Not out of Sheba

By Razib Khan | June 22, 2012 3:56 pm

Liya Kebede, Credit

There is a new paper, Ethiopian Genetic Diversity Reveals Linguistic Stratification and Complex Influences on the Ethiopian Gene Pool, which is being sensationalized in the media. For example, the BBC headline: ‘DNA clues to Queen of Sheba tale’. I assumed that this was just the media, but to my surprise the authors themselves mention the ‘Sheba tale’ in their discussion for various reasons. This is unfortunate. Though it is true Ethiopians have a legend of descent from the queen of Sheba (and through her relationship to king Solomon the ancient Hebrews), if there is a scholarly consensus about the location of Sheba, it is probably in southwest Arabia (i.e., modern Yemen). But the reality is that it is probably just as likely that the story in the Hebrew Bible is an interleaved synthesis of legend and reality, and that disentangling the nuggets of truth so as to establish the location of the real Sheba is going to be impossible (it is just as likely that the real queen of Sheba, if she existed, was a Levantine notable who was given a more exotic provenance by the redactors of the Hebrew Bible).

As for the paper itself, it is of some interest. I’ve blogged and analyzed Ethiopian data myself, but the sample coverage here is awesome. Additionally, the authors attempted to ascertain time since admixture in relation to the Ethiopian population for their West Eurasia and African ancestral components, as well as sniffing around for signatures of selection in the genome. The highlights:

  • As first observed by Dienekes (to my knowledge) the Ancestral Sub-Saharan (ASS) component of Ethiopian ancestry is not in any way shape or form related to that modal in the Bantu or in West Africa. And, upon further exploration, it seems that it is separable from the Nilotic element as well, though this is less assured (one has to be careful when overloading a data set of a particular group of populations)
  • In Ethiopia it seems that Omotic ethnic groups are the modal reservoir for this component. This is of interest since Omotic are liminal members of the Afro-Asiatic language family
  • The major find here is that the non-African component of the ancestry of Ethiopians seems to have an affinity to Egyptians and Levantines, more than Yemenis
  • Additionally, there is some possible suggestive evidence for selection. Unsurprisingly Ethiopians carry a high proportion of the “European” variant of SLC24A5
  • Finally, the time since admixture is ~3,000 years BP (they used ROLLOFF)

In terms of selection, I am curious about what they found in the regions around the highland adaptation loci. One might predict that these regions should be enriched for indigenous African ancestry if the alleles are old. In contrast, if the alleles are newly arisen in the genetic background then there is no expectation that they should exhibit bias in their local genomic ancestry. The high frequency of SLC24A5 in a tropical population with West Eurasian ancestry is not surprising. South Indians have the derived variant on the order of ~50% frequency as well. The authors speculating about sexual selection seems like a deus ex machina. If sexual selection was strong for the derived variant and light skin then the allele should have become decoupled from the rest of the genome in terms of phylogeny (spreading to populations with lower levels of West Eurasian ancestry).

Two major criticisms. First, I am not clear that the comparison with non-African Ethiopian genomes was with the non-African genomes of the non-Sub-Saharan African populations. To get at what I’m saying, if you compare the West Eurasian ancestry of Ethiopians with various West Eurasian groups, then the proportion of West Eurasian ancestry in those groups is going to effect your Fst. Non-Jewish Yemenis have a high load of Sub-Saharan African ancestry. The relative closeness of the non-African component of the Ethiopians to Egyptians and Bedouins may simply be a function of the lower African ancestral load in these populations in comparison to the Yemenis. If the authors found greater genetic distance from Yemeni Jews I would be much more convinced, because the Jewish population in Yemen has a far lower proportion of African admixture than the non-Jews.

Second, like Dienekes I am not quite sure of ROLLOFF’s power in terms of generating a good peg for the time of admixture in this chronological window of time. The recent admixture events (e.g., North Africa, African Americans) are obviously right. But is it plausible that large numbers of West Eurasians were pushing their way into the highlands of Ethiopia as late as ~3,000 years ago? Perhaps. The depictions by Egyptians of the people of Punt seem to suggest they were of mostly West Eurasian ancestry. It could be that ~4,000 years ago the admixture had not been so thoroughgoing. There are two reasons I’m skeptical though. First, if there is one part of the world where we have some documentation of population movements ~3,000 years ago, it is the Near East. All we have to go on at this point is ROLLOFF. Second, like Dienekes I think we should be careful about relying on ROLLOFF alone. I have a hard time accepting ROLLOFF’s estimate for the admixture between West Eurasians and indigenous ancestral Indians ~3-4,000 years ago as well. Rather, I think that ROLLOFF is either biased toward underestimating the admixture time, or, picks up the last major pulses and misses the “peaks” of admixture. I would push both Ethiopian and Indian admixture events back several thousand years at least from what ROLLOFF is implying (or, perhaps more precisely, the inferences that some researchers make from ROLLOFF).

Frieda Pinto, Credit

Which brings me to an interesting point: there are strange correspondences between the demographic history of Ethiopia and South Asia. In both situations you have a population which seems to have arisen out of a balanced admixture between a distinctive indigenous population and a West Eurasian group which was intrusive. The ancient and medieval Western thinkers sometimes confused Ethiopia and India because of their marginal geographical position in relation to the Mediterranean world and the existence of dark-skinned people in both locales. The Greeks did differentiate though between the lighter skinned Indians of the north and the darker skinned ones of the south, with the latter resembling Ethiopians the most, except that their hair form was not curly (in reality, “north” would be the Punjab and Sindh, while the “south” would be Kerala and Tamil Nadu, because of the nature of Greek commerce and trade). Today some South Indians apparently get confused for being Ethiopian, and no doubt the reverse occurs, especially for women who straighten their hair somewhat.

That’s all I’ll say for now. The data is online, in convenient pedigree format. So I’ll be weighing in more in the near future….

Citation: Ethiopian Genetic Diversity Reveals Linguistic Stratification and Complex Influences on the Ethiopian Gene Pool, Pagani et al.

  • A N N A

    “Today some South Indians apparently get confused for being Ethiopian…”

    Hello to you, too. 😉

  • ohwilleke

    * A single admixture event model for West Eurasian admixture in Ethiopia is almost surely wrong.

    Uniparental markers like Y-DNA haplogroup T and mtDNA haplogroups M1 and U6 clearly show a pattern consistent with admixture much older than 3,000 years ago, and in the case of mtDNA M1 and U6, traces of this West Asian back migration to Africa appear throughout North African down the Nile and into East Africa.

    Ethiosemitic linguistic evidence favors an admixture date somewhere in the vicinity of 3,000 years ago, which has some support in a male dominated pattern from a prior uniparental marker study in Ethiopia. The widely held linguistic assumption is that almost all Ethiosemitic linguistic populations used to be Cushitic language speakers until the Ethiosemites arrived.

    Together, these data points make a strong case for one West Eurasian admixture event at least 8,000+ years ago (and by many estimates twice as long ago) and another more like 3,000 years ago, and those two waves may not have been the only ones. A three wave model could include an Epipaleolithic wave (i.e. just before farming and herding arrive) perhaps 12,000 to 16,000 years ago, a second early Neolithic wave (i.e. with the arrival of herding at least (farming comes later in Ethiopia than herding does) about 8,000 to 6,000 years ago, and a third Ethiosemitic wave, about 4,000 to 3,000 years ago.

    But, the dating model used only allows a date to be produced as the output of its calculations so it is inherently incapable of producing results that look like the most plausible models based on historical, linguistic and uniparental genetic evidence. And, a date produced by a model inherently incapable of giving the answer most likely to be correct should be distrusted.

    * Another puzzle for this data (or at least an interpretation of it that attributes European admixture in Ethiopia primarily to an Ethiosemitic wave contribution ca. 3,000 years ago) is that speakers of Ethiosemitic languages and speakers of Cushitic languages (which linguistic evidence makes absolute clear have a much more ancient presence in the region by thousands of years), seems to have almost the same amount of West Eurasian admixture – a result contrary to the Ethiopian autosomal genetic data that you blogged previously.

    This just does not fit any sensible model with a major demic migration giving rise to West Eurasian autosomal genetic components around 3,000 years ago. Any migration like that should have had much more impact on Ethiosemitic linguistic populations than it did on Cushitic linguistic populations that didn’t experience language shift at that time. Maybe this is a product of a flaw in Rolloff. Maybe there is some systemic reason lurking in the data collection reason for this to be the case. But, it just doesn’t make any sense at all.

    If you had a uniform Ethiosemitic and Cushitic level of admixture and a much oldest estimated admixture date, you could argue convincingly that there was lots of old West Eurasian admixture at or before the time that the proto-Cushitic language emerged, but that the Ethiosemitic language shift was elite dominated, in much the same way as the language shift that causes the Hungarians to speak Hungarian instead of an Indo-European language was elite dominated and left no real genetic traces of Uralic speaking populations in the general population genetics of the country.

    * If one must insist on matching this data to something described in ancient writings historical or legendary as the case may be, even if it is on the more ancient end of the margins of error, a more plausible link would be to the Bronze Age Levantine originating Semitic language speaking Hyskos rule of much of Eygpt (ca. 1650 BCE to 1550 BCE). Hyskos rulers might have permitted proto-Ethiosemitic migrants to pass through or split off from them up the Blue Nile and into Ethiopia.

    Unlike the Iron Age Queen of Sheba story, which appears mostly in stories that were intended to have a substantial religious component, the references to the Hyskos rule of Egypt are to sources intended to be historical (such as king lists) and widely credited by historians of who have expertise in ancient Egypt as true.

  • Lank

    Good criticism.

    I’m not convinced of the supposed similarity of the non-African component to Levantine people. The Yemeni samples (I presume from Behar et al.) they’re using have significant Bantu admixture. Looking at the pairwise genetic distance (Table S3), the non-African component is actually most similar to Egyptians, then Bedouins and Saudi. In other words, it is more similar to Arabians than proper Levantine people (excluding Egyptians).

    The similarity to Egyptians may be explained by common Afroasiatic ancestry (K=8 identifies a Northeast African component shared between Ethiopian Cushitic- and Semitic-speakers, and this incorporates the African ancestry in Egyptians). Cushitic is believed to have entered Ethiopia from the north, as it is more closely related to the rest of Afroasiatic than Omotic. The oldest branches of Cushitic (Beja, Agaw) are also the northernmost Cushitic branches. A spread of Cushitic from farther north also fits with certain uniparental markers (Y-DNA E-V32 seems to have spread from Egypt or nearby, perhaps mtDNA M1 and some N lineages as well)

    The Punt depictions look similar to the modern people to me, but this is obviously difficult to judge. I agree with ohwilleke, any major admixture event from 3 kya outside Ethiosemites and their closest neighbors is implausible.


    As I have stated on this blog in the past, a large part of the differences previously seen between Cushitic and Semitic samples mostly had to do with their geographic origin. Most of Behar’s Oromos are from far southern Ethiopia, and the linguistic ancestors of Oromos expanded from southern Ethiopia in the past 500 years. Most Oromos, such as the new samples from this study, are probably a mixture of highland Cushites and Oromos, with some Habesha ancestry farther north. Some of these new Oromo samples are fairly similar to Semitic speakers, and they aren’t even from that far north.

    The “Afar” samples in this study are apparently mislabeled Agaw Central Cushites (they speak Xamtan and are agriculturalists from a region inhabited by Agaws, according to Table S3). They seem to be indistinguishable from the new Amhara samples. This is not a big surprise as Central Cushites are the most likely ancestors of Semitic speakers, and many Central Cushitic Agaws are known to have switched languages to Semitic in the past couple of centuries. Tigrayans have an older Semitic-speaking history than many Amharas, and appear to deviate very slightly toward West Eurasia relative to Amharas and the mislabeled “Afar” (Table S1A) samples. This is in accordance with mtDNA.

    Also, the kingdom of Axum had a documented Sabean presence during some time periods. Axum was located in Tigray and the Eritrean highlands (today inhabited by Tigray-Tigrinya people), and the official language was Ge’ez, a North Ethiosemitic language, like Tigrinya. After Axum declined, the highland civilization moved south, and Agaws took over (Zagwe dynasty).

  • marcel

    But the reality is that it is probably just as likely that the story in the Hebrew Bible is an interleaved synthesis of legend and reality, and that disentangling the nuggets of truth so as to establish … is going to be impossible

    The 5th question: How is this story in the Hebrew Bible different from all others?

    Answer: It’s not.

  • Lank

    By the way, “Ethiopia” in ancient times referred to the areas south of Egypt (i.e. Sudan). The Axumites used the term after conquering the Kingdom of Kush in the 4th century.

  • John Emerson

    When I was in college I had Yemeni and Ethiopian friends, and the Queen of Sheba was an issue with them, though my own friends just thought it was an amusing thing to chat about, like Germans and French arguing about Charlemagne, etc.

    The Indian Ocean sea routes are remarkably easy. You just wait for the monsoon, go one way, wait for the monsoon to turn around, and go the other way. I don’t know if a one year cycle was practical, but two years was. You just had to cross before it turned, which apparently wasn’t that hard.

    There was extensive Indian Ocean sea trade in every period back to the earliest antiquity. It’s not as well known in the West as the Silk Road because it was only Western (Greek and Roman) during a fairly brief period — mostly it was Persians and Arabs (or early Arablike peoples). As far as I know it’s still an open question as to how much pre-Greek African sea trade with India there was, but it was not technically difficult.

    Ethiopia (whatever that meant) was part of the early Egyptian world for many centuries — the Ethiopians were regarded by the Greeks as holders of ancient (Egyptian) wisdom. There were Ethiopian Pharaohs ca. 700-800 BC, before the rise of Athens and classical Greek civilization.

    There was a Keralan at my previous job with the last name of Solomon, as I remember. Not only are Ethiopians and some Keralans Christians, but they have the same theology (Monophysite).

  • ohwilleke

    “the non-African component is actually most similar to Egyptians, then Bedouins and Saudi. In other words, it is more similar to Arabians than proper Levantine people (excluding Egyptians). ”

    Of course, there has been a lot of mass migration and demographic shift in the entire region over the last three thousand years. I’m not convinced that the population demographic mix of any place in the Levant or Arabia is all that similar to the mix ca. 3,000-4,000 years ago in the same place. Events like the expansion and collapse of the Hittite empire and Byzantine empires, Mediterranean trade, Aegean based colonial efforts (e.g. the Philistines ca. 1200 BCE in the Southern Levant), Egyptian expansion and retreat, Roman era migration, Islamic expansion and intra-empire trade, the Crusdades, Meccan pilgrams who didn’t make it back home, and more have all tweaked the population genetic mix of subregions within the Near East in that time frame.

    Arabia is also so close to the cusp of aridity that it could have, as the Eastern Sahara is documented to have had archaeologically, multiple periods of population followed by near complete depopulation, followed by repopulation from not necessary genetically continous peoples. Major Near Eastern droughts ca. 2000 BCE and 1200 BCE and probably also ca. 800 CE, could have easily fueled cycles like this.

    In short, the assumption that the genetics of the late ancient past resemble the genetics of the present is weaker here than it would be in many places that have been continuously habitable (e.g. the more fertile parts of the Indus Rivery Valley).

    Linguistic evidence also suggests that all branches of Semitic that have left traces in Arabia are not terribly ancient (maybe early iron age or late Bronze Age). Arabia is a pretty plausible region for a hypothetical unattested pre-Semitic language that died at a fairly late date before it could leave written records.

    “The Indian Ocean sea routes are remarkably easy. ”

    There is certainly a case that this kind of sea trade existed in some form ca. 3,000-4,000 years ago. There is documented Harappan-Sumerian trade in this era and there are strong archaeological hints of trade between Ethiopia and Yemen and beyond in this era – although the evidence is still emerging from digs currently in progress on these points. A recent paper by Dorian Fuller suggests, moreover, that Yemeni trade may have been dominated by fairly small scale operations relative to trade between the more urbanized civilizations of that era – rather than being parts of big centrally organized palace/temple centered trading states.

  • Minilik

    1. Yemen was a vassal of Ethiopia. Read the renowned historian Gibbon’s book, the Rise and Fall of the Roman Empire. This fact is also reported on the holy Qu’ran. In both books read, for instance, about the war of Abraha who was an Ethiopian ruler of Yemen. From these and similar other facts, you could have concluded that Queen of Sheba used to rule Ethiopia and her vassal Yemen.

    2. There were some Ethiopian rulers in India like Malik Ambar (1549 – 1626). The Siddi in India claim they are from Ethiopia. Given these facts, you would not be surprised by the similarity between some Indians and Ethiopians.

  • Razib Khan

    siddi are not ethiopian, but more generic east african stock. this has been studied. and the vassalage of yemen to ethiopia far post-dated the era alluded to hear. unless 1,500 years doesn’t mean jack shit to the holy fucking koran.


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