Ancient DNA | South Asia | The Aryan debate

Did the Aryans invade India?

For a century and a half, one question has divided historians, politicians and the internet: did Indo-European speakers from the steppe conquer India, or did the Vedic culture grow from the Indus civilisation itself? Ancient DNA now gives an answer, and it suits neither of the old stories. We tested it in Global25.

No steppe ancestryin the only genome from an Indus city, a woman buried at Rakhigarhi some 4,500 years ago, nor in the people who shared her ancestry
1900 to 1500 BCthe window in which steppe ancestry mixed into the people of the northwest, after the great Indus cities had declined
Up to 30%of the ancestry of some South Asian groups today comes from the steppe, highest in the northwest and in some priestly communities

Hero image: AI generated illustration of herders of steppe descent, with horses, cattle and a two-wheeled chariot, arriving in a green valley of the northwest of South Asia in the second millennium BC, watched by local farmers outside a mudbrick village. Clothing, animals and the scene are approximations, and it is not a reconstruction of a documented event or of the individuals whose genomes are discussed here.

Few questions in history are as charged as this one. In one version, fair-skinned "Aryan" warriors swept down from the north in chariots, destroyed the cities of the Indus and imposed their language, their gods and their caste system on the people they conquered. In the other, there was no invasion at all: the Aryans of the Vedas were the heirs of the Indus civilisation, and the Indo-European languages may even have spread out of India rather than into it. Both stories have been told with passion, and both have served politics more than science.

Since 2019, ancient DNA has entered the debate. The first genome from an Indus city, more than 500 ancient genomes from South and Central Asia, and now thousands of present-day Indian genomes tell a story that is less dramatic than an invasion and more surprising than continuity. We read the studies and ran the data ourselves in Global25.

From the Indus cities to the VedasKey dates in the debate over the Aryans, from the Bronze Age to ancient DNA3050 to 2065 BCThe Indus Periphery genomesPeople with an Indus-like ancestry buried far from home,at Shahr-i-Sokhta in Iran and Gonur in Turkmenistan.2600 to 1900 BCThe Indus cities at their heightHarappa, Mohenjo-daro and Rakhigarhi: planned brick cities,drains, seals and a script no one has deciphered.about 2100 to 1800 BCSintashta and the chariotHerders of the southern Urals build fortified settlementsand bury spoked-wheel chariots in their graves.1900 to 1500 BCSteppe ancestry arrivesThe date of mixing estimated from the genomes of Swat;the Indus cities have already declined.about 1500 to 1200 BCThe RigvedaThe oldest Sanskrit hymns, by speakers who calledthemselves arya. Linguists' usual estimate.about 1380 BCIndra in a Syrian treatyA treaty of the Mitanni kingdom invokes Mitra, Varuna,Indra and the Nasatyas, gods of the Rigveda.1947 and 1964A massacre that never wasWheeler writes that "Indra stands accused"; Dales showsthe Mohenjo-daro skeletons were not one massacre.2019 and 2025The genomesNarasimhan, Shinde and colleagues publish the firstIndus genome and 523 ancient genomes; 2025 brings more.Dates are approximate. The Rigveda and Sintashta dates are the usual scholarly estimates.
Key dates in the debate over the Aryans, from the Indus cities to the ancient genomes. Dates are approximate; those of the Rigveda and of Sintashta are the usual scholarly estimates.

Two myths about one word

In the Rigveda, the oldest text in Sanskrit, arya is a word that the authors of the hymns used for themselves, a cultural and religious label, roughly "the noble ones" or "our people". From the late eighteenth century, European scholars noticed that Sanskrit, Persian, Greek, Latin and the Germanic languages share a common ancestor, and in the nineteenth some of them turned the word into the name of a race. The twentieth century showed where that idea could lead. Today linguists speak of Indo-Aryan languages, a branch of the Indo-European family, and geneticists speak of ancestry, not race.

The invasion story also had its archaeology. In 1947 Mortimer Wheeler, then head of the Archaeological Survey of India, pointed to skeletons found by earlier excavators in the upper levels of Mohenjo-daro as the victims of a final massacre, and wrote that "on circumstantial evidence, Indra stands accused", Indra being the Vedic war god who smashes forts in the hymns. In 1964 the American archaeologist George Dales re-examined the evidence and showed that the skeletons came from different levels and different periods, with no sign of a battle, no weapons and no burned citadel. The massacre had never happened. Most archaeologists now see the decline of the Indus cities after about 1900 BC as a slow process, not a conquest.

That left the field open to the opposite thesis, often called Indigenous Aryanism or Out of India: the Indus people were the Vedic Aryans, and nothing important came from outside. Genetics can now test what both sides claim about people, though not, as we will see, everything they claim about languages.

The Harappans had no steppe ancestry

Ancient DNA barely survives in the heat and humidity of South Asia. At Rakhigarhi, in Haryana, the largest known Indus city in India, the team of Vasant Shinde and David Reich screened 61 skeletons; only one, a woman who lived some 4,500 years ago, gave enough authentic DNA. Her genome, published in 2019, fits as a mixture of two ancestries: a large Iranian-related component, from a lineage that split from the ancestors of the farmers and hunters of the Iranian plateau before farming began, and an indigenous South Asian hunter-gatherer component, related to the Andamanese, that geneticists call AASI. What she lacks is just as important: no detectable ancestry from the steppe, and none from the farmers of western Iran or Anatolia. Farming in South Asia, the authors concluded, arose largely among local people, not through a mass migration from the west.

The woman of Rakhigarhi was not alone. Her profile matches 11 earlier genomes that stood out from their neighbours at two sites in contact with the Indus world, Shahr-i-Sokhta in eastern Iran and Gonur in Turkmenistan. Narasimhan, Patterson, Reich and colleagues call them the Indus Periphery and see them as probable migrants from the Indus world. Eight of these genomes, dated between 3050 and 2065 BC, are in Global25, and we tested them. The Rakhigarhi genome itself is not.

The result agrees with the published one. We modelled each of the eight as Iranian-related plus Andamanese-related ancestry, with a steppe source available. With four of the six Iranian-related sources we tried, the steppe share is exactly zero in all eight; with the other two it is 0 to 8%, and adding steppe never improves the fit by more than three units. One honest caveat: a single ancient genome in Global25 is too noisy to prove a strict zero, and in some of our models a steppe share of 10 to 15% could partly go unseen. The published analysis, with formal statistics on the full data, found no detectable steppe ancestry; that is the headline, and our test is consistent with it.

Then the herders came

At the end of the third millennium BC, a new world was forming on the steppe east of the Urals. The Sintashta culture built fortified settlements and buried spoked-wheel chariots with its dead, around 2100 to 1800 BC, and related herding groups spread across what is now Kazakhstan. Their ancestry came from the Yamnaya, the herders of the Pontic-Caspian steppe, mixed with about a third from European farmers, picked up in eastern Europe on the way. In 2025, Iosif Lazaridis and colleagues traced the deeper source of the Yamnaya to a population of the steppe between the Caucasus and the lower Volga, around 4500 to 3500 BC, which they link to all the peoples who speak Indo-European languages today.

The first traces of this ancestry in the south appear in Central Asia. The main population of the great Bronze Age towns of Turkmenistan and Uzbekistan, the Bactria Margiana Archaeological Complex, carried none of it; but by about 2000 BC outlier individuals with steppe ancestry turn up at its sites. In our data, a man buried at Gonur around 2100 BC stands far above his neighbours. Then the movement continued south.

The decisive evidence comes from the Swat valley, in northern Pakistan, where archaeologists, many of them from the Italian mission in Pakistan, excavated cemeteries of the late second and early first millennium BC. The people buried there carry about a fifth of steppe ancestry, and the way the two ancestries are interleaved along their chromosomes dates the mixing to between about 1900 and 1500 BC. To show the change, we measured every ancient genome with one and the same model.

Before and afterSteppe-related ancestry in individual ancient genomes, measured with the same model for everyone (our analysis)0%10%20%30%40%50%3000 BC2000 BC1000 BCAD 1AD 1000mixing1900 to1500 BCmodel background, about 4%Indus Periphery, 8 genomes0 to 8%Gonur outlier, about 2100 BCIron Age Swat, 82 genomesaverage 25%Historic Swat, 20 genomesaverage 24%Medieval, 4Indus PeripheryGonur outlierSwat, Iron AgeSwat, historicSwat, medievalModel: Iranian Neolithic + Paniya + Central Steppe MLBA. It gives about 4% even to Indus Periphery genomes with nothing added;a 10% steppe spike reads as 13%, 20% as 23%. Read the jump, not the exact values.
Steppe-related ancestry in individual ancient genomes, all measured with the same model (Iranian Neolithic, Paniya and Central Steppe MLBA sources; our analysis). The model gives about 4% even to Indus Periphery genomes with nothing added, so the Indus Periphery sits at background, while the Iron Age people of Swat average 25%. Spiking 10 or 20% of steppe ancestry into the Indus Periphery average reads as 13 and 23%.

The contrast is stark. The eight Indus Periphery genomes score 0 to 8%, which is the background of the model: it gives about 4% even to their average with nothing added. The 82 genomes of Iron Age Swat, buried between about 1270 and 760 BC, score 11 to 35%, 25% on average. A more precise model, with the Indus Periphery as the local source, gives 26 to 29% for the three largest cemeteries, Loebanr, Katelai and Udegram, a little above the published figure of about 20%: Global25 tends to read slightly higher than the formal statistics of the original study.

Did the steppe ancestry perhaps travel the other way, out of India and onto the steppe? The genomes rule it out. The steppe ancestry of South Asia has the same profile as that of Bronze Age eastern Europe, including the European farmer share picked up in Europe. And the steppe genomes themselves carry no Indus ancestry: in Global25, the Yamnaya and the people of Sintashta take 0% from either the Indus Periphery or the Andamanese source, a test that would pick up about 10% if it were there.

The genomes also say something about who moved. In the Swat Iron Age, the steppe ancestry is under-represented on the Y chromosome, which suggests it had entered those communities largely through women. In present-day South Asians it is the opposite: the Y chromosome carries more steppe ancestry than the rest of the genome, the signature of a migration in which men were over-represented. The paternal lineage most closely tied to these migrants, R1a-Z93, is today among the most common in northern South Asia.

Who carries the steppe today

Steppe ancestry is now found across most of South Asia, but in very unequal amounts. We modelled present-day groups as a mixture of the Indus Periphery, the Central Asian steppe herders of the Middle and Late Bronze Age, the Paniya of Kerala as a southern pole (a tribal population with little or no steppe ancestry), and an East Asian source for the east.

How much steppe ancestry in South Asia today?Present-day groups modelled with Indus Periphery, Central Steppe MLBA, a southern Indian tribal pole and East Asian sources (our analysis)0%25%50%75%100%fitKalash, Pakistan604044Ror, Haryana584212Pashtun Yusufzai, Pakistan613835Jat Sikh, Punjab673214Kohistani, Pakistan682924Khatri, Punjab693023Bhumihar, Bihar52272218Kashmiri Pandit702624Brahmin, Uttar Pradesh58231927Brahmin, Gujarat66241013Sindhi, Pakistan742422Brahmin Iyer, Tamil Nadu61152516Bengali, Bangladesh39943919Gujarati6492721Chamar, Uttar Pradesh415628Telugu583819Reddy, Andhra Pradesh712621Kallar, Tamil Nadu544416Irula, Tamil Nadu207911Indus PeripherySteppe (Central Steppe MLBA)Southern tribal (Paniya)East Asian (Dai)Steppe share stable within 2 points with Sintashta as the steppe source, Paliyan as the southern pole, or a wider Indus Periphery average.
Present-day South Asian groups modelled as Indus Periphery, Central Steppe MLBA, a southern tribal pole (Paniya) and East Asian (Dai) sources (our analysis). The steppe share changes by 2 points at most with Sintashta as the steppe source, Paliyan as the southern pole or a wider Indus Periphery average. Published estimates (Narasimhan 2019) reach about 30%; Global25 runs a little higher.

Three patterns stand out. The first is geography: steppe ancestry is highest in the northwest, 29 to 42% among the Kalash, the Ror, the Pashtuns, the Jat Sikhs, the Kohistanis and the Khatris, and falls to a few percent in the south and east. The second is community: within the same region, groups differ. In Uttar Pradesh and Bihar, Brahmin and Bhumihar samples carry 23 to 27%, the Chamar 3%; in Tamil Nadu, Iyer Brahmins 15%, the Kallar 2%. Narasimhan and colleagues found the same thing: groups that see themselves as traditionally priestly tend to have more steppe ancestry than other groups, even after accounting for other factors, and the Brahmins have been the custodians of texts in Sanskrit, an Indo-European language. Reich called it a new argument for a steppe origin of the Indo-European languages of South Asia. The third is that nobody is mostly steppe: even in the northwest, 58% of the ancestry or more is Indus-related.

A word on the numbers. Our Global25 estimates run higher than those of Narasimhan and colleagues, who put the steppe contribution at up to about 30% of the ancestry of present-day groups, but they stay within the range of 0 to 45% found across 2,762 present-day Indians sequenced in 2025 by Elise Kerdoncuff, Laurits Skov, Priya Moorjani and colleagues. The ranking, however, is the same, and it barely moves with the choice of sources: within 2 points with Sintashta instead of the Kazakh herders, a different southern pole or a wider Indus Periphery average. The Kalash and Pashtun fits are looser (35 to 44 units), partly because small, isolated communities drift away from any mixture of sources.

These differences were frozen in place by endogamy. Moorjani and colleagues showed in 2013 that the mixing between the northern and southern ancestral populations of India took place between about 4,200 and 1,900 years ago, after which marriage within the community became the rule, even between closely related groups. The 2025 study found the genetic signature of strong founder events across India, a legacy that today raises the risk of some recessive diseases in particular communities. None of this says anything about the worth or identity of any person; it records who married whom over the last 3,000 years.

The Harappans never left

The other half of the answer matters as much. The ancestry of the Indus Periphery, and of the woman of Rakhigarhi, is the largest single source of ancestry in South Asia today. In our model it accounts for 52 to 74% of most groups, before counting the Indus-related ancestry that the southern populations themselves carry. In the words of Reich, this ties people in South Asia today directly to the Indus Valley Civilisation. After the decline of the cities, the Indus-related population mixed with newcomers carrying steppe ancestry in the northwest, and with groups rich in indigenous hunter-gatherer ancestry in the south and east. The northern and southern ancestral populations of India, long known to geneticists as ANI and ASI, were both born from that Indus-related foundation.

No living population is a close match for the Indus Periphery: the nearest in Global25, the Toda of the Nilgiri hills, sit 47 units away, followed by communities of Kerala. Four thousand years of mixing have reshaped everyone. The Iron Age people of Swat, by contrast, have close living look-alikes.

Who looks most like the Iron Age people of Swat?Global25 distance from the Swat Iron Age average (Loebanr, Katelai, Udegram) to present-day populations (our analysis)08162432Kohistani, Pakistan18Khatri, Punjab18Arain, Punjab20Gujar, Punjab23Awan, Punjab23Gujar, Swat24Arora, Punjab25Rajput, Potohar25Sindhi28Distances x 1000. Under 20 is close for population averages: the Iron Age people of Swat have living look-alikes in the northwest.
Global25 distance from the Iron Age Swat average (Loebanr, Katelai, Udegram) to present-day populations (our analysis). The population that formed in the northwest after the steppe migration is still recognisable there today.

The Kohistanis of northern Pakistan and the Khatris of Punjab sit about 18 units from the Swat Iron Age average, closer than many modern populations are to each other. The population that formed in the northwest after the steppe migration, around 3,000 years ago, is still recognisable there today.

So, did the Aryans invade India?

If "invasion" means an army that conquered the Indus cities and put their people to the sword, the answer is no. There is no massacre at Mohenjo-daro, the cities had declined before the steppe ancestry arrived, and the genomes show mixing over centuries, not replacement. Even today, steppe ancestry is a minority almost everywhere in South Asia.

If the question is whether people from the steppe moved into South Asia in the second millennium BC, the answer is yes. Their ancestry is absent before 2000 BC and present in every Iron Age genome from Swat; it arrived in the same window in which linguists place the first Indo-Aryan speakers, whose gods appear in a treaty of the Mitanni kingdom of northern Syria around 1380 BC and whose hymns became the Rigveda. Its link with the priestly communities that preserved Sanskrit is one of the strongest clues that the Indo-Aryan languages came with these migrants.

Both camps keep something. Those who insist on continuity are right that the Harappans did not vanish: they are the main ancestors of South Asians, and the woman of Rakhigarhi is closer to the roots of a Tamil or a Punjabi than any herder from the steppe. Those who defend a migration are right that it happened, and that it helps explain the spread of Indo-Aryan languages. What the genomes rule out is the two extreme versions: the conquest that wiped out the Indus people, and an India untouched by any movement from outside.

Some questions remain open, and honesty requires saying so. A single genome from an Indus city cannot describe a whole civilisation. There are almost no ancient genomes from the Ganges plain in the second and first millennia BC, when the Vedic culture took shape. Genes are not languages: DNA cannot tell us what the Harappans spoke, and the Indus script is still undeciphered. Narasimhan and colleagues suggest that the Indus people may have spoken an early form of the Dravidian languages now spoken in the south, but they present it as one possible scenario, not a finding.

Key points

  • The only genome from an Indus city, a woman from Rakhigarhi, and the 11 Indus Periphery genomes that match her have no detectable steppe ancestry; their ancestry is Iranian-related plus indigenous South Asian.
  • Steppe ancestry reached South Asia from Central Asia after the decline of the Indus cities; in the Swat valley it mixed in between about 1900 and 1500 BC.
  • On a single yardstick in Global25, Indus Periphery genomes score no more than the model background, while 82 Iron Age genomes from Swat average 25% steppe ancestry.
  • The steppe ancestry came from the steppe, not the other way round: the steppe genomes carry no Indus ancestry, and the South Asian steppe ancestry matches Bronze Age eastern Europe.
  • Today steppe ancestry is highest in the northwest and in some priestly communities, and falls to a few percent in the south; published estimates reach about 30%, ours run a little higher.
  • Indus-related ancestry remains the largest source of ancestry in South Asia: there was a migration and long mixing, not an invasion that replaced the Harappans.

Global25 coordinates used in this article

The model sources, the Iron Age Swat average and a few present-day populations. They can be pasted directly into the Vahaduo tool.

Indus_Periphery_ShahrISokhta_avg,0.044391,-0.035137,-0.167894,0.093218,-0.104881,0.056559,0.004418,0.004061,0.004295,0.005430,-0.004547,0.006204,-0.004222,-0.001321,0.009392,0.011986,-0.000939,0.001064,0.002715,-0.011731,0.001897,-0.011920,0.001183,-0.010339,0.001485
Central_Steppe_MLBA_Kazakhstan_avg,0.124693,0.101488,0.056644,0.082180,0.006468,0.030018,0.003103,0.002929,-0.016629,-0.032143,-0.002230,-0.001868,-0.002559,-0.022848,0.022925,0.012346,-0.004961,0.000038,-0.000593,0.001507,-0.005537,0.003189,0.000794,0.006858,-0.002922
Russia_Sintashta_MLBA,0.125888,0.116617,0.057448,0.078640,0.011335,0.029098,0.005820,0.004361,-0.017425,-0.028271,-0.002306,0.001219,-0.002190,-0.021230,0.022878,0.012450,-0.005011,0.000338,-0.000352,-0.000338,-0.005889,0.001822,0.002674,0.006929,-0.003660
Iran_GanjDareh_Neolithic,0.043025,0.067431,-0.153488,0.005556,-0.123962,0.024375,0.015464,0.000277,-0.081605,-0.054270,-0.003248,-0.001619,0.005382,-0.007845,0.031949,0.056775,-0.005815,0.007576,0.014405,-0.032741,0.007661,-0.030048,-0.010920,-0.038777,0.022944
Paniya_southern_pole,0.005193,-0.175370,-0.188678,0.141959,-0.046740,0.052710,-0.006918,0.017177,0.091422,0.058543,0.005257,0.000290,-0.003884,0.015285,-0.024218,-0.029940,0.012582,-0.000230,-0.003307,0.030820,0.007214,0.014274,-0.010276,0.007494,-0.007918
Swat_IronAge_avg,0.068418,0.000148,-0.110724,0.069091,-0.080771,0.042924,0.003345,0.001778,-0.018270,-0.015016,-0.004902,0.001293,-0.001846,-0.009214,0.012892,0.011817,-0.001196,0.000099,0.000192,-0.009818,-0.001738,-0.006230,0.001739,-0.005276,0.003354
Kalash_today,0.083883,0.024991,-0.084032,0.066595,-0.071679,0.040306,0.003116,0.002017,-0.030927,-0.025156,-0.005592,-0.000495,-0.002392,-0.010902,0.016611,0.008762,-0.013821,0.002104,0.000672,-0.012832,-0.003922,-0.005392,0.002861,-0.003426,0.003353
Ror_today,0.084001,0.018076,-0.074494,0.084992,-0.060483,0.045459,0.001770,0.004354,-0.003231,-0.011700,-0.005630,0.000060,-0.000585,-0.010808,0.011998,0.010139,-0.004607,0.000329,0.000344,-0.007862,-0.003743,-0.002671,0.002481,0.001960,0.000575
Kohistani_today,0.073757,-0.006499,-0.102464,0.068961,-0.074219,0.043702,0.003517,0.003138,-0.009626,-0.010940,-0.006809,0.000410,-0.001328,-0.007523,0.010876,0.010576,-0.001878,0.000473,0.000348,-0.009000,-0.003972,-0.003825,0.002490,-0.002161,0.003848
Brahmin_Uttar_Pradesh_today,0.067440,-0.028435,-0.123601,0.108407,-0.072283,0.062227,0.000940,0.008423,0.025182,0.009522,-0.006049,0.001442,-0.003531,-0.002770,0.008873,0.002039,-0.003146,0.001964,-0.003174,-0.008457,-0.001404,0.000031,0.003143,0.004413,-0.006302
Telugu_today,0.042519,-0.080325,-0.170787,0.112707,-0.077533,0.064172,-0.000766,0.013220,0.041749,0.027147,-0.003981,0.001557,-0.000484,0.003272,-0.006375,-0.004281,0.004067,0.000282,-0.000097,0.001795,0.003196,-0.000666,-0.000330,0.000820,-0.003048
Irula_today,0.015935,-0.147760,-0.185129,0.128861,-0.054595,0.056894,-0.003983,0.015184,0.076195,0.050242,-0.000292,0.000809,-0.000282,0.009221,-0.017684,-0.017608,0.010463,-0.000728,-0.001358,0.020441,0.005883,0.007091,-0.005448,0.004211,-0.006688

References

  1. Ancient DNA Narasimhan VM, Patterson N, Moorjani P, Rohland N, Bernardos R, Mallick S, Lazaridis I, et al., Reich D (2019). The formation of human populations in South and Central Asia. Science 365(6457): eaat7487. doi.org/10.1126/science.aat7487
  2. Ancient DNA Shinde V, Narasimhan VM, Rohland N, Mallick S, et al., Reich D (2019). An ancient Harappan genome lacks ancestry from steppe pastoralists or Iranian farmers. Cell 179(3): 729 to 735.e10. doi.org/10.1016/j.cell.2019.08.048
  3. Ancient DNA Lazaridis I, Patterson N, Anthony D, Vyazov L, et al., Reich D (2025). The genetic origin of the Indo-Europeans. Nature 639: 132 to 142. doi.org/10.1038/s41586-024-08531-5
  4. Genomics Moorjani P, Thangaraj K, Patterson N, Lipson M, Loh PR, Govindaraj P, Berger B, Reich D, Singh L (2013). Genetic evidence for recent population mixture in India. American Journal of Human Genetics 93(3): 422 to 438. doi.org/10.1016/j.ajhg.2013.07.006
  5. Genomics Kerdoncuff E, Skov L, Patterson N, et al., Moorjani P (2025). 50,000 years of evolutionary history of India: impact on health and disease variation. Cell 188(13): 3389 to 3404.e6. doi.org/10.1016/j.cell.2025.04.027
  6. Archaeology Dales GF (1964). The mythical massacre at Mohenjo-daro. Expedition 6(3): 36 to 43. penn.museum/sites/expedition/the-mythical-massacre-at-mohenjo-daro
  7. Data Global25 coordinates by Davidski (Eurogenes), and the Moriopoulos Collection 2026 of averages. Admixture models and figures: ExploreYourDNA.