A small clan, perhaps two thousand people, living among the Eneolithic farmers and steppe nomads of the North Caucasus and the Volga-Don basin sometime around 4400 BCE. This is the newest candidate for the ancestors of the Proto-Indo-European speech community, published in Nature in 2025 and immediately garlanded as the Harvard lab’s “Steppe Forward” - the paper that finally, supposedly, resolves the riddle of where Indo-European began. Four hundred ancient genomes, careful clustering, a genuine technical achievement. And at the end of it, a population is named. A language is not.
This is worth sitting with, because the gap between the two is the entire argument. A population geneticist can tell you with increasing precision who mixed with whom, when, and where the ancestry components of a sampled skeleton trace back to. A population geneticist cannot tell you what language that skeleton’s owner spoke, what he called his rivers, what he sang at his fire, or what he believed happened to him after death. Ancestry is inherited in blood. Language is inherited in a community of speech, and communities of speech can leap, cross, be adopted, be abandoned, be imposed and resisted, entirely apart from what DNA is doing underneath them. The Volga-Don paper is a genuine contribution to population history. It is being sold, in the press coverage if not always in the paper itself, as a solution to a linguistic and civilizational question it cannot by its nature answer. And the moment we notice that gap, an older account - one that has been sitting in plain sight for three millennia, telling us in its own words who its speakers were, where they lived, and which direction they moved - deserves to be heard again as a primary source rather than a folk tale to be explained away.
That account is the Rgveda, and the civilization it describes is the Sindhu-Sarasvati.
The river that should not exist
Start with the plainest falsification available, because it rarely gets the weight it deserves. The Rgveda names Sarasvati as naditama, the mightiest of rivers, greater than the Sindhu itself, a river so central to the world its poets inhabited that entire mandalas orient themselves around her banks. This is a geographic anchor as precise as any river-name in any ancient text anywhere, load-bearing rather than incidental. And the river dried. Hydrological and satellite work going back decades, and now sharpened by the Archaeological Survey of India’s own 2025 excavation of a paleochannel twenty-three metres under Bahaj village in Rajasthan’s Deeg district, has traced her course from the Himalayan foothills through the Ghaggar bed down to the Rann of Kutch, and dated her final desiccation to somewhere around 1900 BCE. The Bahaj site itself gives five continuous occupation layers - Harappan-successor, Mahabharata-period, Mauryan, Kushan, Gupta - laid one on the other without the gap a foreign replacement event would leave behind.
A river described as mighty and alive cannot have been described by people composing after it died. This single fact, on its own, retires the 1500 BCE composition date that the Aryan Invasion Theory requires, because no civilization sings hymns to a river it has never seen flow. The theory’s proponents know this, which is why so much energy over the last forty years has gone into either denying the Sarasvati identification altogether or insisting the Ghaggar-Hakra was always a modest seasonal stream unworthy of Vedic praise - a claim that gets harder to sustain with every new paleochannel the ASI digs up. The river was not modest. The theory that requires her to have been modest is the one under strain, not the text.
A civilization, not two
Once the Sarasvati is restored to her actual scale, the entire architecture that the invasion model depends on - a nomadic Aryan people arriving from outside to conquer or absorb a separate, prior, non-Vedic Harappan urban world - loses its load-bearing wall. There are not two civilizations here needing to be stitched together by a migration event. There is one continuous civilization, urban and sophisticated, whose textual self-description and whose excavated cities sit on the same rivers at the same time. Michel Danino’s reconstruction of the Sarasvati’s hydrology and its correlation with the settlement density of the so-called Harappan zone makes this continuity a matter of measurement, not sentiment: the greatest concentration of sites anywhere in the Sindhu-Sarasvati world sits precisely along the dried bed of the river the Veda calls mightiest. Call this civilization by the name that respects both halves of the evidence - Sindhu-Sarasvati, not Indus Valley, not Harappan - because both names smuggle in a decision about what the site-builders were not, before we’ve even asked what they were.
The markers that Aryan Invasion Theory once treated as decisive evidence of a late foreign arrival - the horse, the spoked-wheel chariot - have both turned up in Indian contexts that predate the theory’s own timeline, forcing a slow retreat from “no horse before 1500 BCE” toward vaguer and vaguer hedges. Genetics has replaced horse bones as the newest supposed clincher, and it will not hold the weight either, for the same reason it never could: steppe ancestry components detected in modern and ancient South Asian genomes tell you that gene flow happened. They do not tell you its direction of civilizational causation, its social meaning, or which population was doing the naming and which was being named. A trickle of ancestry moving into a region over centuries is compatible with dozens of historical scenarios, only one of which is invasion. It is equally compatible with an outward-radiating population whose furthest frontiers mixed back into the homeland over the following millennia - which is, as it happens, exactly the direction the linguistic evidence favors.
The Veda’s own itinerary
Shrikant Talageri did the work that the genetics debate keeps skipping past: he read the Rgveda’s ten books not as a flat, undifferentiated block but as a compositional sequence, using the internal genealogical and hymn-cross-referencing evidence to establish a relative chronology - Books 6, 3, and 7 earliest, then 4 and 2, then the family books give way to 5 and finally the late compilation of 8 through 10. What the rivers named in each layer show is not ambiguous. The oldest books know eastern rivers, the Ganga-Yamuna region. Western rivers - the Sindhu’s tributaries, the rivers of what is now Afghanistan and beyond - enter the text only in the later books. A people composing their oldest poetry about eastern waters and their newest poetry about western ones is a people whose center of gravity is moving west, not a people who arrived from the west and are only belatedly noticing the rivers behind them. Add to this the Mitanni kingdom of northern Syria around 1500 BCE, whose kings bore Sanskrit names, whose treaties invoke Mitra, Varuna, Indra, and the Nasatyas by name, whose horse-training manual uses Sanskrit numerals - a diaspora community carrying a fully formed Vedic religious vocabulary a thousand miles west of the Sindhu at exactly the date Aryan Invasion Theory says Vedic religion was still being assembled inside India. That timeline does not work in the invasion direction. It works, cleanly, in the outward one.
Subhash Kak’s and David Frawley’s work on the astronomical content embedded in the Vedic texts - equinox and solstice markers, nakshatra sequences that only align with the sky at specific epochs - pushes the composition of the earliest material back toward 3000 BCE or beyond, an order of magnitude earlier than the colonial-era timeline still taught as default. The dating is not the kind of thing that will ever be settled with the finality a chemistry equation has. But the direction of the error is consistent across every independent line of evidence - hydrology, settlement archaeology, internal textual chronology, comparative religion, astronomy - and it all points the same way: older, more continuous, and outward-moving, not younger, ruptured, and inward-arriving.
Why the invasion needed inventing
None of this evidence is new, and none of it is particularly obscure. Koenraad Elst’s contribution has been to explain why it keeps getting rediscovered and then quietly re-buried: the Aryan Invasion Theory was not derived from Indian evidence in the first place. It was assembled in the nineteenth century inside a frame that needed “Aryan” to be a race before philology ever made it a language family, in the service of a colonial administration that found it convenient to explain India’s civilizational achievements as the residue of a conquering fair-skinned elite rather than as the native production of the people the British were governing. That the framework survived decolonization at all is itself a story worth telling honestly: it was inherited almost intact by a generation of Indian historians trained in Marxist historiography, for whom a foreign-origin Vedic religion served a live political purpose - it let the tradition itself be treated as an import, an ideology laid over an indigenous population rather than that population’s own self-articulation, which made caste, ritual, and religious authority all analyzable as instruments of a conqueror class rather than as a civilization’s internal development. This is a description of incentives, openly discussed in the historiography of the debate, that explains why “most scholars believe” has functioned for decades as a substitute for evidence rather than a summary of it.
What the clan on the Volga actually tells us
Return, now, to the population geneticists and their small clan on the Volga-Don. Their finding may well be correct on its own terms - that a Copper Age population in that region contributed significantly to the ancestry later found across the Yamnaya horizon and into Anatolia among Hittite speakers. What it cannot do, no matter how the press release is worded, is tell us that this population’s speech, rather than some other population’s speech somewhere else entirely, is the one that became Sanskrit, Avestan, Greek, Latin, and the rest. Language family reconstruction and ancestry-component reconstruction are different instruments measuring different things, and the field has spent two decades quietly substituting the second for the first because ancient DNA produces numbers that feel more solid than comparative philology. Numbers are not more solid. They are solid about a different question.
Set beside that ghost clan - known only through statistical clustering of skeletal fragments, its language, laws, and self-understanding entirely unrecoverable - stands a civilization that left its own testimony in its own words: rivers named, tribes named, kings named, a religious cosmos fully described, sacrificial and domestic life recorded in enough granular technical detail that we know what its people ate, how they married, what they feared. The Rgveda is the single richest first-person civilizational document that survives anywhere from this depth of time, describing itself in the region it still names for its own descendants, on a river whose bed the Archaeological Survey of India dug up again last year almost exactly where the hymns said it would be - no translation into the terms of an ancestry graph required before it can be believed. Given a choice between a language reconstructed backward from a modern comparative method and then handed to whichever ancient population’s bones happen to cluster correctly, and a civilization that names its own rivers, its own gods, and its own compositional history in a continuous unbroken tradition down to the present day, there is no honest methodological reason to prefer the ghost.
The stakes of getting the direction right
This is not an antiquarian argument about who gets credit for a verb conjugation. A civilization’s account of its own origin is a load-bearing part of how that civilization understands its own continuity, its relationship to its land, and its right to interpret its own texts without translation through a foreign framework first - never mere decoration on top of the history. Every year the invasion model persists in textbooks, Bharatiya civilization is taught to its own children as a graft rather than a native growth, and every corrective - the Sarasvati’s hydrology, Talageri’s internal chronology, the Mitanni evidence, the astronomical dating, and now even the fine print of the ancestry papers themselves - gets absorbed into the same tired holding pattern where the invasion or its migration-flavored cousin remains the default and the counter-evidence remains a footnote to be managed rather than a body of proof to be reckoned with.
Talageri has done the philological work. Danino has done the hydrology. Elst has done the historiography. Kak has done the astronomy. The Archaeological Survey of India, digging in Rajasthan in 2025, has done the spadework. What is left is the argument about direction, and on every axis where the evidence can actually speak for itself rather than being read through the lens the theory demands, it points the same way: out of the Sindhu-Sarasvati, not into it. The small clan on the Volga may turn out to be exactly right about who their ancestors were. They will never be able to tell us what those ancestors called their river. The Rgveda already has.