The Scythians: Who Were They?

Genetic relationship and origin of the Scythian groups

The eastern sample group (n=113) can be divided in four cultural subgroups chronologically dispersed over the 1st millennium BCE (Fig. 2). Analysing mtDNA, we found no significant genetic distance between those groups (Supplementary Table 8). We then used approximate Bayesian computation (ABC)28 to test for continuity between the earlier (#4: Zevakino-Chilikta and #5: Aldy Bel; n=26) and the later (#6: Pazyryk; n=71) Scythian period in the East. The Tagar/Tes group (#7) had to be excluded because of their imprecise dating. These analyses revealed that they were most likely derived from one single population that was expanding over the time period considered here, that is, the two samples are unlikely to represent two independent populations that diverged earlier than 108 generations before present (g BP) or ∼2.7 ky BP. This scenario was highly supported in our model selection procedure (Supplementary Table 9, logistic regression, P=0.995, neural networks P=0.568, cf. confidence in model choice in Supplementary Table 10, model parameter posteriors in Supplementary Table 11), whereas two scenarios that assumed that the eastern Scythian sample groups were derived from two previously diverged populations received very little statistical support (cumulative posterior probabilities: logistic regression, P<0.001, neural networks, P=0.001).

Since the genetic distance between the combined Scythian groups of the east versus those of the west is relatively low (FST=0.01733; P-value=0.02148±0.0045), we used ABC to further assess if the eastern and western Scythians might share a common origin (Fig. 3). For these analyses we included contemporary samples representative of genetic diversity on the extremes of Eurasia (Supplementary Note 1). According to our model selection algorithm, a multiregional model provided the best fit to the empirically observed diversity patterns (Supplementary Table 12, 0.5% closest simulations, posterior probability P=0.708 for logistic regression, P=0.715 for neural networks method), while a model of western origin also received some support (Supplementary Table 12, 0.5% closest simulations, posterior probability P=0.286 for logistic regression, P=0.267 for neural networks method). Therefore, in contrast to the eastern origin model, a western origin cannot be fully discounted by our analysis. In addition, a pairwise comparison through the computation of Bayes factors reveals a substantial to strong (logistic regression) or weak to substantial (neural network) support for the multiregional origin over the western origin model (see Supplementary Note 1, Supplementary Table 14 and Supplementary Fig. 5b). These results suggest that western and eastern Scythian groups arose independently—perhaps in their respective geographic regions—and thereafter experienced significant population expansions (during the 1st millennium BCE). Importantly, our simulations support a continuous gene flow between the Iron Age Scythian groups, with indications of asymmetrical gene flow from western to eastern groups, rather than the reverse (see Supplementary Tables 13 and 15 for details).


Figure 3: Candidate scenarios for the origin of Scythian populations.

Because population movements across Central Asia during the Bronze Age are often archaeologically associated with the spread of the Andronovo culture29, we used ABC to fit a sample of Middle Bronze Age nomadic groups from western Siberia, most of them associated with the Andronovo culture, onto the preferred demographic model for the origin of Scythians. For this purpose—and based on low FST values between these groups—we combined 40 samples related to the Andronovo culture in the west Siberian forest steppe30 and nine samples from the same culture in the Krasnoyarsk region31, all of which were dated to the first half of the 2nd millennium BCE. The results provided very strong support for a linkage between these Middle Bronze Age groups and eastern Scythians (Supplementary Tables 16 and 17). However, these simulations were not able to fully capture the patterns of genetic diversity observed in the Bronze Age populations, suggesting that the true demographic history of the ancestry of Iron Age populations may have been more complex than considered here (see Supplementary Note 1 and 32 for details).


Figure 4: Principal component analysis.

Genetic diversity and ancestry of the Scythian groups

Haplogroups found in the Iron Age nomads are predominant in modern populations in both west (HV, N1, J, T, U, K, W, I, X) and east Eurasia (A, C, D, F, G, M, Y, Z). The mitochondrial haplotype diversity in our sample set ranges from 0.958±0.036 in the Tagar/Tes sample (group #7 in Fig. 2) up to 1.000±0.039 in the Early Sarmatians (#3; Supplementary Table 7).

Using nuclear SNP data, we performed a principal component analysis33 (PCA) of 777 present-day west Eurasians26,34,35 onto which we projected the eight newly reported Iron Age Scythian samples as well as 167 other ancient samples from Europe, the Caucasus and Siberia from the literature17,34,36 (Fig. 4). The two Early Sarmatian samples from the West (group #3 in Fig. 2) fall close to an Iron Age sample from the Samara district34 and are generally close to the Early Bronze Age Yamnaya samples from Samara34 and Kalmykia17 and the Middle Bronze Age Poltavka samples from Samara34. The eastern samples from Pazyryk (#6), Aldy Bel (#5) and Zevakino-Chilikta (#4) are part of a loose cluster with other samples from Central Asia17, including those from Okunevo, Late Bronze Age and Iron Age Russia, and Karasuk. These samples contrast with earlier samples from the Eurasian Steppe belonging to the Andronovo17, Sintashta17 and Srubnaya34 groups, which overlap Late Neolithic/Bronze Age individuals from mainland Europe34,35 and are shifted downwards in the PCA plot towards the early farmers of Europe and Anatolia34.

Descendants of the Iron Age Scythians



A multidimensional scaling (MDS) plot based on Reynolds’ distances (Supplementary Fig. 1) suggests that the ancient Scythian populations from the eastern and western part of the Eurasian Steppe are genetically closer to each other than are the modern populations of the respective regions. AMOVA analyses carried out for modern and ancient groups of the eastern and western steppe provided further support for this finding. We found FCT values to be higher between modern populations of the East and the West (FCT=0.0835) than between ancient populations of similar regions (FCT=0.0262).

A continuity test was performed between the two Iron Age groups (‘West’ and ‘East’) and a large set of contemporary Eurasian populations (n=86, Supplementary Table 19). For western Scythian-era samples, contemporary populations with high statistical support for a genealogical link are located mainly in close geographical proximity, whereas contemporary groups with high statistical support for descent from eastern Scythians are distributed over a wider geographical range. Contemporary populations linked to western Iron Age steppe people can be found among diverse ethnic groups in the Caucasus, Russia and Central Asia (spread across many Iranian and other Indo-European speaking groups), whereas populations with genetic similarities to eastern Scythian groups are found almost exclusively among Turkic language speakers (Supplementary Figs 10 and 11).



Phenotypic markers

Derived alleles of pigmentation markers that are under selection in Europeans are present in eastern and western Scythians, including individuals who are homozygous for the derived alleles at selected SNPs in the HERC2, SLC24A5 or SLC45A2 (ref. 43, 44, 45). At the two LCT loci associated with lactase persistence, the derived allele is observed only in heterozygotes, only in the eastern Scythian samples, and at low frequency (2–3%). The ancestral alleles at ADH1B rs3811801 and rs1229984 are nearly fixed in the Scythian dataset, as they are in modern Europeans (the derived alleles, which confer some resistance to alcoholism, are under selection in East Asians46,47). We observe the derived allele at rs3827760 in the EDAR gene in a single Pazyryk individual (#6 in Fig. 2). This EDAR derived allele, which is related to tooth and hair morphology, is selected and at high frequency in modern East Asians (87%)48, and very rare in modern Europeans (∼1%)48, although it has been observed in prehistoric hunter-gatherers from Sweden (7.9–7.5 kya)34. Thus, the results of the examination of phenotypic SNPs that show frequency differences between Europe and East Asia are consistent with gene flow across the steppe territory.



Our genomic analyses reveal that western and eastern steppe inhabitants possess east Eurasian ancestry to varying degrees. In our ADMIXTURE analyses we find an East Asian ancestry component at K=15 in all Iron Age samples that has not been detected in preceding Bronze Age populations in either western or eastern parts of the Eurasian Steppe. Another ancestral component that is maximized in the north Siberian Nganasan population becomes visible from the 2nd millennium BCE onwards in the eastern steppe (Okunevo, Karasuk, Mezhovskaya). This component appears later in all Iron Age populations but with significantly higher levels in the eastern steppe zone than in the West. These findings are consistent with the appearance of east Eurasian mitochondrial lineages in the western Scythians during the Iron Age, and imply gene-flow or migration over the Eurasian Steppe belt carrying East Asian/North Siberian ancestry from the East to the West as far as the Don-Volga region in southern Russia. In general, gene-flow between eastern and western Eurasia seems to have been more intense during the Iron Age than in modern times, which is congruent with the view of the Iron Age populations of the Eurasian Steppe being highly mobile semi-nomadic horse-riding groups.



In the East, we find a balanced mixture of mitochondrial lineages found today predominantly in west Eurasians, including a significant proportion of prehistoric hunter-gatherer lineages, and lineages that are at high frequency in modern Central and East Asians already in the earliest Iron Age individuals dating to the ninth to seventh century BCE and an even earlier mtDNA sample from Bronze Age Mongolia49. Typical west Eurasian mtDNA lineages are also present in the Tarim Basin16 and Kazakhstan8 and were even predominant in the Krasnoyarsk area during the 2nd millennium BCE31. This pattern points to an admixture process between west and east Eurasian populations that began in earlier periods, certainly before the 1st millennium BCE13,50, a finding consistent with a recent study suggesting the carriers of the Yamnaya culture are genetically indistinguishable from the Afanasievo culture peoples of the Altai-Sayan region. This further implies that carriers of the Yamnaya culture migrated not only into Europe26 but also eastward, carrying west Eurasian genes—and potentially also Indo-European languages—to this region17. All of these observations provide evidence that the prevalent genetic pattern does not simply follow an isolation-by-distance model but involves significant gene flow over large distances.

Nature Communications 8, Article number: 14615 (2017)