Well, obviously not via Maykop, because it was already present in the pre-Maykop Khvalynsk culture. And also not from Iran_Hotu or Zagros_N -both were shown to be a rather poor fit in Lazarides e.a. 2016. This, in turn, probably has to do with Zagros_N containing some Onge admix that Colchian CHG is lacking (c.f. Lazarides e.a. 2018), and also being more Basal than the latter.
West Georgia (Colchis) as source?
Alright, what about some Upper/ Final Paleolithic or Mesolithic migration from Colchis? Archeology suggests that both took place: As concerns the Paleolithic, layers 3 (around/ shortly after the LGM) and 1 (Younger Dryas/ Epi-Paleolithic) of Kamennaya Balka near Rostov/Don have been archeologically connected to West Georgia, in the latter case with backward linkage to the late Zarzian site of Shanidar in the N. Zagros.
Moreover, while there seems to have beeen a settlement hiatus in NW Caucasia during the LGM, the Late / Final Upper Paleolitic “Imeretian Culture” that is represented a/o by Satsurblia Cave is described to have from ca. 15-14 ky BC onwards expanded into the NW Caucasus, e.g. Gubs Cave (Golanova e.a. 2014, Leonova 2014).
The Mesolithic Kukrek Culture of Crimea and the Southern Ukraine is commonly seen as offspring of the Imereti (Colchian) Mesolithic. The problem here is that we already have aDNA from the latter, namely Kotias (KK1). So, if the CHG in Steppe EMBA (Yamnaya etc.) came via the Kukrek Culture, KK1 should provide an almost perfect match for it. Unfortunately, it doesn’t. [However, the Kukrek Culture offers a plausible path for yDNA J, yet unattested in ANF/EEF, into MN cultures such as Lengyel. Moreover, the unexpected early “Steppe signal” in the Varna outlier may in fact rather reflect an acculturated Mesolithic SW Ukrainian / Moldavian than an early form of Steppe migration from, say, Khvalynsk or Samara.]
So, problem solved ? Actually, not really. With respect to the recent Damgaard e.a. (2018) paper, many, I myself included, initially focused on the Botai and Hittite aDNA, plus the linguistic annex. But that paper holds additional “goodies”, including a demographic model of East Eurasia that a/o dated the Colchian HG (KK1) – “Steppe” CHG split to around 20 kya, i.e. around the LGM. In fact, depending on the specific set-up of their model, they come up with different split dates, but these dates rather tend to be earlier than 20 kya (see their Supplement S. 2.13). Even when allowing for some leeway downwards, a (pre-)KK1 – “Steppe” CHG split as late as the Final Upper Paleolithic, when NW Caucasia was re-colonised from W. Georgia, is clearly not supported by their analysis. Hence, we need to look for a different source that in all likelyhood hasn’t yet been sampled for aDNA.
The South Caspian
Where might that source have found refuge during the LGM? Paleo-climatic models such as the one displayed here from Arpe e.a. 2011 indicate that there was some possibility to survive the LGM on the Balkans, around the Bosporus, in Colchis, and also on the Lower Don, but where one really wanted to be was either NW Iberia, or Central Italy/ the Adriatic Coast, or the Southern Caspian. Of the above, only the Southern Caspian deserves consideration. A CHG-like refugium on the Lower Don should have resulted in much higher CHG frequencies in Mesolithic Ukraine, and around the Bosporus, Barcin-like ancestry is to be expected.
A putative CHG refugium on the South Caspian has three issues:
(1) Paleo-climate models suggest that there may have been a communication channel between West Georgia and the S. Caspian even during the LGM. However, more recent research points towards LGM glaciation of the Caucasus reaching down as low as 500m a.s.l. (which would a/o explain the LGM occupation hiatus at Dzudzuana cave that is located 560 m a.s.l.). In that case, the Colchian and South Caspian refugia would have been effectively isolated from each other, as well as from other presumptive refugia such as the Levante and Mesopotamia, until the the climate improved again around 15 ky BC.
(2) Well evidenced as LGM refugium for a number of species, including the Persian Walnut, the South Caspian so far lacks confirmed evidence of LGM human settlement. The earliest safely dated site is Final UP Komishan (12-11 ky BC), though the lower levels at Kamarband (Belt Cave), reflecting a somewhat different tool composition, may be older.
For lacking connection to world oceans, the Caspian Sea is notorious for substantial sea level fluctuation, with lowstands a/o during the pre-LGM glaciation build-up, and highstands after post-LGM/ post Younger Dryas glacier melt-downs. While the precise timing and extent of these fluctuations is still disputed (compare, e.g. Tudryn e.a. 2016 vs. Chepalyga e.a.), there is consensus about several massive coast line changes since the LGM. South Caspian LGM coastal human sites, if they existed, may now lie some 150 m below current sea level buried under several sediment layers, and even sites closer to or on the current shore are likely to have been covered by Holocene flood sediments [c.f. the Azerbaijani Kura plain that equally witnessed several Holocene floodings, so it took pipeline construction to uncover late 5th mBC Leila Tepe Culture settlements.]
(3) While we don’t have Upper Paleolithic aDNA from the South Caspian so far, there is Mesolithic aDNA, namely Iran_Hotu – and that aDNA doesn’t qualify as source of “Steppe CHG” (see above). However, Iran_Hotu may reflect a late arrival from the Zagros: For inconsistent AMS dates that reach as low as ca. 6,100 BC, Lazarides e.a. 2016 warn that, instead of being Mesolithic, the sample may have been introduced from overlaying Neolithic deposits.
Furthermore, the Final Upper Paleolithic assemblage documented by Jayez/ Nasab 2016 bears little resemblance to contemporary Zarzian assemblages from the Central Zagros, or the Central Georgian Trialetian. It lacks, e.g., burins, (arrow) points, and large geometrics (triangles, trapezoids). The uppermost layer of Komishan, bulldozer-disturbed and as such chronologically unreliable, however, presents a very different picture in terms of technology (pressure-flaking) as well as tool design. A/o, retouched pieces replaced backed ones, and geometric forms appeared – both typical features of the Zarzian. This suggests that Zagros_N – like ancestry only reached the S. Caspian in the EP / early Mesolithic, i.e. contemporary with the Zarzian expansion towards Shanidar in the N. Zagros.
Pre-Neolithic Trans-Caspian connections
In this context, the issue of pressure-flaking as evident a/o in the disturbed Komishan layers is interesting. Pressure-flaking maximises the cutting edge in relation to raw material input and has therefore been the preferred Neolithic knapping technology. It appeared in W. Iran during the early 9th cBC, where it marks the transition between the Zarzian and the Iranian proto-Neolithic (see Tsuneki e.a. 2007), and subsequently spread from there a/o to the Caucasus, pre-pottery Neolithic Anatolia, and possibly Mergarh.
Independent invention of pressure-flaking is a possibility envisaged a/o for cases such as the N. African Capsian that lack apparent connection to other cultures using that technique. However, pressure-flaking requires special skills and, more importantly, evolved into several sub-technologies such as the North East Asian Yubetsu method (attested in Jomon from ca. 13 ky BC) that worked on assymetrical cores, or the symmetrical “bullet shaped” cores preferred in West Eurasia. “Bullet-shaped” cores appeared earliest in the SE Uralian Yangielsk(aya) culture, building onto pressure-flaking of wedge-shaped cores as attested on the Upper Angara and Yenisey (Afontova Gora etc.) from ca. 16 ky BC onwards (Goebel 2002). Acc. to Goebel, pressure-flaking arrived on the Upper Yenisey with Trans-Baikal area populations that re-settled SW Siberia after the LGM, while Vasil’ev 1993 rather sees connections with Mal’ta. Noteworthy in this respect is that SW Siberian “bullet cores”, obviously linked to ANE, did not only spread towards the Near Eastern (and eventually European) Neolithic, but also along a north-western path towards the Mesolithic Baltic Sea (Kama-Perchora, Butovo, Kunda, eventually late Magelmose / Duvensee – good discussion in Szymczak 2002, updated in Harz e.a. 2010).
Iran_Hotu incroporates some 9% EHG (Lazarides e.a. 2016, in f4-stats MA1 is preferred over EHG). Underhill e.a. 2014 report that 21 of their 24 R1a(xR1a1) samples stemmed from Iran, especially Iranian Azeris, and E. Turkey. In all likelyhood, this refers to Near-Eastern-specific (Georgia to Yemen) R1a2b (YP5018) that wasn’t tested for in their study. Acc. to yFull, R1a2b formed some 12.1 kya, i.e. at the end of the Younger Dryas. This suggests epi-ANE/proto-EHG migration out of the SE Urals towards the S. Caspian around the Younger Dryas, eventually introducing pressure-flaking of “bullet-shaped” cores into the Iranian proto-Neolithic.
Details of the migration path are yet unknown. Turkmenistan/ NW Uzbekistan provides an obvious link, and the region’s Kelteminar Culture (best evidenced at Ayakagytma/UZ) is well evidenced for using “bullet-shaped” cores, but Kelteminar is Neolithic and post-dates the arrival of pressure-flaking in Iran. Nevertheless, several researchers have speculated about such path (e.g. A. Collins: “The Cygnus Key..” 2018, Brunet 2012). A problem here is the complete absence of Mesolithic finds from the area, which in turn may relate to frequent floodings of the Aral-Caspian basin not only during the late Pleistocene but also the Holocene (see above). Moreover, under such scenario one would expect R1a2b to also be present in Central Asia that, however, doesn’t seem to be the case (c.f. Underhill e.a. 2014, Table S4).
So, alternatively, one should consider a direct connection. Mesolithic S. Caspians possesed sea-worthy boats, as attested from petroglyphs at Gobustan (see picture above, note the bow-equipped hunter just below attesting Mesolithic rather than Neolithic context) that should have allowed for crossing the Caspian Sea. In fact, they might not even have needed boats, as the Younger Dryas reduced the Caspian Sea to a record lowstand (the Mangyshlak regression) that completely exposed the continental shelf. [During the following flood, at the onset of the Holocene, however, seaworthy boats were certainly of advantage.]
There are also indications of epi-Paleolitic and Mesolithic trans-Caspian Sea contact in the opposite direction: The rather poorly documented SE Uralian Yangielsk(aya) Culture is generally described to display South Caspian influence (e.g. Matyushin 2009, p 137). Acc. to Szymczak 2002, citing Matyushin 1986, the tools from Shikaevka II (on the Tobol somewhat north of Kurgan) correlate with the earliest layers of Kamarband (Belt Cave) that precede the Final Upper Paleolithic (12-11 ky BC) Komishan sequence. Shikaevka II has been C14-dated to ca. 18 ky BP (uncalibrated), which calibrated would be around end of the LGM (Y. Kuzmin 2008), as such (i) placing early Belt Cave into the LGM and (ii) providing a West Siberian entrance point for CHG-like ancestry that is compatible with the Damgaard e.a. 2018 demographic model. However, the dating is questionable:
- It is based on a single measurement of a mammoth bone from the cultural layer. As Kuzmin himself notes, UP humans often scavenged bones from mammoth cemetaries as fuel or for tool making, and there are several cases where mammoth bone dates contradicted other dates from the same site (c.f. Graf 2009 for a more general criticism of Kuzmin 2008). And, in fact, other authors trying to date Shikaevka II based on geomorphology and stratigraphy had come up with dates around 13 ky 14C BP ~ 13,600 calBC;
- It just doesn’t make sense to during the LGM leave the apparently quite cosy South Caspian coast for mammoth hunting in the Taiga 3000 km further north. OTOH, the post-LGM flooding of the Caspian Basin that set in around 15 ky BC provided lots of reasons for migrating north, and the low-lying Turgay pass that was created by a 90 kya glacial spillover into the Caspian basin (Mangerud e.a. 2004) provides a convenient route from Turkmenistan/ W. Kazakhstan to the Tobol river.
As such, I think a 14mBC (13 ky uncal. BP) dating of Shikaevka II, as proposed based on geomorphology and also put forward by Powers 2012, aligns best with both the S. Caspian archeological evidence and climatic/ hydrological considerations.
Acc. to Mosin 2000, the majority of (Late) UP sites of the Southern Urals have been radiocarbon-dated to 12-11 mBC, which raises some concern to which extent Shikaevka II can be regarded as representative, and whether Matyushin’s construct of a Yangielsk Culture that unites Shikaevka with those later attestations is sensible. Nevertheless, for Mosin “the uniform cultural space between the territories of the Transurals in north and the areas East Caspian and Aral region in the south probably at this time is made out.” The author sets forth: “The basic attributes the material culture of the Mesolithic Southern Transurals almost coincides with the characteristics of the Mesolithic Upper Tobol (Evgenjevka I, Duzbay UI, Ubagan III, etc.) and Average Transurals (Seryj Kamen, Vyjka II etc.). On the basis of general characteristic features Mesolithic sites of the specified territories can be united in the Transurals Mesolithic cultural generality. The nearest analogies the sites of the Transurals Mesolithic generality can find in north — with the Mesolithic of the Western Average Ural and in the south — with the Mesolithic sites of the Eastern Caspian.” He/she concludes: “During of the early Holocene the territory of the Transurals in north and the Eastern Caspian and Aral region in the south were closely connected among themselves.“
In conclusion, there seems to have been a trans-(east-)Caspian communication sphere in existence since at least the 14th mBC that a/o transferred South Caspian geometrical and lunate forms northwards, and Siberian pressure-flaking of “bullet-shaped” cores southwards. This communication left visible ANE genetic imprints in N. Iran (Hotu, R1a2), and may also have been a source of the EHG admixture in Namazga_BA , modelled as 79% Iran_N, 21% EHG by Damgaard e.a. 2018.
Mesolithic CHG North of the Caspian Sea ?
Its genetic impact on SW Siberia and NE Europe, however, is uncertain. CHG wasn’t lasting in SW Siberia, as evidenced by Botai aDNA (Botai sites extended towards the upper Tobol, see Brunet 2011), but Botai possibly represens an early Holocene arrival from further east, e.g. the Altai or Baikal.
More suspicious in this respect is the apparent absence of CHG-like ancestry in early Holocene (post-Younger Dryas) Sidelkino north of Samara. That sample is poorly described archeologically, the best I could find is Vasylev e.a. 2014. Nevertheless, general descriptions of the EP/ Early Mesolithic lower and middle Volga stress a dualism of “Western” and “Uralic” traditions. The former is alternatively related to Swiderian (Volotkin 2003) or Ahrensburgian (Galimova 2005) influence, neither of which is particularly likely to have been dominated by EHG, so Sidelkino_ME probably represents the “Uralic” tradition. That tradition is primarily associated with the spread of pressure-flaking of “bullet-shaped” cores, which at the time of Sidelkino_ML acc. to Harz e.a. 2010 had already reached the Early Butovo Culture further NW. The spread of pressure-flaking most likely involved a demic component – Mittnik e.a. 2018 report up to 35% EHG in the (late Kunda) Baltic Mesolithic.
But is CHG really absent from Early Holocene samples? I asked Alberto for some G25 runs – the table to the left presents his best outputs with distal sources. And, apparently, Sidelkino and WestSiberia_N contained some CHG ancestry – not much, but above noise level. The homeopathic CHG in SHG and Botai disappears when Sidelkino and WestSiberia_N are added as sources, meaning it travelled onwards with them.
So, in the early Holocene we find traces of a Trans-Caspian CHG-EHG cline, marked by 9% EHG in Iran_Hotu and 3% CHG in Sidelkino, which mirrors the exchange of stone knapping technologies between the South Caspian and W. Siberia. In theory, somewhere in-between there should have been a point with some 50-50, “Steppe-like” mix. That point may have been the Upper/Middle Tobol (Yangielsk Culture), where the corresponding population either died out during the Younger Dryas, and/or was replaced by “eastern” Botai and WSHGs. But the “equilibrium point” may also have been further south. Steppe admix may have been well present in Kelteminar, as hypothesized by Naramsinhan e.a. 2018 in order to explain early “steppe” signals w/o EEF, i.e. unrelated to Steppe_MLBA, in the Gonur outlier. Moreover – so far we only have one sample from the Caspian coast, namely Iran_Hotu. The 9% EHG there may just have been the tip of the iceberg. IOW – the EHG signals in Armenia_CA and Armenia_EBA may theoretically stem from populations that were already present in the South Caspian during the early Holocene.
So – there remain lots of uncertainties, until we get more Final Upper Paleolithic and Mesolithic samples from the areas in question. Easier said than done – so far there are neither South Caspian archeological sites that can securly be dated to the LGM, nor Final Upper Paleolithic and Early Holocene sites known from the West Asian (Kazakhstan/ Uzbekistan) Steppe (both of which possibly has to do with the frequent cycles of floods and lowstands of the Caspian and Aral Seas). As long as these uncertainties prevail, it is advisable to not instantly attribute each occurence of EHG-CHG admix to the European Steppe (Yamnaya etc.).
The Sidelkino sample, however, makes us reasonably certain that during the early Holocene CHG ancestry wasn’t yet widespread around Samara. The bulk of it only arrived during the Neolithic. When and from where I shall explore in the second part of this series.