been previously noted by various archeological experts in the field), who according to modern research would have been from the area of what is now Eastern Sudan and Eritrea. While I wouldn’t make any assumptions based on eyeballing ancient art, the aforementioned supports the increasingly obvious likelihood that the Ancient Egyptians, at least in appearance (and in turn biologically), where more Afroasiatic East African than anything else. People tend to disregard the fact that the majority of the Ancient Egyptian population, throughout the dynastic period, were resident in between what is now Aswan (Nubia) and Luxor in southernmost Egypt; with a minority in Lower Egypt proper, i.e. the area just south of Cairo in the Fayuam region, while the Delta was only sparsely populated, possibly by peoples outside the Egyptian cultural continua. Today, Delta Egyptians make up approximately 65% of the population, while the historically dominant Aswani-Luxorians make up only about 6% of the population. So even without foreign admixture, modern Egypt is definitely not representative of Ancient Egypt; modern Aswani-Luxorians are still more “African” appearance, yet they’re the minority in their country, while persumingly the majority of Egyptians descent largely from people who were on the fringes of Ancient Egypt, outside it’s northern core population center south of Cairo.

    Besides the Puntites, Egyptians also resembled certain riverine Nubians and Libyans. In respect to the first case, Nubians tended to vary significantly, while some riverine Nubians resembled the Egyptians and likewise the Puntites, others possessed features that were more “Nilotic” in nature; modern Nubians and Sudanese Arabs possess the same variety of features, a testimony to the fact that they’re a mixture between Nilo-Saharans and Afroasiatics. Secondly, Egyptians also resembled southern Libyans from the southern Sahara; who were these people? Tebu?

  • Maju

    Joshua: there must be some West Eurasian ancestry (surely quite older than the pyramids) in important parts of Tropical Africa.

    First of all, Ethiopians look generally very mixed although their admixture to my eyes looks very old – because a local homogeneous Etiopian component shows up at some point, however it is intermediate between Tropical Africans and West Eurasians by Fst.

    Then we have all that haploid genetics in East and Central Africa that must have arrived from Eurasia at some point in prehistory: mtDNA M and N/R, Y-DNA R1b and T…

    And finally the Maasai and similar populations show a reduced fraction of the Ethiopian kind of admixture very insistently in all studies at all K-levels. Also in your link when populations other than Africans are considered that happens as well.

    I don’t know what’s the issue with pretending that “races” are closed boxes and that clines are unreal: clines do exist and there is African blood in West Eurasia (Europe included) and West Eurasian blood in Africa South of the Sahara. It may be not much but it is.

    Re. the panel you just showed, I would have to see an original photo: it’s too easy to recolor things so they appear the way one wants and the Internet is full of cheaters and manipulators. In fact a quick search shows several versions of the same panel or cut versions of it with totally different shades for Egyptians particularly (but also for Asians and even Nubians). However we do have a host of actual trustable images of Ancient Egyptians made by Ancient Egyptians. They have diversity of shades, possibly depending on the region or even artist’s preferences.

    Whatever the case the color of the Puntites (better than “Puntians” I guess) is clearly black/brown to me, not “Arab white”, whatever that is (there are very light and very dark Arabs in fact).

  • Joshua Gatera

    @ Maju

    I agree! If you go back to my post, no where did I state that Eurasian admixture was somehow absent in NE Africa. Quite the contrary, in my opinion, besides it limitations, Tishkoff et al. 2009 provided the most reasonable and realistic results, primarily due to the overwhelming emphasis of African genetic diversity. As previously stated in my response to Karl Zimmerman, Tishkoff et al. identified moderately sized Eurasian admixture in parts of NE Africa, in particular the Eastern Sudan and Ethiopian highlands, i.e. Eritrea and Ethiopia. The Egyptian-Eastern Sudanese-Eritrean Beja for example were approximently 1/3 Western Eurasian, while the Beta Israel were closer to about 1/4 Western Eurasian. But in respect to more inland, lowland Cushitic speaking groups like the Oromo and Rendille (a group closely related to the Somali) that Western Eurasian fraction dropped to approximently 1/8 in the Rendille and even less in the Oromo, Burji, and Konso who are all inland Afroasiatic speakers. These aforementioned numbers are pretty significant; the Afroasiatic cluster in Tishkoff et al. didn’t overestimate African admixture in Arabia or elsewhere, indicating that it’s not a “mixed” cluster as is the case with the “Southern” or “Red Sea” component from Dodecad for example. The Yemeni Jews, Bedouins, and Palestinians were only about ~10% East African according to the global run, the Mozabite were about 30% African; the aforementioed figures could be in fact underestimates, I’ve seen slightly higher numbers in the past.

    Your right about the Eurasian parental lineages as well, no disagreement here; in my previous post I incorporate those various lineages in what I believe, based on the evidence that I’ve seen, happened in NE Africa, i.e. 3 seperate episodes of Eurasian gene-flow into NE Africa. The first one was likely associated with lineages like M1, U6, (in addition to other notable mtdna lineages) and ydna T, in addition to African E-M78, by way of Egypt, the second one seems to be exclusively derived from the Arabian Neolithic (ydna J1 is key candinate, alongside some other important mtdna lineages), and lastly the introduction of Semitic, while less successful also played a role.

    R1 seems to have been introduced after the split within the Erythrean (ex. Omotic) sub-moiety of the Afroasiatic phylum; R1 would’ve been incorporated into a Northern Erythrean group (Berber-Chadic, Egyptian, Semitic0, in contrast to a southern (i.e. Cushitic). But while modern Chadic speakers possess high frequencies of the R1 marker, a relic of their Northern Erythrean origin, they possess little to no autosomal Eurasian admixture due to extensive mixture with Nilo-Saharan and Niger-Kordofanian speakers in the region. The aforementioned would have took place immediately after the split between Berber and Chadic as the latter expanded into the territory of indigenous Nilo-Saharan speakers, starting from what is now southern Libya.

    The Maasai for example are about ~50% NE Afican; but they descend from southern Ethiopian groups like the Oromo and not people like the Tigray for example. Therefore, the fact that they’re only about ~50% “Oromo-like” or even “Somali-like” wouldn’t translate to any notable Eurasian component among them, i.e. very low legitimate Eurasian signals for groups like the Maasai and Samburu.

    Again, we honestly don’t really even disagree with one another. imho Low-(Dominant-Moderate) Eurasian admixture is evident among Berbers (2/3rds – 1/half), Delta Egyptians (2/3rds), Riverine Nubians and Sudanese Arabs (1/3rd or less), Ethiopian Highlanders (1/3rd or less), and Cushites (on average at about 1/8th). No to mentioned the exceptional Wodaabe and other Eastern Fulanis who among the former in particular are on average 1/4 Western Eurasian (although they do vary, some of the Cameroonian Fulanis for example lacked NW African admixture). That’s pretty significant.

    Lets leave Ancient Egypt to the egyptologists. lol


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar