Notes on South Asian genetics, 2018

A “pure” Tamil Brahmin, Chandrasekhar

In the post below Zach observes that the progressive author of a piece criticizing Ajit Pai has to note she too is a Gaud Saraswat Brahmin. Of course, she is progressive and opposes casteism no doubt. But to me “caste-dropping” that you are a Brahmin is like criticizing standardized testing, while observing that you also aced your standardized test. Not that that matters. Or that it proves anything.

But I’m posting this because there was a section on the genetic purity of Gaud Saraswat Brahmin’s of Karnataka. It caught my attention because I knew it was likely false. I’ve looked at South Indian Brahmins, and they generally look like they have gene flow from other South Indians. Also, if you use something called your eyes you can see that some South Indian Brahmins do not look like pure Indo-Aryan specimens at all.

Several years ago my friend Zack collected a bunch of data via his Harappa project. We’ve come further since then, but it’s still one of the best sources of information we have. Looking at the data there, and elsewhere, we can say a few things about South Asian genetics.

  • Jatts are different. I don’t know much about Jatts personally, aside from the fact that they are quite proud of being Jatt online. But in Zack’s data, and my own analysis in the SAGP, Jatts are highly inflated for “European-like” ancestry compared to populations around them. They have the highest proportions in their part of South Asia. Even higher than Pathans.

If you asked me to say why, at this I do think Jatts do have a more recent gene flow than other groups in South Asia. If you talk to Jatts online about their history, you will know what their hypothesis for this exotic element is.

  • Brahmins are different from other South Asians, and from each other. It will surprise no one that Brahmins are often somewhat different from non-Brahmins genetically. But, they also differ from each other.

Both South Indian and Bengali Brahmins mixed with the local population. Probably on the order of ~25% of the ancestry of these two Brahmin communities can be attributed to the local substrate. But, if you correct for East Asian admixture Bengali Brahmins are actually quite similar to the Brahmins of the Gangetic plains to the west. This comports with history.

A similar fraction seems reasonable for South Indian Brahmins, though perhaps more. The key issue that I have in this case is that the “European-like” proportion of South Indian Brahmins is about half of that of North Indian Brahmins. This would indicate half dilution. The admixture was probably from the higher end of the non-Brahmin caste hierarchy.

To get a sense of what I’m talking about, here are some percentages:

Ethnicity Dataset N SIndian Baloch Caucasian NEEuro NEEuro ratio
ap-brahmin xing 25 49% 36% 3% 6% 6%
iyengar-brahmin harappa 8 47% 37% 4% 6% 6%
iyer-brahmin harappa 11 47% 37% 5% 5% 5%
brahmin-tamil-nadu metspalu 2 47% 38% 6% 5% 5%
tn-brahmin xing 14 47% 38% 6% 4% 5%
karnataka-brahmin harappa 5 46% 35% 5% 6% 7%
oriya-brahmin harappa 2 45% 35% 2% 8% 9%
kerala-brahmin harappa 1 43% 39% 4% 6% 6%
brahmin-uttar-pradesh metspalu 8 42% 36% 5% 12% 12%
bengali-brahmin harappa 8 41% 33% 5% 10% 11%
up-brahmin harappa 4 39% 37% 7% 11% 12%
bihari-brahmin harappa 1 39% 38% 5% 11% 12%
rajasthani-brahmin harappa 2 34% 36% 8% 12% 13%
punjabi-brahmin harappa 3 34% 39% 10% 11% 11%
kashmiri harappa 3 30% 37% 14% 9% 10%
pashtun harappa 7 19% 34% 20% 11% 13%
maharashtrian harappa 6 46% 35% 5% 5% 6%
tamil-nadar harappa 5 57% 31% 2% 0% 0%
gujarati-patel harappa 2 55% 41% 0% 0% 0%
bengali harappa 11 47% 27% 2% 4% 5%
ap-reddy harappa 6 54% 36% 3% 0% 0%

Don’t take the percentages as literal populations.

  • Some groups that think they are special are not so special. Kashmiri Pandits, for example, fancy themselves as somewhat better than other South Asians, often because of their West Asian or even European physical appearance. But the genetic data indicates ancestrally they’re not surprising in any way in the context of their geographic locale.
  • Geography is not that predictive. Well, it sort of is. But you see that groups like Chamars in Uttar Pradesh are similar to South Indian populations.

19 thoughts on “Notes on South Asian genetics, 2018”

  1. somewhat better than other South Asians, often because of their West Asian or even European physical appearance

    I don’t think physical appearance has much to do with this (wrong) notion as actually ruling the place. Post-Independence India has been literally run by KP families – for better or, more appropriately, worse.

    they’re not surprising in any way in the context of their geographic locale

    That itself is surprising because I would have expected higher South Indian genetic imprint esp among Brahmins – more than the non-Brahmin population.

    Kashmir’s pre-Islamic history is littered with mentions of local kings resettling large numbers of Brahmins from āryadeśa (modern day E Punjab & Haryana), vanga (Bengal), draviDa (South India) etc. Entire localities of Srinagar were populated by South-Indian brahmins, who since inter-married into local Brahmin communities.

    Perhaps the Rajataranginis were inflating resettlement figures…

    1. It is actually kind of interesting that both India and Pakistan have been “ruled” by Kashmiris. The Nehru family in India and Nawaz Sharif in Pakistan (at least he calls himself Kashmiri). Perhaps this emotive link to the place among the ruling classes makes the dispute harder to solve.

      1. Quite possible…


        At least for Nehru, the importance of Kashmir Valley was very symbolic, as is quite apparent from his writing. Many Kashmiri Pandit families have been very close to the reins of power in North India from the late Moghal period, through Sikh rule and British rule to modern day India. So, their influence on how Indian government views Kashmir is fairly profound and continues to shape Indian policy.


  2. “Brahmins are different from other South Asians, and from each other”

    Can this be said of most South Asian caste groups since they all practice/d endogamy? So Sudra in South India are different from other South Asians and from each other in India as a whole? Vaishyas are different from other South Asians and from each other in India as a whole? Ditto Kashtriya ?

  3. The ethnogenesis of a lot of Indian castes (as they manifest today) is dated to the collapse of the Mughal Empire, and the rise of local kings/chiefs/leaders in its aftermath. Nicolas Dirks and Susan Bayly have a lot of work on this.

    It is usually difficult to find lineage information before this time period.

    Caste identity based on genetic differentiation will have trouble sustaining far longer, especially with the decline in the prevalence of arranged marriages. Yes, arranged marriages still account for the majority of marital unions in India, but their share is declining, with greater involvement of women in choosing their partners.

    The shrinking of families is a major structural challenge for this institution, since an arranged marriage match typically relied on extensive kinship networks to find prospective partners.

    Another point to note is that even if different castes have distinct proportions of gene flow, the reality is that all of them are mixed, just to different degrees. The situation here is quite different from say China or Western Europe, where one particular category is overwhelmingly dominant, although this might be changing in Europe lately.

    So for example, even if Jatts, KPs or GSB’s do have higher admixtures of a certain type, and hence a greater chance of the associated expressions, there exist significant proportions of people within these communities who do not have these distinguishing expressions. So sustaining a differentiation based on facial features (Han) or skin color (white) is quite difficult. In any case, most of these caste groups are pretty small relative to the Indian population as a whole, so the benefits of associating with these identities diminishes, especially once one moves to an urban setting away from their area of origin.

    The interesting question to ask is what will take the place of caste as the principle site for political mobilization in India. I think this question is key to India’s political and social future. If caste is replaced by secular organizations based on professions, with entry regulated by educational attainment (for eg Indian Medical Association, All India Bank Officers Association, Institute of Chartered Accountants of India), who mobilize and engage with the state, we can look forward to a better governed, richer country. Most other options will lead to higher levels of conflict and insecurity.

  4. The ethnogenesis of a lot of Indian castes (as they manifest today) is dated to the collapse of the Mughal Empire, and the rise of local kings/chiefs/leaders in its aftermath. Nicolas Dirks and Susan Bayly have a lot of work on this.

    i’ve read dirks. a lot of his assertions read like bullshit now. caste (jati) endogamy is actually ~2,000 years year old. the genetics on this is clear.

    1. Even if that is true, the points about the decreasing salience of caste and genetic differences in an urbanizing India remain pertinent.

