Sons of the farmers, the story of Japan

By Razib Khan | December 16, 2010 1:52 am

Ainu in 19th century Hokkaido, and rice paddies

Unlike some islands Japan has a long history of human habitation. More interestingly, under the Jomon culture the Japanese archipelago was home to one of the earliest, if not the earliest, societies which used pottery. The Jomon do not seem to have been intensive agriculturalists. Rather, with a widespread marine littoral they likely maintained extremely high population densities, and at least semi-sedentary habitation patterns, simply through a hunting & gathering mode of production. Pacific Northwest Amerindians are likely a good analogy. They also relied on a dense stock of marine life to maintain population densities of a high level and a sedentary lifestyle.

About 2,000 years ago the Yayoi people arrived in Japan. The first Yayoi settlements are in northern Kyushu. These people brought intensive agriculture, in particular rice agriculture, to the Japanese archipelago. The general assumption is that the Yayoi are the precursors of the Japanese who entered into the international system of East Asia during the Tang dynasty in the second half of the first millennium. The Ainu of Hokkaido are presumed to be the descendants of the remaining Jomon people, maintaining a hold in the northern island because of its ecological unsuitability to Japanese agriculture.

The question is: what proportion of the ancestry of modern Japanese is Jomon/Ainu, and what proportion is Yayoi? The dynamics here are nicely constrained by the fact that Japan is a relatively isolated island system. The Yayoi seem to have arrived at one discrete moment in history, and rapidly expanded in ~1,000 years to all the main islands of Japan, though the full settlement of Hokkaido commenced in the 19th century. Interestingly, parts of northern Honshu seem to have had a distinct post-Jomon culture down to ~1000 AD.

Conveniently the HapMap has both Japanese and Chinese samples, but often there hasn’t been too much focus on the differences between these two groups because they’re very close in a global context when compared to the Yoruba or Europeans. In more recent analyses of East Asian groups the coverage seems to be better with various Chinese ethnic groups, but relatively few samples from Siberian populations. The latter are critical because the supposition is that these are the groups which would have the most affinities with the Jomon, due to the culture and contacts of the Ainu which evident during the modern period.

Dienekes most recent post on K = 15 ancestral components in ADMIXTURE clarifies some issues in this regard. There are multiple Han Chinese and Japanese samples, as well as a wide range of East Asian and Siberian groups. I’ve reedited and formatted K = 15 a bit, with the aim of focusing on the relationships of the Japanese in particular.


First, it is reasonable that the Denver and Singapore Chinese sample would have a greater proportion of the orange “Southeast Asian” component, Han in the United States are mostly from southern regions of China. Notice that the Japanese don’t have this component at all. The Japanese have more of the light gray component, which is modal among the Nganassan of the Arctic coast of Central Siberia. Unlike the Chinese the Japanese also have the blue component which is modal in Eastern Siberia, and also found in many North American groups (many omitted). Finally, it is interesting that the Japanese have the light yellow component in both of the samples Dienekes ran through ADMIXTURE. Going through his spread sheet, here’s are some of the populations sorted by this component (number 13):

Paniya (South Indian tribe)6.18
North Kannada1.44
Malay, Singapore1.34
Sakilli (South Indian)1.02
Japanese #20.7
Chinese, Beijing0.06
Chinese, Singapore0.03

This is basically Melanesian. Strangely, though at low proportions, the Japanese have much more of this than mainland Chinese, or most East Asians period. They’re in the same range as Cambodians, and topped only by maritime Southeast Asians and Indians. I find this interesting because Japanese Y chromosomes are often of haplogroup D, also common among the Ainu. This lineage spans various isolated regions of eastern Eurasia, such as the Andaman Islands. The implication is that D is a relict of a large set of populations which have slowly been absorbed by expansions of other groups. Additionally, some physical anthropologists have observed similarities in the morphology of the remains of the Jomon people and Australian Aborigines. I am inclined to chalk that up to the general robustness of non-farming populations, but it is something to consider.

But my primary point about writing this post was to offer some judgment as to the provenance of the modern Japanese. I think that looking at these results, and also keeping in mind other results on East Asian genetics, I’m of the mind that Japan looks to be a classic case where farmers totally marginalized hunter-gatherers. In other words, the modern Japanese are predominantly descendants of farmers from the Korean peninsula who arrived ~2,000 years ago.  The alternative is that the Jomon were more similar to mainland East Asians in Korea and China than they were to Siberian peoples, which seems unlikely if the Ainu give us any clue as to the culture of the Jomon. This seems more implausible in light of the fact that the Japanese do seem to have some affinities to Siberian people, to a greater extent than the northern Chinese of Beijing (Beijing is a large city, so likely this sample has people who are descendants of ‘reverse colonists’ from the South as well).

If this inference is correct then it is an amazing instance of demographic expansion and replacement. In much of the civilized world intensive farmers replaced extensive farmers in fertile regions in prehistoric times. Or at least during periods when textual evidence is thin on the ground. In contrast, in Japan the process persisted right up until the margins of written history. Additionally, we can peg the arrival of the Yayoi culture very well chronologically because of the discontinuity. Assuming 25 years per generation, in 40 generations the Yayoi had totally extirpated the post-Jomon cultures across all of the Japanese islands except Hokkaido. And, it is notable that the Jomon are generally judged to have been a relatively numerous for hunter-gatherers. Their longevity in Japan as a continuous culture also attests to their success. This narrow specific case may have larger implications for the demographic-genetic patterns we see in the rest of the world.

CATEGORIZED UNDER: Genetics, Genomics

Comments (21)

  1. Steve Tripp

    The one remaining mystery is that there are almost no (or no) remnants of the Jomon language in Japanese, assuming Jomon was related to Aizu. There are Ainu words in place names in Hokkaido, but there are no similar words in Honshu, AFAIK.

  2. Sandgroper

    Really interesting – this answers a lot of questions I have had.

    The guy on the left of the Ainu photo could look a bit like an Australian Aborigine (aside from skin colour), but it’s a stretch.

  3. bioIgnoramus

    That link to the Jomon takes us to: “The Early and Middle Jōmon periods saw an explosion in population… during the prehistoric Holocene Climatic Optimum (between 4000 BCE and 2000 BCE), when temperatures reached several degrees Celsius higher than the present…. After 1500 BCE, the climate cooled, and populations seem to have contracted dramatically.”

    That obviously can’t be settled, concensus science.

  4. AG

    Fascinating stuff. Is there any relation with pacific islanders?

  5. Laassi Hippeläinen

    I expected some connection with Polynesians, because they seem to have origianated in or near Taiwan. Island hopping along the Ryukyus to Kyushu would have been easy. And the Japanese language has features that can be associated with Austronesian languages.

  6. Razib: Might the implications of the outcompetition of the Jomon be more proof that genetic diversity among humans today has shrunk considerably from the period predating the expansion of agricultural civilizations?

  7. #7, i suspect so.

    The one remaining mystery is that there are almost no (or no) remnants of the Jomon language in Japanese, assuming Jomon was related to Aizu. There are Ainu words in place names in Hokkaido, but there are no similar words in Honshu, AFAIK.

    two thoughts

    1) HG languages can be quite diverse. look at the diversification in the new world. the jomon residence in japan is on the same order of magnitude

    2) the same phenomenon seems to have occurred in england, where celtic place names are strangely absent

  8. pconroy


    While I largely agree in the case of England, I also think there has been some deliberate effort to shift derivation of some words and places to Latin, rather than admit that they were more likely Celtic/Brythonic – as many Latin/Celtic words have the same root.

    But to give a solid example, the Bard of Avon – lived by a river called “Avon” – in Gaelic River=Abhain, pronounced “Aw-an”, so that’s clearly an example.

  9. Caroline Macafee

    There are various rivers called Avon in Scotland, which place-name scholars recognise as being from the same root as the word for ‘river’ in modern Celtic languages. I don’t have any source to hand to check on Shakespeare’s Avon, but river names are known to be amongst the oldest place-names, and the name of any large river is likely to be Celtic or even pre-Celtic, and certainly not Latin.

  10. includedmiddle

    Like Laassi, I immediately thought of the supposed link between Japanese and the Austronesian languages. A similar link with the Altaic languages is often proposed as well; indeed, some have suggested Japanese is an Austronesian-Altaic creole!

    But given that these are both minority components of their ancestry, it seems odd that the Japanese would speak the language of people they so thoroughly displaced.

  11. guys, the connection/likely origin of austronesians in taiwan is pretty well established. probably nothing to do with japan. it looks pretty likely that they came from korea, that’s the group closest to the japanese. that being said, on the PCA the japanese deviate “toward” the altaic/siberian groups more than the koreans. i suspect that’s the jomon substrate.

    but, as it said to my surprise, it’s a pretty minor substrate. though it would be closer to 25%.

  12. Scott the mediocre

    @Steve Tripp and Razib –

    The wikipedia article on the Emishi (which gives the impression of being well sourced) says that there _are_ “Ainoid” placenames in Tōhoku. But the likely supporting references are mostly in Japanese, which I can’t read to any useful degree.

  13. 1. The autosomal DNA data here pretty much replicate the Y-DNA and mtDNA which likewise show Jomon/Ainu contributions to be minor relative to Yaoyi contributions, although my recollection was that the older substrate was not quite so small in that sample. Y-DNA and mtDNA also show very distinct strata and so are relatively easy to distinguish from recent arrivals. (It would be helpful to have a sample to confirm that the light gray component really is frequent in the Ainu population, as opposed to being a remant of population structure or origin in the Yaoyi, in which case Ainu replacement would be far more dramatic.)

    2. Japan is normally considered one of the most ethnically homogeneous countries in the world, and for good reason shown here, but it was once a multi-ethnic pan regional empire, even though this lasted only briefly. My intuition is that some of the small outlier components may be remnants of this brief East Asian Co-Prosperity Sphere era.

    The small Melanesian (Yellow) component could be a product of the Japanese occupation of New Guinea in World War II with some locals brought back to the homeland as laborers. Without knowing the extent to which it appears at a fixed percentage in the Japanese population or varies a great deal in individuals, it is hard to know.

    The small East Siberian (Blue) component could likewise be a product of the prolonged Japanese occupation of Manchuria and Korea in the 1930s and 1940s, and it wouldn’t be hard to assign some of the Southeast Asian (Orange) component with Japan’s short lived effort to become a pan-Asian empire to its holdings from South China to Burma in that era.

    The percentages are small, the practice of taking warbrides and importing foreign workers and foreign notables did persist very briefly (followed by a long continuous period of xenophobia), there has been considerable population expansion in Japan since its expansionist period started, and low status minority populations sometimes have higher than average fertlity rates. There was also a long standing, low level practice of providing exotic concubines for elites for a long period of Japan’s history that could go towards explaining these populations.

    I’d be skeptical of assigning those components to deep history or grander population movements without more convincing evidence.

    3. It is also a little hard to tell if the Atlaic dark gray is the same as the Han dark gray, or is different, and if there is a difference is the similarities in color is coincidental or simply a function of the limited number of colors that a computer chart can display. The absence of the green, Siberian reds and blues and possibly of the Altaic dark gray, if it is a separate color, collectively seem to make a pretty strong case of a lack of a genetic identity between Altaic populations and the Yaoyi. The lack of as strong a genetic connection to Altaic which has been controversial linguistically in any case, isn’t necessarily a big surprise.

    Since China hadn’t consolidated its Northern origin kingdom into the South 2000 years ago, the lack of a strong Southeast Asian component in Japan also isn’t too surprising.

    But, the very strong genetic identity between the Han Chinese and the Yaoyi at K=15 is a surprise.

    My expectation would be that at some greater K level than 15 that the Chinese Han component which is basically “North East Asian” would separate out into a Northern “Barabarian to China” component that would include Yakut, Yaoyi and Korean populations on one hand, and a Southern “Civilized Han” component.

    If for no other reason, this would be surprising because there is so little linguistic commonality between Japanese and Chinese except in cases where the borrowing in the historic era that can be documented, and there seems to be so little of a Jomon substrate in Japanese. If the Japanese are really 90% Northern Chinese in origin, why isn’t the Japanese language a clearly Tibeto-Burmese language?

