Demographic replacement in Southeast Asia during the Holocene

Well sometimes you feel silly, and it’s not your fault. Yesterday our podcast on Sundaland went live (we talked about Doggerland and Beringia too!). Though I expressed a fair amount of skepticism, I took the argument that Stephen Oppenheimer presented in Eden of the East, that modern Austronesians are long-term residents of Southeast Asia, seriously.

The alternative view, most forcefully put by Peter Bellwood in books such as First Farmers, is that Austro-Asiatic and Austronesian people were agriculturalists issuing out of southern China that transformed the region over the past 4,000 years (the Austronesians from Taiwan specifically, though during the Pleistocene Taiwan was connected to the mainland).

I lean toward Bellwood’s view, and today a preprint came out which basically confirms it in totality, Ancient Genomics Reveals Four Prehistoric Migration Waves into Southeast Asia. The abstract:

Two distinct population models have been put forward to explain present-day human diversity in Southeast Asia. The first model proposes long-term continuity (Regional Continuity model) while the other suggests two waves of dispersal (Two Layer model). Here, we use whole-genome capture in combination with shotgun sequencing to generate 25 ancient human genome sequences from mainland and island Southeast Asia, and directly test the two competing hypotheses. We find that early genomes from Hoabinhian hunter-gatherer contexts in Laos and Malaysia have genetic affinities with the Onge hunter-gatherers from the Andaman Islands, while Southeast Asian Neolithic farmers have a distinct East Asian genomic ancestry related to present-day Austroasiatic-speaking populations. We also identify two further migratory events, consistent with the expansion of speakers of Austronesian languages into Island Southeast Asia ca. 4 kya, and the expansion by East Asians into northern Vietnam ca. 2 kya. These findings support the Two Layer model for the early peopling of Southeast Asia and highlight the complexities of dispersal patterns from East Asia.

The transition to full-fledged rice agriculture occurred in Vietnam ~4,000 years ago. In First Farmers Bellwood reports on an archaeological site dating to that period where skeletal evidence has been adduced to record the presence of both Northeast Asian and Australo-Melanesian types. These results make clear though that these hunter-gatherers in Southeast Asia are more similar to the Onge of the Andaman Islands, as well as the Negritos of the interior of the Malay peninsula. They’re totally in alignment with the earlier morphological results (also, readers might be curious to know that one site of the Hoabinhian culture is in Yunnan, China). This shouldn’t be surprising, as the Andaman Islands were a peninsula which extended from southern Burma during the Pleistocene.

Already the most accepted model for the introduction of intensive agriculture into Southeast Asia is that it was brought by Austro-Asiatic peoples. These results confirm that. Additionally, it seems clear that Austro-Asiatic ancestry made it to island Southeast Asia, whether directly or through Austronesian admixture before arriving in island Southeast Asia. Java and Bali have some of the higher fractions ancestries most closely associated with Austro-Asiatic groups on the mainland.

Deeper digging into the admixture distributions has long made it pretty evident that some areas had much higher Austronesian fractions in Indonesia than others, and it wasn’t just a function of distance from the Phillippines. Why? My own hunch is that Austronesians brought social and cultural systems which were better adapted to island Southeast Asia, and were more fully able to exploit the local ecology. Meanwhile, aside from a few fringe areas such as the Malay peninsula and coastal Vietnam, they were not successful on the mainland.

The authors also detect migrations into Southeast Asia besides that of the Austro-Asiatics and Austronesians. One element seems correlated with the Tai migrations, and another with Sino-Tibetan peoples, most clearly represented in Southeast Asia by the Burmans. The excellent book, Strange Parallels: Volume 1, Integration on the Mainland: Southeast Asia in Global Context, c.800–1830, recounts the importance of the great migrations of the Tai people into Southeast Asia ~1000 A.D. Modern-day Thailand was once a flourishing center of Mon civilization, an Austro-Asiatic people related to the Khmers of Cambodia. The migrations out of the Tai highlands of southern China reshaped the ethnography of the central regions of mainland Southeast Asia. The Tai also attempted to take over the kingdoms of the Burmans. Though they failed in this, the Shan states of the highlands are the remnants of these attempts (tendrils of the Tai migrations made it to India, the Ahom people of Assam were Tai). Vietnam, shielded by the Annamese Cordillera, came through this period relatively intact. It is also well known that Cambodia’s persistence down to the present has much to do with the shielding it received from France in the 19th century in the wake of Thai expansion.

There are two bigger issues that this paper sheds light on. One is spatial, and the other is temporal.

They detect shared drift between Austro-Asiatic people and tribal populations in northeast India. This is not surprising. A 2011 paper found that Munda speaking peoples, whose variant of Austro-Asiatic is very different from that of Southeast Asia, are predominant carriers of Y chromosome O2a. This is very rare in Indo-European speaking populations, and nearly absent in Dravidian speaking groups. Additionally, their genome-wide patterns indicate some East Asian admixture, albeit a minority, while they carry the derived variant of EDAR, which peaks in Northeast Asia.

One debate in relation to the Munda people is whether they are primal and indigenous, or whether they are intrusive. The genetic data strongly point to the likelihood that they are intrusive. An earlier estimate of coalescence for O2a in South Asia suggested a deep history, but these dates have always been sensitive to assumptions, and more recent analysis of O2a diversity suggests that the locus is mainland Southeast Asia.

Now that archaeology and ancient DNA confirm Austro-Asiatic intrusion into northern Vietnam ~4,000 years ago, I think it also sheds light on when these peoples arrived in India. That is, they arrived < 4,000 years ago. As widespread intensive agriculture came to Burma ~3,500 years ago, I think that makes it likely that Munda peoples arrived in South Asia around this period.

I now believe it is likely that the presence of Austro-Asiatic, Dravidian, and Indo-Aryan languages in India proper was a feature of the period after ~4,000 years ago. None of the languages of the hunter-gatherer populations of the subcontinent remain, with the possible exception of isolates such as Nihali and Kusunda.

The temporal issue has to do with the affinities of these peoples, and how they relate to the settling of Eastern Eurasia. All the Southeast Asian groups after the original Australo-Melanesians share more of an affinity with the Tianyuan individual than Papuans. The implication here is that Tianyuan is closer to the ancestors of various agriculturalists in Southeast Asia than just some random basal Eastern Eurasian. But, since Tianyuan dates to 40,000 years ago, and, is from the Beijing region, it is hard to make strong inferences from comparisons with only it. The heartland of ancient Chinese culture in Henan was to the south of the Tianyuan, after all. More samples are needed before one can truly tease out the pattern of isolation-by-distance vs. admixture that led to the emergence of the proto-farmer populations which settled Southeast Asia.

In the podcast above one thing that came up is that a lot of genetic data indicate decreased diversity as one moves from the south to the north in East Asia. This has long been taken to mean that humans migrated north, and so were subject to bottleneck effects. I pointed out that this may simply be a consequence of admixture between two very different groups of people in Southeast Asia, elevating diversity statistics.

And yet as the map at the end of the preprint suggests it is highly plausible that Pleistocene Asia was marked by a south to north dynamic of migration. The Austro-Asiatic peoples who migrated south during the Holocene may simply have been backtracking the migration of their ancestors. What these results, and ancient DNA more generally, tell us is that humans were often on the move. The Pleistocene world of climate change probably meant that humans had to be on the move.

17 thoughts on “Demographic replacement in Southeast Asia during the Holocene

  1. This seems to make it pretty unlikely that the IVC was munda or even para-munda speaking.(The munda loans in the veda will have to be accounted for though) The explosion in ancient DNA studies is transforming what we know about the human past. Apparent continuity in archaeological cultures doesn’t seem to rule out migration. Wonder when linguists, archaeologists and historians will starting revising their models since their assumptions of population continuity since the paleolithic stand refuted ?

  2. This seems to make it pretty unlikely that the IVC was munda or even para-munda speaking.(The munda loans in the veda will have to be accounted for though

    yeah. seems like it.

  3. Re our previous discussion on Amerindian origins, how do you interpret this?

    “We also investigated the affinity between certain Australasian populations
    and particular Native American groups, like the Surui (45, 53, 54). When computing D(Mixe,Surui, X,Yoruba), we find that the Group 1 samples had some suggestive but non-significant affinity to Surui relative to Mixe (Z = -2.18 when X = Ma911, Z = -2.48 when X = La368; Table S19), although the signal is not as robust as observed for Tianyuan (Z = -3.53), Khonda Dora (Z = -3.04) and Papuans (Z = -3.02), among others (53, 54). We note, however, that there are much fewer SNPs to compute this statistic on Group 1 samples than on the other populations (La368: 191,797; Ma911: 47,816; Tianyuan: 295,628, Papuan: 471,703, Khonda Dora: 496,097), thus we may be underpowered to detect this signal.”

  4. stuart, obv they are looking to detect the weird patterns in amazonians that were discovered a few years ago. i think the key problem is that we don’t have many ancient samples from eastern eurasia so we can’t interpret things before well.

    remember in 2012 when ppl were trying to figure out why/how europeans had a siberian affinity? then ma’lta and EHG were discovered and it made more sense….

  5. Razib,

    Agreed, more will be better. For now, it seems to me the key project, with what is available, is to disentangle, allele by allele, the relationships of Tianyuan, Malta/Afontova Gora, Devil’s Gate, Onge, Papuans, Anzick and the Upward Sun River infants.

    FWIW, archaeology indicates a dramatic cultural separation, starting before ca. 25,000 cal BP, between northern China (with microblades linked to northern Eurasian blade-making traditions all the way to Europe) and Hoabinhian-connected southern China.


  6. FWIW, archaeology indicates a dramatic cultural separation, starting before ca. 25,000 cal BP, between northern China (with microblades linked to northern Eurasian blade-making traditions all the way to Europe) and Hoabinhian-connected southern China.

    i’ve read that.

    lazaridis et al. i believe so that mesolithic WHG have east asian admixture somehow some way….

  7. One thing that surprises me is on Figure 3, Page 25, where they attempt to fit ancient and present individuals. They note that Ami (Taiwanese aboriginal group – basal Astronesian) gets modeled as a mixture of Onge-like and Tianyuan-like. This in itself isn’t surprising – I’ve always thought it’s likely there was some foundational admixture of an Austro-Melanesian type population into basal East Asians. However, the graph clearly shows the Ami are modeled as 76% Onge-like, and only 24% Tianyuan-like. This seems to suggest that Tianyuan Man is not particularly close to the “main branch” of East Asian ancestry.

    Like you, I am eager for more ancient DNA from China – particularly “China proper.” We know that several genetically (and linguistically) differentiated populations left South China for southeast Asia, but we still don’t know why they are genetically distinct. Did they have thousands of years of relative genetic isolation? Low levels of admixture from various outside groups like ANE or something even more westerly? I presume that such studies must be in the works, but I know if you do have any such information it would be embargoed.

  8. Frankly, it is inexcusable that ancient DNA from sub-Saharan African and SE Asia–two of the most adverse regions on Earth for its preservation–has been published before anything significant from China. China has many cold, dry areas, burgeoning wealth and scientific societies, and the world’s largest population.

    I don’t wish to sound conspiratorial, but is there some official blockage going on here with DNA research? Or are people afraid of angering the Chinese, like the situation with India?

  9. “Australo-Melanesian types.”

    This phrase is used in the paper and may be “technically” correct in the same sense that European Americans are called “Caucasians” even if they have little or no ancestral origin in the Caucasus.

    But, it is a phrase that implicitly assumes continuity and ancestral identity between Asian Negrito populations and the population ancestral to Papuans and Australian Aboriginal populations over about 60,000 years.

    At a minimum, this connection has not yet been credibly established. Realistically, this identity of ancestry is more likely to be wrong than it is to be correct.

    While the two ancient DNA specimens from Mainland Southeast Asian hunter-gatherers do show some pretty modest percentage of Papuan genetic origins in an ADMIXTURE analysis and by other statistical tests, the models showing that aren’t great fits in any case (no simple three population mixes are a good fit to the data), and the inferred percentages are low enough to seem to be a decent fit to a model where pre-Papuan people are mostly (80%+) replaced by a subsequent Onge-like population that is closer to the Tianyuan individual than to the Papuans, even though these early Holocene hunter-gathers have some Papuan affinity, while later agriculturalists have almost none. And, the paper establishes fairly well that this isn’t simply an artifact of Denisovan admixture that is absent in Southeast Asians to the West of the Wallace line and in Mainland Southeast Asia.

    Given the Tiayuan individual’s antiquity (i.e. 40kya), the naive inference would be that the Tianyuan individual is part of a second wave of modern human migration to SEA and EA that took place after pre-Papuans migrated ca. 70kya, but before 40kya, which isn’t too far afield from fitting estimates based upon mutation rate and phylogeny analysis of uniparental markers in modern populations.

    Of course, any time one hunter-gather population largely replaces another, it is much harder to figure out what gave the replacing population an edge than it is in the case of food producers who have much high population density than hunter-gatherers, or metal age people relative to Neolithic farmers – because swords, axes, metal tipped spears, metal maces, metal shields, and metal armor.

    My best guess would be that we know that pre-Papuan people didn’t have domesticated dogs, but that dogs were domesticated not that long into the Upper Paleolithic era. So maybe the Tiayuan individual, the Onge, the Ancestral South Asians who no longer exist in pure form, and Mainland Southeast Asian Negrito populations were all part of a wave of hunter-gatherers who utilized domesticated dogs as partners, which allowed them to conquer pre-Papuans who didn’t, not long after dogs were domesticated.

  10. “I now believe it is likely that the presence of Austro-Asiatic, Dravidian, and Indo-Aryan languages in India proper was a feature of the period after ~4,000 years ago.”

    If Dravidian is a recent arrival in India proper, the logical date for it to arrive/emerge would be more like 2500 BCE, one date commonly given for the South Indian Neolithic revolution, than 2000 BCE (i.e. the 4000 years ago), although the basic thrust of this argument remains a sound one. Dorian Fuller makes a pretty good case for a 2500 BCE date in “DATING THE NEOLITHIC OF SOUTH INDIA: NEW RADIOMETRIC EVIDENCE FOR KEY ECONOMIC, SOCIAL AND RITUAL TRANSFORMATIONS” by Dorian Q Fuller,, Nicole Boivin, and Ravi Korisettar (2007)

    See, in particular, Table 9 on the last page of the paper.

  11. Lipson et al. make this surprising statement about Juang (contemporary Austroasiatic speakers in India):

    For Juang, the inferred western Eurasian component is almost certainly itself admixed, but for the purposes of our model, it fits best as closely related to the Ancient Northern Eurasian lineage forming part of the ancestry of Native Americans. The deep eastern Eurasian component splits close to the same point as Onge, East Asians, and the indigenous Austroasiatic component. A mixture of two components of this type is characteristic of Indian populations today; Juang, however, also traces ancestry to a third,Austroasiatic-related source.

  12. @Stuart, what’s the surprising part? As a reference for West Eurasians, Their model only include Native Americans as a Mixe, in which their ANE stream is “West Eurasian”, so it’s not like there was a lot of choice, or many stats that would go wrong from this simplification (you need lots of real West Eurasian populations, preferably ancient on the graph to be able to distinguish between ANE vs other ancient West Eurasians).

    They may have been able to improve on the model (on page 3) by giving Juang a “basal Eurasian” edge, but since Juang get 13% ANE only, then, how much can, what, 2-3% Basal Eurasian into Juang at most really improve the graph’s fit? (By very, very, very marginally making them further from Ust Ishim than the graph would as published suggest).

  13. I was interpreting their ANE component as Mal’ta/AG (which constitutes about 30-40% of Mixe and other Native Americans). See Figure S3. As I read it, 63% of the Juang genome derives from a modeled ancestral group that was comprised of 79% ancestral Onge and 21% ANE (from the source population, i.e. ANE, contributing 30% to Mixe).

    And all this raises the question, how did those genes get into an aboriginal group in India? From an early common ancestor before the ANE headed farther north? Or maybe late introgression from high-caste Indians with a remote Yamnaya-derived ANE ancestral component?

  14. Yeah, but the issue is that with their topology they can’t really tell the difference between ANE and other forms of West Eurasian ancestry. It doesn’t actually mean that the West Eurasian ancestry in Juang is completely ANE, it’s just a limitation of the model. They can’t really tell the difference between Iran Neolithic, CHG, EHG, ANE, etc. with the limits of the model

    Hence “for the purposes of our model” though it “is almost certainly itself admixed”.

Leave a Reply

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