Evidence page · genetics, read carefully
Moorjani et al. 2013 dated ANI–ASI admixture in 18 groups. The numbers get quoted as endogamy dates. For most of the groups they are not endogamy dates at all — and the groups where they are not are almost exactly the upper-caste, northern and Indo-European ones. That distinction changes the meaning of the whole table, and it is in the paper's own Table 2.
Descendants of the first modern human dispersal, c. 50,000 years ago. Distantly related to Andaman Islanders. No unmixed AASI population survives anywhere — it is reconstructed, never sampled directly.
Best modelled on Sarazm_EN. Mixes with AASI to form the population Narasimhan et al. 2019 call Indus Periphery. This is the wave that makes the Indus-era population.
Sintashta/Andronovo-derived, strongly male-biased, carrying R1a-Z93. This is the wave that creates the ANI end of the cline.
Populations at both ends of the cline mix with each other, pervasively, affecting every group including isolated tribes. Then it stops.
The paper cannot date this and does not try. It is inferred from three things: the LD decay in those groups fits two exponentials better than one; they have higher ANI; and Metspalu et al. 2011 found multiple West Eurasian components in northern India. Candidates span the Indo-Greek, Śaka, Kushan, Hūṇa and later Central Asian movements — none of them tested here.
Note: the two-source ANI/ASI model is the 2013 framing. The current model (Narasimhan et al. 2019) is three-source — AASI, Iranian-farmer-related, Steppe — and ANI and ASI are themselves already mixtures. The dates below are still the best available for the mixing-to-endogamy transition.
Read naively, the table says a middle-status southern merchant caste locked its boundaries first and Brahmins locked theirs last. That reading is an artefact. A group with one old mixture and then isolation gets an old date. A group with an old mixture plus a later wave gets a younger date, because the recent wave dominates the linkage-disequilibrium signal. The younger date is an average, not a closing date.
And even for the groups where the date is real, the genetics cannot say who imposed it. A boundary can be policed from inside to preserve rank, or imposed from outside as exclusion. Chamar — a Dalit leatherworking community, oddly labelled "tribal" in the paper's own table — has an early date. That is at least as consistent with being shut out as with shutting others out. The DNA records the boundary. It is silent on the direction of force.
| Issue | What it does to the number |
|---|---|
| Error bars | Vysya is 144 ± 27 generations — roughly ± 780 years. Several groups overlap each other entirely once the error is drawn. |
| Sample size | 6 to 17 individuals per group. Six people are carrying the estimate for one row. |
| Instantaneity assumed | The method assumes mixture happened in a pulse. The paper states the date "may actually be an average of multiple dates spread out over a substantial period." |
| Old mixture is invisible | If initial admixture were older than ~10,000 years, the resulting LD would have decayed below detection. Earlier mixing would simply not appear. |
| Coverage | 18 of 73 groups dated. No Kerala, Bengal, Maharashtra or Northeast; Austroasiatic and Tibeto-Burman speakers excluded from the cline analysis outright. |
| Mixture ≠ migration | The authors state no West Eurasian group has been found sharing ancestry with ANI within the last 12,500 years, and that ANI and ASI may both have been in or near South Asia long before they mixed. |
What survives all of that: pervasive mixture within the last few thousand years, affecting every group including isolated ones, followed by a transition to endogamy. The ordering of individual groups within a few hundred years of each other does not survive.
The Arya Vysya of Andhra Pradesh and Telangana — a Telugu-speaking merchant and trading community, historically associated with the Komati name, with their own goddess, Vasavi Kanyaka Parameshwari, and their own temple network and caste associations. The paper classes them "middle caste". They are not an oppressed group, which is worth stating plainly given how the chart reads at first glance: the oldest date in the table belongs to a merchant caste, not a marginalised one.
Their genetic distinctiveness is well documented beyond this paper. A strong founder event — a population bottleneck followed by strict endogamy — has left recessive variants at elevated frequency, including a well-known butyrylcholinesterase deficiency that causes prolonged paralysis after certain surgical muscle relaxants. Endogamy is not an abstraction here; it is legible in a hospital. Nakatsuka et al. 2017 identified comparable founder events in dozens of South Asian groups.
The data does not show that. UP Brahmins have high ANI (61–63%) and reject the single-pulse model — meaning their profile reflects more West Eurasian input, later, not ancient isolation. High ANI here is a record of additional gene flow, not of preserved purity. The one northern upper-caste group with a genuinely old date, Kashmiri Pandit at 103 generations, also fails the single-pulse test.