      Also, the frequency of last names like Cook, Taylor, Butler, Smith, Carpenter, Clark etc indicate that similar endogamy might have been seen in other Indo-European societies. We just never got the chance to confirm it in places like England.

      1. The Korean have like 4 surnames; surname concentration is apparently a phenomenon especially in societies that have had surnames for a very long time..

        The Chinese also have very concentrated surnames etc..

    1. Tens of thousands of samples from when ? Pre industrial or modern industrialized England ? I think it makes a big difference.

      Think about the numbers you have for Indian castes. Two or three generations of inter mixing between castes should make inferences regarding the ‘original proportions’ as you state above quite difficult.

      From your table above, consider the following sequence:
      brahmin-tamil-nadu + gujarati-patel up-brahmin + maharashtrian
      Male child + Female child
      Grandchild whose sample is taken

      Are you telling me that from the grandchild’s sample, one can infer that their lineages are a convergence of the four specific groups their grandparents belonged to ?

      I dont think that this is possible since the grandparents were themselves composed of varying distinct groups. All you should be able to find are varying levels of those original groups.

        1. Razib,

          People have a right to have a different opinion. They maybe wrong but you can put it across politely, can’t you ? Do you have to behave like an a*%h@le ? Truly pathetic of you. You are not the only one writing on this blog so why don’t you stop behaving as if you own this place and stop talking to people in a demeaning manner ?

          1. fuck you too 😉

            don’t talk about shit you don’t know on my threads.

            i won’t threaten to ban you cuz that’s unfair to the others who contribute here. but i can delete comments.

  5. According to the Vedic literature, the region of Haryana and Western UP is the Vedic homeland. It is known in later literature as Brahmavarta – a holy land. So it is interesting that the people living in this holy land – the Jatts have the highest proportion of the ‘European’ component.

    Another interesting thing is that Haryana also has the highest proportion of Pre-Harappan and Early Harappan sites such as Kunal, Bhiranna, Farmana, Girawad etc and it also has the largest Harappan site known so far – Rakhigarhi.

    Also, here is an article with some interesting maps –

    See how the lactase tolerance is highest in exactly the same regions as those where the Indus civilization prospered with the highest concentration in Haryana which is not only the Vedic homeland and also the place with a large no of pre- and early Harappan sites as well as the largest one known so far. Europeans and Soutb Asians share the gene for lactase persistence, so why is this gene showing a geographic spread more aligned with the Indus civilizational spread in India ?

  6. @Razib
    Paper from Sep 2017 on Jat Y-Dna.

    L – 36.8%
    R – 28.5%
    Q- 15.6%
    E,G,H,I and T- 9.5%

    “The results of haplogroup analyses revealed that MRCAs of 302 Jats in our dataset belonged to nine different haplogroups—E, G, H, I, J, L, Q, R, and T—with nine different geographic origins.

    The same nine haplogroups were used to compare the Jats with other ethnic groups of the Indian subcontinent. The haplogroups of 302 Jats and 1,855 other men in 38 ethnic groups of Bangladesh, India, and Pakistan are displayed in Table 2.”

    “The results signified that the Jats shared an underlying genetic unity with several other ethnic communities in the Indian subcontinent with the same MRCAs and geographic origins. About 90% of the Jats and about 75% of the other 38 groups in the study belonged to the same four haplogroups J, L, Q, and R.”

    I am a bit surprised that H is not one of the lineages that accounts for 90% of Jat samples, and such high % of Q .

  7. pure Indo-Aryan specimens

    Assuming there ever was such a thing, if by “specimen” you’re hinting at information encoded in one’s genes. I think the earliest attested Indo-Iranian itself has non-IE linguistic substrate, which hints at elite dominance by an IE speaking culture of non-IE people around BMAC. The correlation, if any, between genetics and language is questionable even before India proper had IE speakers. The so called Aryans were quite probably mongrel IEers to begin with.

    And Indo-Aryan was a language and a culture – the word is specialized linguistics terminology. One isn’t born Indo-Aryan (anymore that one’s born Latinate), one becomes Indo-Aryan by learning (Sanskrit) – shit that all humans are pretty damn good at doing – especially Tam Brahms 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *