Many waves, not one: ancient DNA and the structure of Indian admixture

The Moorjani LD-decay method that opened this series gives one number per population: roughly when, on average, the mixing of ANI and ASI happened ^[S4]. It's an enormously useful number, and Story 1 walked through how it's measured. But there's a catch in the math the paper itself flags. The method assumes a single pulse of mixing — one moment when two populations meet, after which their genomes recombine over generations. If the actual history involved multiple waves of mixing layered on top of each other, the inferred date is a smear of the underlying events, not a single sharp answer.

Whether the mixing happened in pulses or as a continuous process matters. It's the difference between two populations meeting once and then settling, versus an ongoing demographic flux that lasted centuries.

Ancient DNA can answer this directly. By sampling individuals from before, during, and after the proposed mixing window, you can watch the ancestry composition of South Asian populations change over time. You don't need to infer the structure of mixing — you can read it.

What you'd see if mixing was one event

If two distinct populations met once and then stopped exchanging genes, the aDNA record would show a sharp transition: pre-mixing samples carry only one ancestry, post-mixing samples carry both, with the proportions stable across subsequent centuries.

Recombination over generations would shorten the alternating ancestry chunks in each individual's genome — that's what Moorjani's LD-decay method picks up. But the proportions of the two ancestries would stabilize quickly and stay stable across post-mixing samples.

What you'd see if mixing was many waves

If multiple groups arrived at different times, each carrying somewhat different ancestry profiles, the aDNA record would show progressive change: ancestry composition shifting over centuries as new arrivals layered new ancestry into the existing population.

Some samples would have only the earliest waves. Others, sampled later, would carry signatures of all the waves that had reached their region by then. Ancestry proportions would drift over time, not stabilize abruptly.

What we actually see

The Narasimhan 2019 paper, which sequenced 523 ancient individuals across Central and South Asia spanning 9,000 years, provides the strongest test of these models ^[S1]. The pattern is unambiguous.

Pre-Indus samples — individuals from before about 5,000 years ago, mostly from sites in the Iranian plateau and BMAC — carry profiles that look like Iranian Neolithic ancestry mixed with deeper indigenous components, with substantial group-to-group variation.

Mature Harappan-era samples, including the Rakhigarhi individual sequenced by Shinde 2019 ^[S2], carry a fairly stable profile: Iranian-Neolithic-related ancestry plus deep South Asian indigenous ancestry, no Steppe component. This is the pre-Steppe baseline, anchored to a specific time and place.

Bronze Age samples from northwestern South Asia, dated 3,500 to 3,000 BP, suddenly show Steppe ancestry layered on top — but in widely varying proportions. This isn't a single point in time when "Steppe ancestry arrived." Different individuals from sites in the Swat Valley, sampled across centuries, show different proportions of Steppe, Iranian-related, and indigenous components.

Iron Age and historical-era samples continue the pattern: ongoing admixture, gradual homogenization within groups, with Steppe ancestry spreading southward and eastward over the centuries.

What this tells us

The Moorjani LD-decay date — roughly 1,900 to 4,200 years ago — is real, but it's an average across multiple distinct mixing events, not a date for a single one. The Iranian-Neolithic ancestry and AASI components started mixing around 5,000 to 6,000 years ago in the broader Indus region. The Steppe component layered in starting around 3,500 BP. Different waves continued for centuries afterwards in different regions.

The Damgaard 2018 paper ^[S3], looking at the source side from Central Asia, reinforces this picture. Steppe Bronze Age populations were themselves heterogeneous — different subpopulations with different ancestry profiles moving south at different times. There wasn't one Steppe people who arrived in India; there were several related but distinct groups, arriving over generations.

This is one reason the average mixing date gets fuzzy when you push on it. Moorjani's analysis is rigorous, but its single-pulse assumption is a simplification. The reality the aDNA reveals is more textured — and arguably more interesting.

What aDNA can't yet resolve

The pulse-versus-continuous question isn't fully settled even now. Ancient DNA from the actual mixing window — roughly 4,000 to 2,000 years ago in South Asia — is still relatively sparse compared to what we have for Europe and Central Asia. Many of the inferences about pace come from samples slightly before or after the window, with the middle filled in by interpolation.

What's needed: more samples from the 4,000 to 2,000 BP window across the subcontinent, especially from the Hindi belt, Bengal, and the Deccan, where the genetic transition was substantial but where aDNA preservation is harder. Several research groups are actively working on this. The picture is filling in, and Story 1's smear of dates is becoming, as it should, a sequence of waves.