When Did Caste Endogamy Begin? What Ancient DNA Says—and What It Doesn't

Here's a strange fact: your genome can tell a scientist, roughly, when your ancestors stopped marrying outside a particular group — even if nobody wrote that rule down. It's called a founder effect, and it works like this. Take any population and shrink it to a small number of breeding adults. Then close the gates: no new members marry in from outside, generation after generation. Everyone descended from that group starts sharing unusually long, identical stretches of DNA, inherited from the same small pool of ancestors. That shared DNA is a kind of fingerprint, and geneticists can read its length to estimate how long ago the isolation began. The same signature explains elevated rates of certain inherited diseases in populations like Ashkenazi Jews and Finns.

In India, that fingerprint shows up early and often — which is why a cluster of Reich lab studies has become required reading for anyone asking when caste hardened into a genetic boundary. The honest answer, all three studies agree, is: we can date the isolation. We cannot date the rule that caused it.

What the numbers actually say

Moorjani et al. 2013, in the American Journal of Human Genetics, was the first to put a number on it. Using a method that tracks how "linkage disequilibrium" — the tendency of nearby DNA variants to travel together — decays at a predictable rate once a population stops mixing, the study estimated that most Indian groups stopped large-scale intermarriage between two ancestral lineages, which the paper labels Ancestral North Indian (ANI) and Ancestral South Indian (ASI), somewhere between 1,900 and 4,200 years ago. Most groups clustered toward the more recent end of that range. After that point, the paper found, the mixing stopped almost everywhere it looked, and groups became what the authors call "strongly endogamous."

Narasimhan et al. 2019, in Science, backed this up with a much bigger ancient-DNA dataset spanning Central and South Asia. It traced how the ANI/ASI mixture happened in the first place — Steppe pastoralist ancestry arriving and blending with earlier South Asian populations — and confirmed that widespread mixing gave way to endogamy mostly during the first millennium BCE. The authors note this timing overlaps with the period that later Sanskrit textual tradition associates with formalized varna (a fourfold social ranking system) and jati (birth-based sub-caste group) rules. They're careful, though, not to claim the genetics prove that link. Overlap in timing is not proof of cause.

Nakatsuka et al. 2021, in Cell, went a layer deeper, looking jati by jati rather than at broad ancestral categories. Some jatis carry founder signatures strong enough to produce measurable elevated risk for specific recessive diseases — comparable, the paper says, to the well-studied founder populations of Finland and Ashkenazi Jewish communities. Notably, the strength of that signature doesn't track cleanly with caste rank. Some groups historically placed low in caste hierarchies show very strong founder effects; some historically high-ranked groups show weaker ones. What predicts the signal is simpler and less loaded: how small the marrying population was, and for how long it stayed closed.

More recent work has reported different timing estimates using refined methods. Palamara et al. 2022, in PLOS Genetics, using genome-wide patterns of allele sharing, found that in a majority of Indian populations the founder events occurred within the past 600–1,000 years — substantially more recent than the earlier estimates. The method difference matters: different statistical approaches to detecting isolation can yield different date ranges, particularly when founder events weren't simultaneous across all groups.

The gap between "when" and "why"

None of this dates when the word jati was first used, when endogamy became a named social rule, or how strictly it was enforced on the ground — those are questions for textual and social historians, not geneticists. Sanskrit tradition holds that prescriptive statements on marriage and varna appear in texts like the Manusmriti, but pinning down when those texts were composed, and how widely people actually followed them, is a separate and unsettled scholarly problem.

There's also a plainer statistical limit. Linkage-disequilibrium dating estimates when mixture stopped, averaged across many generations — it's not a calendar date with a single year attached. "Around 2,000 years ago" carries a wide margin of error, and populations lumped into that same estimate could have closed off decades or centuries apart from one another.

And scholars debate the direction of causation itself: did formal caste rules cause endogamy, or did endogamy — driven by ordinary social and economic pressures — come first, with textual prescriptions following behind to describe a practice already underway? The ancient-DNA record, by itself, cannot settle that argument. It can only tell you that somewhere in the last few thousand years, mostly clustering in the centuries before roughly 100 BCE, a great many Indian populations stopped marrying across the lines that now define them — and stayed closed long enough that the closure is still legible in blood drawn today.