Knanaya & Kerala: perhaps there is something different down south?

Over the past few months I have been getting together some samples from people from Kerala, with a focus on Knanaya Christians. A subset of the broader St. Thomas Christian community, two things have jumped out in my analyses:

– they are quite endogamous

– they are shifted off the ‘India-cline’

More precisely, like Cochin and Mumbai Jews, they are often shifted toward Middle Eastern populations. This is relevant because the Knanaya believe themselves, like most St. Thomas Christians, descended in part from Jews or Christians from the Middle East.

All that being said, looking more deeply into the data I’m not quite as sure. One of the reasons is that Kerala may not be as “structured” as other parts of India. Some of this is well known. The Nair samples I have are shifted toward South Indian Brahmins, which is plausible in light of connections between Nairs and Brahmins. The Brahmin-adjacent Ambalavasi seem quite similar to Brahmins. These are not surprising. But, Kerala samples I have as a whole seem notably shifted on the India cline more toward the “north” than I would have expected. This could be due to gene flow from without and within Kerala, in a way that is not typical in other parts of the subcontinent.

I say this because even the Ezhava, who were basically what we’d call a Dalit community (no longer today), shows a shift.

Instead of talking, let me post some admixture plots (unsupervised):

Now, supervised:

Now TreeMix:

Here is an admixturegraph (using the Narasimhan et al. right-populations):


I ran f3-stat. Here it is filtered of any z-scores that are > -2.



27 thoughts on “Knanaya & Kerala: perhaps there is something different down south?

  1. A clue to this may be the distribution of y-haplogroup L. It is clearly Indus Valley. And the max current distrubution is along the Indus river and the western coast of India right up to Kerala. As you move eastward from the west coast L reduces proportionately. Also the distribution of L is probably also representative of distribution of Indus valley in India. Keralites must exhibit higher proportions of Indus Valley than its eastern neighbours. Hence Kerala will be shifted more to the North.

  2. Because Kerala has had ancient trade relations with the Middle East, the majority of the population, including Hindus, will show a shift. When you think about the historical spice trade this makes a lot of sense. However certain groups like the Knanaya will likely continue to show more shift/higher admixture because they have “actual” Middle Eastern ancestry compared to the other groups. It seems like the Knanaya story holds up. What do you think?

  3. Celebrated 14th century Muslim scholar Ibn Battuta stated the following about Kerala’s ports (Kozhikode/Calicut in particular) “…one of the largest ports in the world that was visited by merchants from China, Java, Ceylon, the Maldives, Yemen, and Persia”. I think we’re starting to see the reality of this in genetics too (I agree with Harazi entirely).

    Certain groups like the Knanaya because of their descent from a small party of foreigners, seem to maintain more admix than the average population. What an interesting group! To think that Canaanites exist in Kerala is insane.

  4. Is the north shift restricted to Kerala or is it seen throughout western coast? That is right from gujrat through Konkan upto Kerala relative to people living on Deccan plateau on their east.

    This can be route used in post IVC migration. The L haplogroup as written in one comment above peaks in Maharashtra if I remember correctly.

  5. @iamVY

    Yeah it looks like post-IVC migrations from the north made a greater impact on the West coast compared to the East. West coast groups like Patel, Bunt, and all the Kerala groups score more Steppe and less AHG than the Velama. The second supervised admix plot also shows a stronger ‘N. European’ signal in the Keralite groups (except knanaya?) compared to Velamas who are a sister population.

    I suspect that both K_Vaniya and K_Varma are the product of more recent migrations from the North.

  6. The ezhava are not a real dalit/adivasi group at all. They are a lower middle caste group at worst. They would be regarded as sudras in the brahminical worldview. They have origin myths of being Buddhists from Sri Lanka (at least of partial descent). R1a1 is around 27% in their community. Mixing with namboothri Brahmins possibly affected the whole of Kerala.

    The pulayar are dalits.

  7. I know it’s tedious but it helps a lot if you could label the PCA plots. It’s highly preferable to distinguishing the faded colours.


  8. I think geography and trade help explain this, and I would be interested in seeing data re other groups along the West Coast (e.g., Mangalore, Goa, Mumbai).

    Geographically, the West Coast Plains are quite narrow — roughly ~50km in width. On one side of the WCP, you have the Arabian Sea.

    On the other side, you have the Western Ghats. The Western Ghats have a steep incline — there’s not much in the way of foothills before you start approaching the peaks. The Western Ghats are sometimes not thought of much since the Himalayas are so much bigger, but they’re still big especially since it’s a steep incline from pretty much sea level. It peaks at 2.7km in elevation and averages about 1.2km in elevation.

    Interestingly, the Western Ghats increase in height as you go further south — meaning that Kerala is perhaps most affected by it (although there are a couple of passes through the Ghats).

    So you’ve got a narrow coastal plain that’s fairly secluded from other parts of the subcontinent geographically.

    It’s also among the most diverse parts of the country — the religious demographics speak for itself. Kerala is 55% Hindu, 25% Muslim, 20% Christian. Goa and Mumbai are 66% Hindu (compared to the national average of 80% Hindu). And that’s probably because it faces Europe and the Middle East. Some groups probably have admixture with these traders.

    Also, the climate is different because of this geography. The WCF are humid, rainy, tropical, verdant, etc. This could have some phenotypic impacts, I don’t know.

    I guess my point is that it’s a very distinctive region within India.

  9. I’d like to second Anik’s request. I asked a few months ago, and you did it for a couple of weeks. That was much better.

    With this post, I had to give up trying to understand the plots seriously. But the black marks with the different shapes helped for the relevant parts.

  10. The enhanced geneflow within Kerala among various castes compared to other parts of the country seems very right. Kerala always seems to have had a very distinctive culture in the medieval period (which included weird practices like not allowing lower-caste women to cover the upper parts of their bodies). The rest of India (especially the other parts of south India) probably had a more old-fashioned religious conservatism when it came to the endogamy aspect of perpetuating caste system and upper caste men probably did not exploit lower caste women as much as we would expect from a feudalistic society, or maybe they did and I don’t know. On first glance it seems like Kerala was not also that extremely feudalistic or something but maybe it was, I don’t know at all. And I also don’t know if the other parts of south India were any less feudalistic with lower incidence of sexual exploitation of lower-caste women by upper-caste men – it’s just my impression what I noted above – a pertinent question might be to see what is the ranking of the internal homogenization vs. heterogeneity of all caste populations speaking a given language for various Indian states. Even in other parts of south India, particularly in river valley and delta regions like that of Kaveri, Krishna-Godavari, etc., things like Devadasi system existed so perhaps historical lower castes in these regions show a steppe/Indus_Periphery shift more than what might be expected?

    Regarding gene flow from without Kerala, isn’t this the similar kind of problem as that old no-East-Asian-ancestry-in-Jatts-for-them-to-have-been-Sakas, etc.? Would not we expect clearcut non-Iranian/northwestIndian-mesolithic-related West Asian ancestry if the gene flow was from outside India (I’m conflating “without Kerala” with “without India” here as I wondered what new distinguishing, particularly north-ish ancestries would Kerala get from other parts of India, since it seems equally isolated from all the other inland regions of India in history – maybe Gujarati seafarers of some kind exclusively contributed some gene flow from a more northern place within India to Kerala?) The dominant discussion in the air always seems to indicate to me that all/most of these surviving Indian populations don’t really have any significant components after that initial ethnogenesis of 1000s of caste groups involving Indus, Steppe, and AASI; things like the northwestern Iranic kingdoms even in the northwest of India and the northwestern Deccan, anything from Southeast Asia related to trade in the southeastern states of Odisha, Andhra, TN, etc. so I’d think even Kerala did not get much gene flow from regions like Arabia, etc. except for some religious groups like the Kerala Jews, Knanaya Christians, etc.

    The thing with the possible west-east geographical cline of Indus vs. AASI ancestry in south India is also a very interesting problem. I usually think the Indus vs. AASI cline is mostly along the lines of caste (the south Indian version) with agriculturalists and some elite artisan castes also perhaps having similar levels of Indus ancestry throughout south India but is this not right? Do Telugu agriculturalist castes, let’s say, have lower amount of Indus ancestry than those of Karnataka? If that’s the case I can still perhaps try to explain the phenomenon by tying it down to the original (south)western and (north)eastern Dravidian dialectal division in Proto-South-Dravidian in the form of Tamil-Tulu (southwestern) and Telugu-Kui (more northeast) and then hypothesize that the Tamil-Tulu overwhelmed the western AASI more than the Telugu-Kui did/could the eastern AASI (and the Munda-related Southeast Asian component, etc.), but it would be more complicated than I would like it to be lol. Or it could all be even more complicated and the hypothetical geographical cline could be the result of even earlier migrations of Indus mesolithics or someone into the western parts of the Deccan who were not really speaking any Dravidian languages but some other different thing altogether which along with the languages of the AASI (if they existed) went extinct because of language shifts to the South Dravidian language of the later-migrating Indus peoples.

  11. 3 funny stories.

    1. My father noticed during a documentary that what you wear during a hajj at the Kabba is traditional kerala wear. There is even a certain way your are supposed to wear it (tucking it either left or right, I don’t remember) which is the same as what he remembered from his grandfather.

    2. In Nepal, I found a lot of architectural elements that were also found in found in very old temples. I mean onion domes are found in a lot of places as well.

    3. Strangely, as a light skinner nair (Shashi Tharoor is a cousin) I get mistaken for kashmiri quite a bit. Even by Kashmiris. Hell, Shashi married one. I bought a nice Kashmiri girl down to Kerala — her name reveals her hindu origins although she’s Muslim now. Again she was seeing similarities all over the place as well.

    Kerala is a very strange place. I won’t even go into the swastikas on the Cochin synagogue.

  12. Hi Razib,

    I believe you indicated in earlier posts that South Indian Brahmins are all very similar to each other genetically.

    However, in the chart you posted here, it seems that even Kerala Iyers are distinct from Tamil Brahmins (who are predominantly Iyers, as well). This seems like the two communities of South Indian Brahmins who should have the least genetic differences between them.

    The difference I’m noting here is the steppe to ahg ratio, where K Iyers are more steppe skewed and Tamil Brahmins are more ahg skewed.

    Any thoughts?

  13. It seems that Tamil Brahmins from Tamil Nadu are uniquely less steppe-shifted. Cultural memory points to Kerala Iyers being descended from TamBrahms, but I wonder if it’s possible that the record is either wrong or that Kerala Iyers are representative of that population prior to their admixture with the local population.

    I wonder, is this decreased steppe admixture also evident in Brahmins from Andhra Pradesh as part of the West => East phenomenon?

  14. Kerala Iyers are TamBrahms. My grandfather was a Kerala Iyer, but spoke Tamil his entire life, and the Kerala Iyers in my family think of themselves as TamBrams.

    However, Nambudiris are entirely a Kerala and Malayali population in terms of language and culture.

  15. @ Mr. Iyer,

    Agreed, Kerala Iyers are Tamil Brahmins and still speak Tamil at home in Kerala.

    But then what explains the genetic variance you see here between K. Iyer and Tamil Brahmin?

  16. I think both Iyer and Tam Brahms are invaders to our land. As great Periyar said “Build that wall, and make these Tam Brahms pay for it”.

  17. @Hoju, Mr. Iyer

    Can think of three possibilities:

    (1) Kerala Iyers are an indicator of what Tamil Brahmins looked like before local admixture.

    (2) Kerala Iyers are a subset of Tamil Brahmin internal diversity and the range of steppe ancestry between Tamil Brahmins is highly variable, perhaps dependent on different semi-endogamous sub-groups.

    (3) Tamil Brahmins who came to Kerala mixed with a higher steppe ratio population which led both populations to have more similar steppe components (maybe Nambudiris).

  18. @suryavansa,

    Option 2 looks most likely. Iyers only moved into Kerala relatively recently, mostly under the pressures of Tipu Sultans expansions. This is long after strict endogamy set in.

    Prior to the late 19th century, Iyer subgroups did not frequently inter marry, so there was a bit more diversity. I know that Palakkad Iyers are mostly, if not entirely in some villages, Aşțasahasram.

    That said, do we know how large these reference samples are? If it’s relatively small, these could be sampling size flukes. I can’t really see any historic reason for this high variation, or mislabelling.

  19. @Mr. Iyer, suryavansha

    Agreed, #2 is most likely.

    For #1, migrations to Kerala are documented and happened within the last 300 years or so I think.

    For #3, before modern times I don’t think there’s been a documented phenomenon of Nambudiri-Iyer marriages (unlike, say, Nambudiri-Kerala Non Brahmin marriages, especially with Nairs).

    For #2, I think even phenotypically it makes sense. There are Tamil Brahmins who look indistinguishable from Tamil Non Brahmins, and there are Tamil Brahmins who look indistinguishable from North Indian Brahmins (although both of those groups can be diverse in appearance too).

    Also, the various semi endogamous sub groups of Iyers seem to be partly based on when they arrived to TN? So maybe the distinction is between earlier arrivals and later arrivals, and that the ones who moved to Kerala belonged to the later arrivals.

  20. “which included weird practices like not allowing lower-caste women to cover the upper parts of their bodies). ” — @Santosh,i thought the nangeli story was fake .What about nair and namubudiri women not covering their breasts ?

    “even the Ezhava, who were basically what we’d call a Dalit community (no longer today)” — How ? Are you saying that based on some incidents of nairs not letting them in their temples ?

  21. @hoju,

    The thing is that these subgroups may also have been relatively short-lived, or recent, groups. While intermarriage between the Aşţasahasram, Vadama, and Brhatcaranam may have been relatively new in 1890, we don’t know if these groupings themselves only came from 1690.

    I’d like to get a better idea of those sample sizes, since some reference labels weren’t exactly great in the past, like the mysterious “North Kannadi” of early Indian genetic research.

  22. td, that thing I read was a long long time ago and I did not revisit any relevant material before writing the comment. However, I believe in many cases my initial perceptions of stuff that I encounter (simple stuff like facts, events anyway) tend to be accurate because I somehow am mostly able to intuitively discern if my source of knowledge for these sorts of things is good or bad; that sort of thing. With more abstract things like arguments and analyses, I of course am easily swayed and can’t tell if they are valid, accurate, etc. (in other words, not much capable of critical thinking). About my memory also I tend to have a bit of a confidence but as I illustrate in both this message below and in a comment I will put on the Open Thread, that belief is proving to have been a mistake. But yeah, even in this case, I don’t remember what was my source of knowledge about this perception that lower-caste women in Kerala were subject to non-consensual activity of this kind as a normal institution in the past, but I am moderately confident that this was a trustworthy source. I now realise that this source must not have been something as basic as Wikipedia even because I haven’t got any memory triggers even as I looked at the Wikipedia article after I saw your comment lol. Anyway, Wikipedia does suggest that something called mulakkaram or breast tax was a thing in Kerala in the past and cursorily, there seem to be ample citations, many of which are to news articles but some are to history books as well.

    Yes, I had in mind the impression of this story of a woman who cut off her breasts as a protest and believed all this while that this was true (haha casts serious doubt on the confidence I said I have in being able to judge my sources properly for simple things like these, doesn’t it?) and Wikipedia states that the story of this woman might not be historically accurate (so thanks a lot to you of course for letting me know that), but I was just talking about the phenomenon of breast tax ultimately. Well, I now realise that I never wrote precisely this because I had never remembered this to this level of detail and instead had believed that they were simply not allowed to cover their breasts at all and did not even have the option of paying a breast tax if they wanted to cover their breasts. (As a person with very base social liberal instincts, I later partially internalized that this was extremely immoral because of the lack of even basic consent by the women subjected to it (to allow an easy, typical social liberal room in my mind for the possibility that this would probably be okay if there was informed consent and no-harm-done-really-if-so and all that)).

    But I also notice several inconsistencies even in the Wikipedia article as it seems to say on one hand that Kerala’s lower-caste women were allowed to cover their breasts if they paid the breast tax, but then it also seems to say that tax collectors went about collecting this breast tax from every lower-caste woman anyway. Which is it? Hope some accurate history book/chapter on the topic presents falls on me in the future regarding this.

    And regarding Nair and Nambutiri women not being allowed to cover their breasts, apparently Nairs were not allowed to do so majorly when in the vicinity of Nambutiri men (as they fell one rung or so (six rungs? six feet? something like that haha) below Nambutiris on the caste ladder) but apparently Nambutiri women were not allowed to cover their breasts only in front of deities. They were otherwise not even allowed to come out of the house very much also, being antarjanams and all, but that’s a different type of a thing. This information from Wikipedia again and that article says that there was a religious dimension to all of this in that people were expected to symbolically show obeisance to entities above their place (like humans of upper castes for humans of lower castes, and God for humans of upper castes) by not covering their breasts so actually there may have been some type of a nice religiously induced harmony in all this and everyone was deluded with lower amounts of psychological pain relatively (I hope so certainly), though the liberal principle of absolute-necessity-of-informed-consent may have not been adhered to?

  23. Also, needless to say, the most unhinged piece of misleading that I did yesterday was in somehow associating this phenomenon of lower-caste women not allowed to cover their breasts with the phenomenon of possible sexual exploitation of lower-caste women in Kerala (and elsewhere). It has become crystal clear that I knew nothing accurately about the former but I of course don’t know anything about the latter too. I had just assumed that the phenomenon of not allowing lower-caste women to cover their breasts (under my understanding of it at the time which did not even take into account the presence of the possibility of women actually allowed to do so if they paid the damn tax) was majorly driven by the lust of the upper-caste men and then continued to consider that this hypothetical lust somehow makes us infer that they actually sexually exploited the women in question as a practice and had northern-shifted progeny who were inducted into their mothers’ castes.

    As I now saw a genuine religious angle seems to have been present for this (of course, whatever did people do in India in the past that is non-religious?) and though I still believe my suspicions may be right in that the motivations of the men of upper-castes may not have arisen fully out of genuinely moral religious principles but at least slightly must have been influenced by immoral abusive impulses, it weakens my belief many fold as you never know how people really think and what their motives are. They may have really did all they did just out of innocent religious belief! (But just for your amusement, this above also illustrates to you my natural biases against religion and towards a progressive advancement of morality in human society with the onset of European modernity really sealing it; I seem to be a typical revolutionary who will be definitely overthrown by his grandchildren (I so hope to God that they at least come about haha) in a bloodier revolution lol!). Irrespective of any possible accuracy in my conclusions and results, I should not have brought up this entire thing originally being very poorly informed about any of these issues.

    But that said, I believe the entire exercise helped me a bit towards discovering a framework for myself to view things in – that many of the genetic homogenisations, if any, of divergent caste populations, etc. in India might have happened because of sexual abuse. The endogamy aspect of caste system is again unpalatable to a liberal leftist (not every type of leftist but just the liberal one) instinctively majorly and also intellectually due to some other reasons but everything of the past looks quite bad – lack of mixing bad (because of the badness of the overarching structure (and also just instinct)) and mixing bad too (because of the seemingly low probability of such mixing being in the background of genuine (or even just sufficiently) loving relationships and not of sexual abuse). This I believe is again driven by my strong bias against all things past, and my belief in the concept of moral progress with time as opposed to the traditional Hindu view which states that there is invariably moral regress with time (for a given cycle of time). But then nothing to be too congratulatory about my “discovery” of the above “framework” either – it’s most probably just because of a lack of sufficient brainpower and imagination to think of other, actually more probable possibilities that explain the observations accurately. And actually deluding myself into thinking that this type of a very nasty kind of a thing actually even played a role. For example, I came to know about this heuristic recently that apparently large groups have to be not considered too morally evil as apparently their emergent morality tends to be of a sort of pleasant nature only overall, because of statistics or something. And I have just realised that a much simpler alternative that may explain the Kerala results is that the caste-endogamy system reached Kerala late, never became very strong, etc. in contrast with invoking anything to do with abuse.

  24. Razib, are the names of the samples all you know about their precise caste backgrounds?


Leave a Reply

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