The Hittites: an Indo-European people without steppe ancestry, or how genetics resolved the Anatolian paradox

The Hittites occupy a unique place in the linguistic prehistory of humanity. Their language, attested by the earliest written records of any Indo-European tongue (the cuneiform tablets of Hattusa, around 1650 BCE), proves that a branch of the Indo-European family had established itself in central Anatolia by the second millennium. And yet, when the ancient DNA of these populations was finally sequenced, it revealed a spectacular anomaly: the Bronze Age Anatolians, including individuals from the Hittite period, carry almost no trace of the steppe ancestry (the Eastern hunter-gatherer, or EHG, signal) that characterises every other Indo-European population, from Northern Europeans to the Indo-Aryans. A people speaking an Indo-European language but lacking the genetic marker of the steppe: this is the Anatolian paradox, long regarded as the single most serious objection to the steppe theory of Indo-European origins. This article documents that paradox on Global25 with direct figures, then shows how three waves of studies (Damgaard 2018, the Southern Arc of Lazaridis 2022, and the major 2025 paper on the origin of the Indo-Europeans) not only confirmed it but resolved it. The key fits in a single sentence: the Hittites do not descend from the Yamnaya, but from an earlier common ancestor located south of the steppe, who split off before the EHG component was grafted onto the steppe lineage. The absence of a steppe signal is not a flaw in the theory: it is the very signature of the oldest branch of the Indo-European tree.

1. Who were the Hittites?

The Hittites are, strictly speaking, the Hittite-speaking people who dominated central Anatolia during the late Bronze Age. They did not call themselves Hittites: the term is a modern convention derived from the land of Hatti, which was in fact the name of the earlier substrate. They referred to themselves as speakers of Nesili, the language of the city of Nesa (Kanesh). Their capital, Hattusa, was founded on the site of an older Hattic city, and it is in its archives that the oldest written attestations of any Indo-European language, inscribed on thousands of clay tablets in cuneiform script, were found, dated to the seventeenth century BCE.

The Hittite language belongs to the Anatolian branch of the Indo-European family, which it shares with Luwian, Palaic, and later Lydian, Lycian, and Carian. This branch displays a remarkable linguistic peculiarity: it preserves archaic features that all other Indo-European languages have lost, foremost among them the laryngeal consonants, whose existence had been postulated by Ferdinand de Saussure on the basis of internal reconstruction alone, decades before their actual discovery in the Hittite texts. It is precisely this archaism that led, as early as the twentieth century, to Edgar Sturtevant's Indo-Hittite (or Indo-Anatolian) hypothesis: Anatolian would not be a branch like the others, but the very first split, separated from the rest of the family before it developed its common innovations. As we shall see, the genetics of 2025 spectacularly confirmed this century-old intuition.

Politically, the Hittite empire was one of the great powers of the Bronze Age Near East, a rival of pharaonic Egypt (the Battle of Qadesh against Ramesses II, around 1274 BCE, led to the oldest surviving international peace treaty) and of Assyria. Its collapse, around 1180 BCE, was part of the great systemic crisis of the end of the Bronze Age that simultaneously brought down Mycenae, Ugarit, and so many other centres. The Anatolian languages survived it for a few centuries in the small Neo-Hittite kingdoms of the northern Levant and southern Anatolia, before dying out definitively around the turn of the common era.

2. The Anatolian paradox: an Indo-European language without the steppe

To grasp why Hittite DNA landed like a bombshell, the dominant model must be recalled. Since the work of 2015 (Haak, Allentoft), a consensus had formed: the Indo-European languages spread with the Yamnaya pastoralists of the Pontic-Caspian steppe, and their characteristic genetic marker is a roughly equal mixture of Eastern hunter-gatherer (EHG) and Caucasian ancestry (CHG). Wherever ancient Indo-European languages are found, in Northern Europe, in Central Asia, in India, this steppe ancestry is found too. The implicit reasoning had become: steppe equals Indo-European.

The Hittites should, according to this scheme, have carried a substantial dose of steppe ancestry, since they spoke the most anciently attested of all Indo-European languages. Yet the first sequencing, published by de Barros Damgaard and colleagues in 2018, showed exactly the opposite. To use the phrasing of their abstract: the Bronze Age Anatolian samples, including those from Hittite-speaking settlements associated with the first written evidence of Indo-European languages, show genetic continuity with the preceding Anatolian Chalcolithic samples and possess a substantial Caucasian hunter-gatherer (CHG) ancestry, but no evidence of direct steppe admixture.

3. Three waves of studies, one resolution

The Hittite genetic dossier was built in three stages, worth distinguishing clearly because they answer different questions.

The overall logic is as follows. There existed, around 4500 to 4000 BCE, a population (or rather a genetic cline) stretched between the Caucasus Neolithic in the south and the lower Volga in the north, suffused with CHG ancestry: this is the Caucasus-Lower Volga cline. From this source, two divergent fates. Toward the north and west, it mixes with Eastern hunter-gatherers (EHG) along the Volga and the Dnipro, giving the Serednii Stih groups and then, around 4000 to 3300 BCE, the Yamnaya: it is the acquisition of EHG that forges the steppe signature. Toward the south, a fraction of the same CLV population enters Anatolia from the east (along the Areni-Arslantepe-Ovaören-Kalehöyük axis), without ever meeting the EHG. The Anatolian speakers descend from this second branch. This is why they have no steppe signal: they left the tree before the steppe branch grew.

4. The Hittites on Global25: a Bronze Age Anatolian, nothing else

When all ancient populations in the Global25 panel are ranked by Euclidean distance from the average individual of the Hittite period, the result is unambiguous: the closest neighbours are all Anatolians, and the order of magnitude of the distances tells the story on its own. The Hittite individual is glued to its local Anatolian predecessors and successors, and astronomically far from the nearest steppe population.

The contrast is numerically overwhelming. The Hittite individual sits at 0.0200 from its Chalcolithic Anatolian predecessor, and at 0.2199 at minimum from the nearest Yamnaya population: it is therefore roughly eleven to thirteen times closer to the local Anatolian soil than to the steppe. More telling still, it is markedly closer to the Caucasus Neolithic (Aknashen, at 0.1009) than to any steppe group. The eastern ancestry of the Hittites points south, toward the Caucasus, never toward the north.

5. The decisive test: EHG offered to the algorithm, and refused

A distance does not prove an absence of admixture: a point can be far from a source and still contain a fraction of it. The rigorous test is to explicitly offer the steppe source to the mixture algorithm (NNLS) and observe the weight it assigns to it. If the Hittites contained even a fraction of steppe ancestry, the algorithm would capture it. Here is the distal model of the average Hittite individual, decomposed over five fundamental sources of the Near East and the steppe, with EHG (the steppe proxy) being one of them.

6. The time transect: the Anatolian shift goes toward the Caucasus, never toward the steppe

The evolution of the Anatolian genome can be tracked through time, from Neolithic Barcin to the Hittite period, by applying the same distal model at each stage. The result is unambiguous: the eastern component (Iran plus CHG) rises steadily, exactly as the contacts between Anatolia and the Caucasus during the fifth and fourth millennia would predict, while EHG stays nailed to zero throughout. The shift that turns the Neolithic Anatolian into a Bronze Age Anatolian is entirely a shift toward the Caucasian southeast.

7. The big picture: one common ancestor, two fates

8. Proximal modelling: even when Yamnaya are offered, they are refused

To close the door definitively, a proximal model can be tested that offers the Yamnaya directly as a possible source, alongside Anatolia_N and the Caucasus Neolithic (the southern end of the CLV cline). If any steppe ancestry existed in the Hittites, this is where it would show up most strongly. The result confirms everything above.

9. An honest nuance: MA2203, and the sampling question

Rigour requires two caveats, which do not overturn the conclusion but sharpen its outlines. The first concerns individual variability. Of the two Hittite-period samples, individual MA2200 returns 0 percent EHG in the distal model, but individual MA2203 returns 2.4 percent. This is the faint residual signal that an analyst like Davidski noted in suggesting a minimal but real steppe infiltration. Honestly, 2.4 percent in a single individual, in a five-source model, remains at the edge of statistical noise, and the average individual falls back to 0.1 percent. It can neither be asserted nor excluded: at most one can say that there is, at best, only a whisper of steppe in Hittite Anatolia, in no way comparable to the half-genome of the Yamnaya.

The second caveat, more important, is archaeological. The sequenced individuals come from Kalehöyük, a Hittite-period site, but they are not identified royal Nesite burials. There is no guarantee they belonged to the Hittite-speaking elite rather than to the Hattic substrate, the non-Indo-European population that occupied central Anatolia before and during Hittite rule. If Anatolian Indo-European was carried by a demographically minority elite that imposed its language on a majority substrate (a frequent pattern in the history of linguistic expansions), then the steppe or Caucasian ancestry specific to that elite could have been strongly diluted, even undetectable, in a general-population sample. The weak CLV input of about 10 percent that Lazaridis 2025 detects in central Anatolians is precisely consistent with this diluted-elite scenario.

10. What the Hittites teach us about Indo-European origins

The Anatolian paradox, long brandished as the decisive argument against the steppe theory, turned into its subtlest support. The 2025 resolution does not say that Indo-European comes from Anatolia (Renfrew's hypothesis remains refuted: Anatolia receives CLV ancestry, it does not export it toward the steppe). It says that the Indo-Europeans and the Anatolians share a common ancestor, the populations of the Caucasus-Lower Volga cline, and that this ancestor sat at the hinge between the Caucasus and the lower Volga, neither quite in the steppe nor quite in the south.

This picture fits remarkably with the Indo-Hittite hypothesis formulated by Sturtevant a century ago on purely linguistic grounds. If Anatolian preserves archaisms (the laryngeals) that all the rest of the family has lost, it is because it split off first, before the development of the innovations common to the rest of Indo-European. The genetics delivers exactly the same topology: the Anatolian branch detaches from the CLV source before the steppe branch, defined by the acquisition of EHG, is constituted. The earliest linguistic split corresponds to the earliest genetic split. The absence of a steppe signal in the Hittites is therefore not a defect in the model: it is the imprint, in the DNA, of their position as the very first branch of the Indo-European tree.

11. The Hittite Global25 coordinate

For readers who wish to reproduce the analysis in Vahaduo or any other Global25 tool, here is the scaled coordinate of the average individual of the Hittite period used throughout this article.

12. Myth versus reality

13. Conclusion

The Hittites were long the stone in the shoe of the steppe theory: how could an Indo-European language be spoken by a people devoid of the genetic marker of Indo-European? The answer, delivered by palaeogenetics between 2018 and 2025, has an elegance that surpasses the initial embarrassment. The Hittites carry no steppe ancestry because they never had any: they descend, like the Yamnaya, from the Caucasus-Lower Volga cline, but they detached from it via the south, toward Anatolia, before the northern branch mixed with the Eastern hunter-gatherers to forge the steppe signature. On Global25 they stay glued to their Anatolian neighbours and infinitely far from the steppe; their genome is one-third eastern coming from the Caucasus and two-thirds Anatolian Neolithic, without the slightest measurable steppe input. This genetic topology matches exactly the Indo-Hittite hypothesis that linguists had deduced, a century earlier, from the laryngeals alone. Anatolian is the first language to have detached from the Indo-European tree, and the Hittites are its living genetic proof: an authentic Indo-European people, but from before the steppe.

References

1. de Barros Damgaard, P., Martiniano, R., Kamm, J., et al. (2018). The first horse herders and the impact of early Bronze Age steppe expansions into Asia. Science, 360(6396), eaar7711. DOI: 10.1126/science.aar7711 Bronze Age Anatolia
2. Lazaridis, I., Alpaslan-Roodenberg, S., Acar, A., et al. (2022). The genetic history of the Southern Arc: A bridge between West Asia and Europe. Science, 377(6609), eabm4247. DOI: 10.1126/science.abm4247 Southern Arc
3. Lazaridis, I., Patterson, N., Anthony, D., et al. (2025). The genetic origin of the Indo-Europeans. Nature, 639, 132 to 142. DOI: 10.1038/s41586-024-08531-5 CLV cline
4. Nikitin, A. G., Lazaridis, I., Patterson, N., et al. (2025). A genomic history of the North Pontic Region from the Neolithic to the Bronze Age. Nature (companion paper). North Pontic
5. Kroonen, G., Barjamovic, G., Peyrot, M. (2018). Linguistic supplement to Damgaard et al. 2018: Early Indo-European languages, Anatolian, Tocharian and Indo-Iranian. Zenodo. DOI: 10.5281/zenodo.1240524 Linguistics
6. Wang, C.-C., Reinhold, S., Kalmykov, A., et al. (2019). Ancient human genome-wide data from a 3000-year interval in the Caucasus corresponds with eco-geographic regions. Nature Communications, 10, 590. DOI: 10.1038/s41467-018-08220-8 Caucasus
7. Davidski (Eurogenes). Global25 scaled coordinates and the Vahaduo Global25 tools. NNLS modelling performed for this article in Python (scipy) on the scaled coordinates. G25 method