    Answering my own question, could it be that Japanese is the closest to archiac North Chinese and that Tibeto-Burmese was a replacement language imposed by a Tibetan elite with little genetic impact (a bit like Hungarian in Hungary) that had not yet wiped out its mainland cousins at the time of Yaoyi migration? Were the Yaoyi in Japan a bit like the Republican Chinese in Taiwan, but two thousand years earlier, i.e. an exiled political class of the old regime utterly defeated by the new regime that migrated out en masse?

  14. There is a nice summary of mtDNA and Y-DNA found here.

    Two Y-DNA haplogroups that are unique to Japan, which argues strongly for their origin not being from Yaoyi immigration to Japan in the last 2,000 years are D2 (36%) and C1 (4%). The antiquity of D2 is also attested by its near ubiquity in the Ainu. Y-DNA haplogroup O makes up 54% of Japanese men and is almost surely not Jomon. There are multiple stories that could be told about N and C3 which together make up about 6% of the population.

    Definitively Jomon mtDNA types (M7 and N9) make up 22% of the Japanese population, about 67% of the mtDNA haplogroups are typical East Asian ones, and perhaps 9% of the mtDNA haplogroups are ones about which one could offer differing stories associated with Siberia. The antiquity of N9 is supported by ancient DNA evidence.

    FWIW, I’m inclined to think that the ambiguous 6% of Y-DNA and 9% of mtDNA each associated with Siberia may be a Northeast Asia superstate component and linguistic source for Japanese that ruled an ethnically Han Chinese (x Southern Chinese component) subject class within the Yaoyi.

    A naive estimate of patrilines and matrilines would suggest an expectation that Japan has roughly 33% overall Jomon ancestry, with the percentages surprisingly (given that Jomon is the substrate and not superstate population) higher in patrilines than matrilines.

    The difference might be suggestive of widespread superstrate polygamy/concubinage in early Yaoyi culture with East Asian imported wives (perhaps 30% of Yaoyi men might have had more than wife) in a colonization scenario, rather than local wives predominating as they do in many conquest scenarios. Alternately, some mtDNA lines assigned to East Asia might also have been present in the Jomon, in which case the predicted Jomon ancestry from uniparental markers would be even higher.

    But, the autosomal estimates from K=15 at Dienekes are eyeballing it, suggestive of a much, much lower Jomon component. It is really hard to reconcile the uniparental and autosomal data here, and the uniparental markers can be affiliated with Japan or other places respectively, in a much more definitive way than the 15 dimensions utilized by Dienekes, because it is finer grained and can be considered in a phylogenetic way that can be more readily compared to ancient DNA. To get that kind of autosomal result and the mtDNA and Y-DNA to reconcile requires some serious population history gender-ethnicity interaction dynamics gymnastics.

    When you are talking about entirely different species interacting at arms length (e.g. the fact that we have 1%-4% Neanderthal DNA but no Neanderthal Y-DNA or mtDNA) it is possible to come up with some plausible theories.

    But, when you end up elevating the percentage of Y-DNA from the substrate Jomon population relative to the autosomal populations, you have to get really creative and since almost all of the two thousand years in question is accompanied by written history to at least some extent, we can’t let our imaginations get too wild in deviating from that history to come up with mechanisms to produce that result.

    If the data really are all right, the best I can come up with is that the Yaoyi come in, take charge, import a lot of women, and then a Jomon insurgency ends up subjegating the invading population and causing Jomon patrilines and matrilines to surge above their automsomal levels. This is a strikingly different story than the one told by the autosomal data alone.

    It seems much more plausible that something in the autosomal analysis by Dienekes (or more accurately by the software he is using) is misclassifying Jomon automsomal markers as Han when they are actually found in both, perhaps because there aren’t enough dimensions being analyzed, or perhaps because the software isn’t designed to handle a scenario with overlapping ancestry profiles.

    This might also be a function of a smaller and less representative sample. Other autosomal studies from Japan have shown considerable regional variation: “Genome-wide SNP genotypes of the Japanese show that people from northern Japan (Tohoku and Hokkaido) are the most distantly related to the Han Chinese, while those from Western Japan (Kyushu, Chūgoku, Kinki) are the closest. This confirms the theory of the continental Yayoi invasion from Kyushu. Okinawan people were shown to be a clearly distinct ethnic group, falling in a separate genetic cluster from other Japanese.”

    The image of fishers like the Jomon being trounced by intensive farmers also is somewhat inconsistent with the history of European farmer expansion in which fishers had the most staying power vis farmers, while land based hunter-gatherers were the ones who got trounced, that was noted in an earlier post at this site.

  15. ohwilleke, it’s not just dienekes. we’ve had japanese and chinese in hapmap stuff for a while now. two-way admixture has a distinctive look in PCA & structure/admixture/frappe. i know all about the uniparental stuff (i’ve blogged in the early 2000s), so this has always bothered me. contrary to your assertion i think we need to be really careful about assigning/inferring from Y/mtDNA. there have been too many obvious mistakes of inferences in situations where we have ancient DNA. there is no ambiguity of the close relationship of the koreans to the japanese.

    anyway, not interested in having a discussion about uniparental vs. autosomal with you. you obviously are attached to inference from uniparental lineages. i’m not so much.

  16. Spike Gomes


    The best agricultural lands in Japan are the coastal plains. Farmers and fishers would be in immediate conflict, especially over freshwater usage.

    Also, I doubt there’s much interplay of Japan’s colonial history leaving much of a mark genetically on the nation. Family records go waaaaaay back. If someone can still encounter difficulty from ancestors who once lived in a burakumin village 300 years ago, someone with relatively recent Korean ancestry, much less non-Asian ancestry, even more so.

    Frankly, I would like to see the Okinawans and moreso the more remote Ryukyuan genetic lineages looked at. Now that would be interesting. There’s a pretty strong phenotypical difference between Japanese and Ryukyuans. The ancedotal way people here in Hawaii sort ’em out is if you meet someone with a Japanese name and full beard, they’re Okinawan.

  17. MNA

    It’s important to remember that many studies in Japan has established that the Yayoi have dual origins. Recent carbon dating methods done by the Japanese National History Museum have established that the origins of the Yayoi period was actually much earlier than previously assumed. (no english translation, but the site is here: (

    Other studies to keep in mind: DNA analysis from several skeletons unearthed in the early Yayoi period from Fukuoka (Japan) and Jiangsu (China) actually closely match each other. Further proof of their linkage is that both skeletons exhibited ritualistic pulled teeth in addition to the DNA evidence.

    Third, more research done by the Japanese National History Museum has found through DNA analysis of rice grains used by the early Yayoi over 2300 years ago was of a type which did not exist in the Korean Peninsula, further damaging the theory that the Original Yayoi came from the Korean Peninsula. link here, but again no English translation available, (

    I think there can be little doubt that there were subsequent waves of immigration from the Korean Peninsula that contributed, and probably overwhelmed, the genetic makeup of the Yayoi *LATER*, but the identity of the *ORIGINAL* Yayoi is still pretty much in doubt. There are lots of studies that point to an actual origination from China, rather than Korea. Most of these studies, unfortunately, are not published in English but are available in Japanese only. Just wanted to say that it’s far from proven who the Original Yayoi are, the debate is still raging in Japan.

  18. Lassi Hippeläinen

    “the connection/likely origin of austronesians in taiwan is pretty well established. probably nothing to do with japan”

    Do you mean that the component you labelled as Melanesian is strictly Melanesian, and excludes other Pasific Islanders? Taiwan is so much closer to Japan than Melanesia…

  19. Karoly

    “In other words, the modern Japanese are predominantly descendants of farmers from the Korean peninsula who arrived ~2,000 years ago.”

    A Japanese friend who speaks quite a few languages and has recently been learning Korean told me that Korean grammar is surprisingly similar to the Japanese one while the vocabulary is completely different. With some exaggeration, a Japanese only needs to replace the words in his Japanese sentences to speak Korean.
    It would be interesting to know how the vocabulary got replaced or how the grammar was imposed on another language.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar