The Lady from Central Asia who was found dead in a Bronze Age Turkish well

 

Site of the Alalakh 'Well Lady'. Courtesy: Skourtanioti et al 2020

The remains of the Alalakh 'Well Lady' were discovered in the archaeological site of Tell Atchana / Alalakh / Alalah, Turkey dating to ~1550BCE. Her aDna was published in 2020 (1) and she was found to be of central Asian origin, far from where she was found. She was found at the bottom of a deep well that was still in use at the time. The individual showed evidence of healed trauma on the skull's frontal bone and two healed fractured ribs, and her manner of death has been suggested as a homicide due to or before being thrown into the well (2). Her teeth also showed clear signs of enamel defects (2), a possible sign of malnutrition during childhood. Her age at death was 40-45 years.

The strontium isotope analysis of her dental enamel surprisingly suggested that she was born and raised locally (3), which means that both her parents were migrants. She had genetically inherited shovel-shaped second incisor teeth, and what's of particular interest to us is that there are three other samples from Alalakh with the same kind of teeth. Their DNA results could not be obtained, but Ingman et al 2021 (3) suggest that those three could be genetic outliers as well based on the peculiar incisor teeth.

Ancestral Sources via qpAdm Rotating Models

Fixed References: Mbuti.DG, PPN, IRQ_PPN, EHG, CHG, Iran_GanjDareh_N, Mongolia_North_N, Serbia_IronGates_Mesolithic, Turkey_N, ONG.SG

Rotating Sources/References: Tarim_EMBA1, WSHG, ARM_Masis_Blur_N, Tajikistan_C_Sarazm, IVCp, Turkey_Alalakh_MLBA, IRN_DinkhaTepe_BIA_A, IRN_Hasanlu_LBA_A, IRN_Hasanlu_LBA_B, IRN_Hasanlu_MBA, Uzbekistan_SappaliTepe_BA, Iran_BA1_ShahrISokhta, ARM_Lchashen_LBA, Russia_MLBA_Sintashta

2-Source Models - Only 1 accepted 

The only working 2 source model suggests that the Well Lady was an admixture of 70% Tajik_Sarazm_eneolithic and 30% Hasanlu_LBA_B (Assyrian - West Iran/Iraq) populations.

Someone on Twitter suggested that 3600BCE Sarazm is too far removed in time to be an actual source. That sounds fair, but we won't know if such populations existed in pockets till much later. Many modern people like Kalash & Ror still require Sarazm (or a similar label) to be modelled successfully. The fact is that the fit is superior to other choices. In any case, I also tested models after removing Sarazm from potential sources.

Fixed References: Mbuti.DG, PPN, IRQ_PPN, EHG, CHG, Iran_GanjDareh_N, Mongolia_North_N, Serbia_IronGates_Mesolithic, Turkey_N, ONG.SG

Rotating Sources/References: Tarim_EMBA1, WSHG, IVCp, IRN_Hasanlu_LBA_B, Uzbekistan_SappaliTepe_BA, Iran_BA1_ShahrISokhta, ARM_Lchashen_LBA, Russia_MLBA_Sintashta, Kyrgyzstan_Aygirdjal_BA, Turkmenistan_Gonur_BA_1, Russia_Steppe_Catacomb

3-source models

Removing Sarazm, 3-source models are required for passing p-values above 0.05. Among the eastern/central Asian sources - both Gonur (BMAC) as well as Shahr-Sokhta (East Iranian) work and form the most significant ancestry components. The second necessary source is something with additional Yamnaya-like or ANE-rich ancestry - like Aygirjdal_BA or Steppe_Catacomb. Aygirjdal is a more likely source as it is proximal to BMAC in the Oxus region. The 3rd source is a minor Assyrian source like Hasanlu_LBA or from Armenia_Lchashen_LBA. Importantly, Sintashta-related steppe sources are rejected.

The Outliers from Megiddo, Israel

Bronze age Megiddo is best known as the site of the epic battle dated to the 15th century BCE between Egyptians and Canaanites and their allies. This region started receiving CHG/Iran-related ancestry before 2400 BCE from the Caucasus region (4), resulting in a homogenous population in the bronze age, labelled as Megiddo_MLBA or Israel_MLBA.

Within the Megiddo samples, there are three outliers of which 2 infants are brother-sister siblings. The authors of the paper labelled them as Israel_MLBA_oCaucasus because they think that's where they came from. The authors note:

These individuals are unlikely to be first generation migrants, as strontium isotope analysis on the two outlier siblings (I2189 and I2200) suggests that they were raised locally. This implies that the Megiddo outliers might be descendants of people who arrived in recent generations. Direct support for this hypothesis comes from the fact that in sensitive qpAdm modeling (including closely related sets of outgroups), the only working northeast source population for these two individuals is the contemporaneous Armenia_MLBA, whereas the earlier Iran_ChL and Armenia_EBA do not fit . The addition of Iran_ChL to the set of outgroups does not change this result or cause model failure. Finally, no other Levantine group shows a similar admixture pattern . This shows that some level of gene flow into the Levant took place during the later phases of the Bronze Age and suggests that the source of this gene flow was the Caucasus.

Now, one of the siblings, the boy, has Y Hg R1a1a1. The 'steppe enthusiasts' in the online genetic community, who probably themselves have R1a+ Y Hg and thus the enthusiasm, jumped at this with the claim that they had proof of Mitanni's presence in the Megiddo war. 

...whereas the bone of one of the three (I10100) was directly dated (1688–1535 BCE, ± 2S)

These outliers are dated to before the formation of Mitanni and much before the Megiddo war. There is 0 archaeological context of them being related to Mitanni. Anyway, let us proceed to find the best ancestry sources for these 3.

The parameters will be

Fixed References: Mbuti.DG, IRQ_PPN, EHG, CHG, Iran_GanjDareh_N, Mongolia_North_N, Serbia_IronGates_Mesolithic, Turkey_N, ONG.SG, Tarim_EMBA1

Rotating Sources/References: Tajikistan_C_Sarazm, IVCp, Turkey_Alalakh_MLBA, ARM_Tavshut_Trialeti_MBA, ARM_Lchashen_LBA, Turkmenistan_Gonur_BA_1,Russia_Steppe_Catacomb, IRN_Hasanlu_MBA, Israel_MLBA, Israel_C, Jordan_EBA, Russia_MLBA_Sintashta, Turkey_Arslantepe_EBA

Allnsnps: YES, inbreed: NO

1240K dataset from Harvard will be used, no SNP filtering.

The sister - Sample ID I2200

2-source models for I2200 (Link for all results is provided later in the article)

The sample I2200 comes from a local Israeli and a bronze-age Armenian. The results agree with that of the paper. Sintashta-related Steppe_MLBA ancestry is NOT involved. All models with Sintashta are either infeasible or with failing p-values. She has 611k SNPs, which is very good and ensures that very very few non-optimal models can pass if at all they pass.

Her mtDna is U3b, first found at Turkey in 5000BCE. Which tells us that her mother was a local and that her father was not. Her father was an R1a1a1 carrying Armenian/Caucasian man.

The R1a1a brother should also have a similar ancestry profile, let us check. His SNP count is a bit on the low side at 130k (~90% missingness), so non-optimal models will pass too.

The brother - Sample ID I2189

2 source models for I2189

As expected, the Israel_MLBA + Caucasus/West Iran models all pass. However, the Israel_MLBA + Sintashta model passes too. In conjunction with R1a1a1, does this mean there is Sintashta-related ancestry present here?

Well, no. Since the 1st relative (sister) only accepts the Israel_MLBA + Catacomb/ArmeniaBA models, that will be taken as the chosen model here. Also, it is unlikely that Sintashta-related ancestry reached the Levant via Oxus and Iran untouched by any other ancestry on the route. This model passes because of the lower power of qpAdm on samples with high missingness (low SNP count) (5).

As discussed earlier for his sister, this boy's father was likely an R1a1a1 man from Armenia. Maykop, just north of Caucasus, had 1 R1a sample from 3100bce. In Armenia, 1 R1a-Z645 sample has been found from 300BCE, giving credence to the possibility of an Armenian man from 1600BCE possessing R1a. Furthermore, 2 R1b (R-M269 of Yamnaya-like origin) have also been found at Megiddo_MLBA, although they don't show any extra Yamnaya-like autosomal ancestry. This suggests minor genetic contact between Caucasus and Levant.

3rd Megiddo Outlier - Sample Id I10100

This sample is labelled Israel_MLBA_oCaucasus by the authors.

3 source model for I10100

For this sample, a 3 source model is needed. Israel_mlba + Tajik_Sarazm or Gonur (BMAC) + Steppe_Catacomb work. A 3rd marginal p-value model with Sintashta also works. The sample is a female with mtDNA T2 (which is local, so again a foreigner father).

The other reason to reject the Sintashta model is that all our previous samples require some Yamnaya-related ancestry rather than a Sintashta-related one. She has 246k SNPs (~80% missingness) so few non-optimal models can pass as well.

Results of all rotating models for all the above targets are found in this folder.

DISCUSSION

Some or all of these 4 samples have been suggested by 'steppe enthusiasts' on online communities as possibly being Mitanni elites.  This is why I took the time to address these samples and critically analyze their ancestries to get the best models. Steppe from the Caucasus (Catacomb culture or via Armenia) cannot be related to the Mitanni Aryans as they are Yamnaya descended whereas the Kurgan theory posits Sintashta-related steppe_mlba ancestry as proto Indo-Iranian. It is essential to build models which can differentiate between the two, otherwise 'steppe enthusiasts' go on confirming their biases on a whim.

My own position is that Iranians in the west of Iran came from the Oxus/East Iran, as is also seen in the Well lady as well the recently published Hasanlu and Dinkha Tepe samples from NW Iran. I also believe that the Mitanni were connected to elites from the East of Iran/ Afghanistan/ NW India region, so the well Lady fits well into this hypothesis. However, there are some problems with the Well Lady's context:

a) There is no archaeological context which proves that she was a Mitanni elite. On the contrary, she was malnourished as a child (tooth enamel deficiency), was either a warrior or was physically abused (healed skull and healed ribs). She was murdered and thrown into a well and she landed on her front.

b) To me this seems like a case of a woman born locally to migrants from the east, and later being abused physically (either by her spouse or by her master if she was a servant).

c) There is a possibility that her injuries are because she was a warrior (and therefore an elite), but this is countered by her dental record (enamel deficiencies during childhood)

As for the Megiddo outliers, there is no context suggesting that they were Mitanni, and the carbon date from one of the samples is before or at the beginning of the establishment of the Mitanni empire. As such, they seem irrelevant to the question of Indo-Iranians.

Why is Mitanni important?

By 1500BCE at the latest, and 1700BCE at the earliest, there are traces of Indo-Aryan words and proper names found in the archaeological record in the near east. While little is known from Mitanni sites, most of the knowledge about the Mitannis comes from the empires they interacted with. For example, in the Alalakh period IV (15th century BCE), the region became a vassal state of Mitanni and various tablets recovered from Alalakh carry information about their dealings with Mitanni and their King. Although the Mitanni subjects were Hurrian speakers, many of the words are Indo-Iranian in nature. 

Witzel 2019 (7) describes 

However, a major milestone is the appearance of early IA speakers in the near east, on the northern rims of Mesopotamia. They have left many loanwords in a non-related Caucasus language, Mitanni Hurrite, from ca. 1600 BCE onwards. The Mitanni kingdom covered northern Iraq and Syria; IA speakers will have reached it from its eastern border in the Zagros mountains. The loanwords extend from horse culture (colors, number of rounds in chariot races), to the throne names of Mitanni kings, and a brief list of four Vedic deities (Mitra, Varuṇa, Aryaman, Nāsatya) attached to a treaty with the Hittites (1380 BCE).

So, a formerly Hurrian region gets a new dynasty and there appear a lot of words of Indo-Iranian origin. While Witzel (and many others) favours an Indo-Aryan nature of these words, Benedetti (8) argues that their origin is Iranian rather than Indo-Aryan in nature. The third option of proto-Nuristani is supported by Diakonoff(1996)(12). Based on some of the Mitanni words like 'sattawartanna' and other loanwords found in Akkadian texts, Diakonoff concludes:

If so, this suggests a presence of proto-Dardic tribes not only in Mitanni but also in northeastern Iran at an early epoch (say, from the middle of the second millennium B.C.E. to ca. the ninth century B.C.E.)

We have tried to show that linguistic data seem to point to the possibility that ancient Iran could have been invaded by earlier Indo-Iranian Dardo-Nuristani tribes before the arrival of Iranian-speaking tribes, that is, the Medes, about the middle of the second millennium B.C.E.

The Dardo-Nuristani option is interesting because the Kalash people need to be modelled with Sarazm as a source, and Sarazm is also the ancestry we see all over these West Asian outliers (Alalakh 'well lady' and Megiddo I10100), and also at Hasanlu.

Rotating Source models for Kalash.SDG using Admixtools 2 on the 1240K dataset. Extract_f2 settings: maxmiss=1, Transitions=False, Transversions=True (228500 SNPs). Transitions are set to false because True gives no working models. This may be due to the differential rate of ancient damage on transition sites of the ancient sources (per Harney et al 2021). Hence we set transitions=FALSE. References and rotating sources are mentioned at end of the article.

Witzel 2019 (7) argues that some of these words were older than the language of the Rig Veda and therefore sets a terminus a quem date of RigVeda to 1400BCE. However, this sees strong refutation by Talageri (10) in his books (and adapted blog posts). Talageri argues that all the Indo-Aryan names found in the Mitanni, Nuzi and Syrian documents match only with the later Mandalas (New Books) of the Rig Veda, and therefore are much later than the earliest part of Rig Veda.

I am not in a position to judge the linguistic aspects of these arguments myself, however, I side with Talageri in this debate. The weakness of Witzel's argument, in my eyes, is that he assumes some monolithic language block across the Indo-Iranian lands, whereas the more likely scenario is that various regions had various dialects and different levels of preservation of Indo-Iranian archaisms and innovations (as is the case for all extant IE languages), some of these dialects may have been lost forever.

Another problem with Witzel's paper is that in 2019, he still quotes the hoax of the Potapovka Dadhyañc myth which was debunked 20 years back and accepted by none other than David Anthony (11). This erroneous supposition which put the Steppe Aryan theory into overdrive in the 1990s still biases his opinion. From Witzel 2019 (7)

Remnants of it have recently been discovered at Krasno-Samarskoye, just west of the Urals. In the same area, some 40 years ago, a grave has been found with a headless body; instead a horse head had been substituted, just as in the Ṛgvedic Dadhyañc myth.

So how did these words get there? There are two options:

1. The Mitanni kings were Indo-Iranians who conquered and ruled over Hurrian subjects (mainstream view). or

2. The Mitanni kings were locals but they borrowed culture and loanwords from their Indo-Iranian neighbours to their east. (6)

In either scenario, we should expect to see the genetic trail of the Indo-Iranians from the east toward Syria or western Iran latest by 1600BCE. This is the weakest link for the Kurgan theory because they should expect to see Sintashta Indo-Iranian genetic trail till western Iran. But nothing of this sort is to be seen. This is why 'steppe enthusiasts' jump at every opportunity and make crude models to prove the presence of minor at least some 3-4% Sintashta ancestry in the near east, to establish the link from Sintashta to Mitanni.

The latest data from Lazaridis et al 2022 (9) shows the presence of fresh Bactria Margiana Complex-related ancestry, along with minor IVC-related ancestry at Hasanlu Tepe and Dinkha Tepe. This inflow can clearly be seen in the 1500-1300BCE samples from this site, whereas it is completely absent in the 2000-1900BCE samples.

Hasanlu and Dinkha Tepe samples and their implications have been covered extensively in this post.

The Southern Arc paper and its data upends the Steppe Theory in multiple ways

Right Pops for Kalash.SDG

Fixed Right (Only in reference, not tested as source):

Mbuti.DG, EHG, CHG, Marmara_Barcin_N, Tarim_EMBA1, WSHG, Iran_GanjDareh_N, ONG.SG, Mongolia_North_N, Serbia_IronGates_HG

Rotating Sources/References (Tested as source as well as kept in reference when not in source):

Nepal_Samdzong_1500BP.SG, Iran_BA1_ShahrISokhta, Tajikistan_Ksirov_Kushan, Uzbekistan_Dzharkutan_BA_1, Kazakhstan_Kangju.SG, Kazakhstan_MLBA_Dali, Russia_MLBA_Sintashta, Sarazm_En, Irula.DG, Turkmenistan_IA, PAK_Loebanr_IA, Pakistan_Loebanr_IA_o, Pakistan_Aligrama_IA, PAK_Saidu_Sharif_H, PAK_Udegram_Medieval_Ghaznavid

Kalash.SDG Results here.

APPENDED: 3-Nov-2022

Updated models for the Megiddo outliers after adding viable Balkan sources in the rotating models.

Using the newer Harvard Allen dataset v52.2, so label names have changed a bit.

Fixed Right Sources: Mbuti.DG, EHG, CHG, WSHG, TUR_Marmara_Barcin_N, IRN_Ganj_Dareh_N, ONG.SG, MNG_North_N, SRB_Iron_Gates_HG, IRQ_PPN, PPN, ARM_Aknashen_N

Rotating Right Sources: RUS_Catacomb, IRN_Shahr_I_Sokhta_BA, TJK_Sarazm_EN, IRN_Shahr_I_Sokhta_BA2, TUR_Hatay_Alalakh_MLBA, IRN_DinkhaTepe_BIA_B, IRN_Hasanlu_LBA_A, IRN_Hasanlu_LBA_B, IRN_Hasanlu_MBA, IRN_DinkhaTepe_BIA_A, RUS_Sintashta_MLBA, ARM_Lchashen_LBA, ARM_Tavshut_Trialeti_MBA, UZB_Dzharkutan1_BA, BGR_NovaZagora_Yamnaya_EBA, BGR_Merichleri_MLBA, ISR_Canaanite_MLBA

Models for Megiddo Sister I2200

With the addition of Arm_Trialeti_MBA, Bulgaria_Yamnaya, and Bulgaria_MLBA as sources, they are chosen as appropriate sources in the new rotating models. Both the sources selected in the original article (Catacomb & Lchashen_LBA) are also rejected in favour of these new sources. Sintashta is again rejected as a source.

Models for Megiddo Brother I2189

Trialeti_MBA, Bulgaria_yamnaya, & Bulgaria_MLBA are again selected for the Megiddo brother. Two other sources are also selected - Sintashta and Catacomb, those can be rejected because the same models don't work for the sister.

Bulgarian sources can also explain R1a1a1 in this boy as I2183 Bulgaria_MLBA is also R1a1a1, and is close in date to the Megiddo site. Indeed, Bulgaria is also closer to Megiddo than north Kazakhstan.

Model for unrelated sample Megiddo_o2 I10100

2 source models do not cross the p threshold of 0.05 so 3 source models are required. Multiple steppe sources work for this sample, however, Arm_Trialeti DOES NOT work. If we want to find commonality with the other two Megiddo siblings, then perhaps we should be looking at Bulgaria_Yamnaya or Bulgaria_MLBA as the source for this sample as well. But this sample may be independent and have other combinations of ancestry as well, so the jury is still out. Whatever the case, this sample surely has excess South Central Asian ancestry that is not present in the Megiddo siblings. This excess SC Asian is contained in labels - Sarazm, Dzharkutan (BMAC), Dinkha_BIA_B, Hasanlu_LBA_A and Shahr_Sokhta_BA.

Link for all the new rotating models, Admixtools parameters used, and all the Sample Ids and data sources used are in Google Sheets here.

REFERENCES

(1) Skourtanioti E, Erdal YS, Frangipane M, et al. Genomic History of Neolithic to Bronze Age Anatolia, Northern Levant, and Southern Caucasus. Cell. 2020;181(5):1158-1175.e28. doi:10.1016/j.cell.2020.04.044

(2) Shafiq R. Come and Hear My Story: The ’Well-Lady’ of Alalakh. In: Yener KA, Ingman T, editors. Alalakh and its Neighbors: Proceedings of the 15th Anniversary Symposium at the New Hatay Archaeology Museum, June 10–12, 2015. Leiden: Peeters; 2020. p. 433–52.

(3) Ingman T, Eisenmann S, Skourtanioti E, Akar M, Ilgner J, et al. (2021) Human mobility at Tell Atchana (Alalakh), Hatay, Turkey during the 2nd millennium BC: Integration of isotopic and genomic evidence. PLOS ONE 16(6): e0241883. https://doi.org/10.1371/journal.pone.0241883

(4) Agranat-Tamir L, Waldman S, Martin MAS, et al. The Genomic History of the Bronze Age Southern Levant. Cell. 2020;181(5):1146-1157.e11. doi:10.1016/j.cell.2020.04.024

(5) Éadaoin Harney, Nick Patterson, David Reich, John Wakeley, Assessing the performance of qpAdm: a statistical tool for studying population admixture, Genetics, Volume 217, Issue 4, April 2021, iyaa045, https://doi.org/10.1093/genetics/iyaa045

(6) von Dassow, E. (2014). Levantine polities under Mittanian hegemony. In Constituent, Confederate, and Conquered Space: The Emergence of the Mittani State (pp. 11-32). Walter de Gruyter GmbH.

(7) Witzel, Michael. (2019). Early ‘Aryans’ and their neighbours outside and inside India. Journal of Biosciences. 44. 10.1007/s12038-019-9881-7. 

(8) Benedetti, Giacomo. “Were the Mitanni Aryans really Indo-Aryans?” Were the Mitanni Aryans really Indo-Aryans?, 24 May 2017, http://new-indology.blogspot.com/2017/05/were-mitanni-really-indo-aryans.html. 

(9) Lazaridis, Iosif, et al. “The Genetic History of the Southern Arc: A Bridge between West Asia and Europe.” Science, vol. 377, no. 6609, 2022, https://doi.org/10.1126/science.abm4247.

(10) Talageri, Shrikant. “Two Papers by the Renowned Indologist P.E.Dumont.” Two Papers by the Renowned Indologist P.E.Dumont, 7 May 2016, https://talageri.blogspot.com/2016/05/two-papers-by-renowned-indologist.html. 

(11) K, Ashish. “David Anthony, Michael Witzel, Asko Parpola & the Potapovka Horse Head Burial Hoax.”, 14 Dec. 2021, https://a-genetics.blogspot.com/2021/12/potapovka-horsehead-hoax.html. 

(12) DIAKONOFF, I. M. (1996). Pre-Median Indo-Iranian Tribes in Northern Iran? Bulletin of the Asia Institute, 10, 11–13. http://www.jstor.org/stable/24048881

(13)  Éadaoin Harney, Nick Patterson, David Reich, John Wakeley, Assessing the performance of qpAdm: a statistical tool for studying population admixture, Genetics, Volume 217, Issue 4, April 2021, iyaa045, https://doi.org/10.1093/genetics/iyaa045