Ancient DNA | Jomon, Yayoi, Kofun | Japan and the Ryukyus

Japan is made of three peoples, not two

Every Japanese schoolbook teaches the same story: hunter-gatherers met rice farmers, and the result was the Japanese. Ancient genomes have now added a third ancestor nobody had counted, found people of the farming era who carried no farmer ancestry at all, and measured how recently Okinawa stopped being Jomon.

19% Jomon ancestry in modern Japanese in our Global25 model, against 15.0% published by Cooke and colleagues
96% Jomon ancestry in two people buried in the Yayoi period, the era of the continental migration
99% to 21% the fall in Jomon ancestry in the Ryukyu islands between 2,800 years ago and today

Hero image: AI generated illustration of a Yayoi period village in northern Kyushu, with rice being transplanted, a raised granary on stilts and a bronze bell, and people of Jomon and continental descent working the same field. It is not a reconstruction of a documented site and is not a portrait of any real person.

In 1991, the anthropologist Kazuro Hanihara published a model so clean that it became the standard account of Japanese origins for a generation. He called it the dual structure hypothesis. The Japanese, he proposed, are the product of two populations: the Jomon, hunter-gatherers and potters who had been in the archipelago since the Ice Age, and migrants who arrived from the Korean peninsula in the Yayoi period bringing wet rice agriculture. The Ainu of Hokkaido in the north and the Ryukyuans of Okinawa in the south kept more of the first; everyone in between was a mixture.

Thirty years later, three studies have taken that model apart and put it back together with an extra piece.

Fifteen thousand years alone

The Jomon are one of the strangest populations in human genetics. Their ancestors separated from other East Asians somewhere between 15,000 and 20,000 years ago, around the time rising seas cut the archipelago off from the mainland. They made pottery from about 16,500 years ago, some of the oldest in the world, and they went on making it, and living by hunting, fishing and gathering, for more than ten millennia, while farming societies rose everywhere around them.

Genetically, they barely moved. Cooke and colleagues estimate that the Jomon kept an effective population size of around 1,000 for thousands of years. We can see the consequence directly in Global25. The distance between an Initial Jomon individual from Shikoku who died around 6870 BC and a Late Jomon woman from Rebun island who died around 1805 BC, five thousand years apart and at opposite ends of the archipelago, is 0.0272. The distance between Koreans and the Han of Shandong today, across a sea, is 0.0313. Five thousand years of Jomon history produced less genetic change than you find between two neighbouring modern populations.

The third ancestor

In 2021, Niall Cooke, Shigeki Nakagome and colleagues sequenced twelve new ancient Japanese genomes, nine Jomon and three from the Kofun period, and tested the dual structure model directly. It failed. Two sources were not enough.

What the data required was three: the Jomon, a Northeast Asian source resembling populations of the West Liao River region, and an East Asian source resembling the Han. The Northeast Asian component came with rice farming in the Yayoi period. The East Asian component arrived later, in the Kofun period, the age of the keyhole tombs and of the first centralised state, between roughly the third and seventh centuries of our era.

Their numbers: Kofun individuals carry 13.1% plus or minus 3.5% Jomon ancestry, and modern Japanese 15.0% plus or minus 3.8%. The ratio between the two continental components flips across the two periods, from about 1.9 to 1 in favour of Northeast Asian in the Yayoi, to about 1 to 2.5 in favour of East Asian in the Kofun. Something large happened during the formation of the Japanese state, and it was not only political.

In 2024, Xiaoxi Liu, Chikashi Terao and colleagues at RIKEN sequenced 3,256 whole genomes from across Japan and found the same three components still legible in living people, with a geography. Jomon ancestry is highest in Okinawa at 28.5%, intermediate in Tohoku in the northeast at 18.9%, and lowest in Kansai at 13.4%. The East Asian component is strongest in the west, nearest the continent, and fades eastward. The dual structure model was not wrong so much as incomplete.

Our Global25 analysis

We modelled every ancient and modern population of the archipelago available in Global25 coordinates as a mixture of three sources: a high coverage Jomon genome, a Bronze Age West Liao River population for the Northeast Asian pole, and the Han of Shandong for the East Asian pole. The three sources are well separated, at 0.303, 0.199 and 0.217, and every fit has a condition number of 7.0.

1. First, does the test work?

Before reporting any percentage we checked the estimator against known answers, by adding measured amounts of Jomon ancestry to continental genomes.

Validation: known amounts of Jomon ancestry added to continental genomes 0% 4% 8% 12% 16% 20% 24% 28% 32% 0% added 1.4% 2% added 3.0% 5% added 5.9% 10% added 10.8% 15% added 15.7% 20% added 20.5% 30% added 30.2% Mean over 15 synthetic mixtures (5 continental hosts x 3 Jomon sources). Below 5% the estimate is unreliable.
Figure 1. Validation of the Jomon estimator. A 30% spike-in reads back as 30.2%, a 10% spike-in as 10.8%, and a 5% spike-in as 5.9%. With no Jomon ancestry at all the model still returns 1.4% on average, so our honest detection floor is about 5%.

Real controls agree. The Han of Shandong return 0.0% Jomon, the Han of Fujian 0.7%, Koreans 3.4%. Two warnings, though. Koreans of Jeju island return 10.4%, and the Ulchi of the Amur river return 19.8%, as much as a modern Japanese. The Ulchi figure is not Jomon ancestry: it is the Jomon pole absorbing an ancient Siberian signal that our three sources do not otherwise cover. The estimator is trustworthy for populations of the archipelago and should not be pointed at continental Northeast Asia.

2. Three layers, from the Ice Age to the present

The three ancestries, from the Jomon period to today Jomon Northeast Asian East Asian Jomon, Honshu, Early (control) 98% Okinawa, Nagabaka, c. 2800 BP 99% Yayoi, Kuma-Nishioda 27% Kofun, Honshu, AD 571 to 625 18% Okinawa, Nagabaka, c. 1800 35% Ainu, Sakhalin 69% Ryukyuan, today 21% Japanese, today 19% Korean (control) 3% Han Shandong (control) 0% Each bar sums to 100%. The number on the right is the Jomon share. Condition number 7.0 for every fit.
Figure 2. The three ancestries across periods and populations. The Jomon control and the 2,800 year old Okinawans sit at the top because they are almost unmixed; the controls at the bottom because they carry no Jomon ancestry.

Modern Japanese come out at 19.3% Jomon, 11.5% Northeast Asian and 69.2% East Asian. Our Jomon figure is four points above the published qpAdm estimate, which is the expected direction for a Global25 model that has to put every leftover somewhere, and we checked that it is not an artefact of our choice of sources: across seven different Jomon genomes the answer moves between 19.1 and 19.8%, and across four different Northeast Asian poles between 16.5 and 20.0%.

The Kofun individuals of Honshu come out at 18.4% Jomon, within a point of living Japanese. Measured as a distance, modern Japanese sit 0.0127 from the Kofun of the sixth century and 0.0334 from modern Koreans. Fourteen centuries of history, including several more waves of recorded migration, have changed the genetic profile of mainland Japan remarkably little. Whatever made the Japanese was essentially finished by the time the keyhole tombs were being built.

The Ainu are the other pole: 69.0% Jomon, with the remaining 31.0% Northeast Asian rather than East Asian, which fits their history of contact with the Okhotsk culture of the northern coasts rather than with the rice farming world to the south. Japan recognised the Ainu as an indigenous people in law in 2019.

3. The Yayoi were not a replacement

The most interesting result is not about averages. In January 2026, Jonghyun Kim, Fuzuki Mizuno, Jun Ohashi and colleagues published four new genomes from the Yayoi period islands off northwestern Kyushu, exactly where the continental migrants would have landed first. We ran each of them separately.

Jomon ancestry in five people buried during the Yayoi period 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Neshiko 16, Hirado, 281 BC 96.8% Neshiko 13, Hirado, 270 BC 96.1% Neshiko 11, Hirado, 146 BC 60.2% Shomura, Iki, 345 BC 44.5% Kuma-Nishioda, Kyushu (n=2) 27.2% Same three source model. These people were contemporaries, in neighbouring places, on the migration front.
Figure 3. Jomon ancestry in five people buried during the Yayoi period. Neshiko 13 and Neshiko 16, from Hirado island, are statistically indistinguishable from unmixed Jomon despite living in the farming era, and may be third degree relatives.

Two of them, Neshiko 13 and Neshiko 16, return 96.1% and 96.8% Jomon. They were buried in the middle Yayoi period, in the third and second centuries BC, on an island within sight of the continental route, and they carry no detectable continental ancestry. Their contemporaries do: Neshiko 11, from the same site but a century later, is at 60.2%; the man from Shomura on Iki island, at 44.5%; the Kuma-Nishioda pair from Kyushu, at 27.2%.

The original authors report the same picture, and add a detail that our autosomal models cannot see. Neshiko 11 carries a Jomon mitochondrial lineage and a Jomon Y chromosome while being 40% continental in the rest of his genome, and the Shomura man carries a continental mitochondrial lineage. Individual families, not populations, were mixing, and they were mixing in both directions. The authors date the start of this gene flow to about 2,500 to 2,600 years ago.

This is what the dual structure model always implied but could not demonstrate: the Yayoi transition was not an invasion that swept the Jomon away. For several centuries, on the same small islands, there were people who had mixed and people who had not.

4. Okinawa, across 2,800 years

Jomon ancestry in Okinawa across 2,800 years 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Nagabaka, c. 2800 BP 98.8% Nagabaka, c. AD 1800 34.6% Ryukyuan, today 20.6% Mainland Japanese, today 19.3% The Ryukyu islands kept an almost unmixed Jomon profile for far longer than the mainland, then changed quickly.
Figure 4. Jomon ancestry at Nagabaka, on Miyako island in the Ryukyus, at three points in time, against the mainland Japanese average today.

The Ryukyu islands give us something rare: the same cemetery, sampled across almost three thousand years. The two people buried at Nagabaka between about 800 and 680 BC come out at 98.8% Jomon, as unmixed as the Jomon of the mainland two thousand years earlier. Their successors in the same islands around 1800 of our era are at 34.6%. Ryukyuans today are at 20.6%, barely above mainland Japanese.

The implication is that the Ryukyus stayed outside the continental gene flow for most of the period when the mainland was being transformed, and then caught up fast. Our historic sample of three individuals and single modern Ryukyuan average are thin evidence for the timing, and the 28.5% that Liu and colleagues report for Okinawa from 3,256 modern genomes is a better figure for living Ryukyuans than our 20.6%. But the direction and the scale of the change are not in doubt.

5. What Global25 cannot do here

One trap is worth showing, because anyone running these models will hit it.

Northeast Asian pole Separation from the East Asian pole Result for modern Japanese
West Liao River, Bronze Age 0.217 19.3% Jomon, 11.5% Northeast Asian, 69.2% East Asian
Haminmangha, Middle Neolithic 0.168 18.1% Jomon, 12.0% Northeast Asian, 69.9% East Asian
Devil's Gate Cave, Neolithic 0.144 17.0% Jomon, 19.8% Northeast Asian, 63.2% East Asian
Modern Koreans 0.031 14.9% Jomon, 85.1% Korean, 0.0% East Asian, condition number 36.2
Table 1. Using modern Koreans as the Northeast Asian source destroys the model. Koreans and the Han of Shandong are only 0.031 apart in Global25, so the two sources are nearly the same vector and the fit assigns everything to one of them. The result looks precise and means nothing.

This matters beyond the technical point. A large part of what circulates online about Japanese and Korean genetics consists of models built on sources that are too close together to be distinguished. If you see a model that splits Japanese ancestry between modern Koreans and modern Chinese and reports two decimal places, the decimals are noise.

The second limit is geography. Global25 gives us one modern Japanese average, so we cannot reproduce the regional gradient that Liu and colleagues measured across 3,256 individuals. For anyone asking whether their own family from Tohoku carries more Jomon ancestry than a family from Kansai, that study, not ours, is the answer: 18.9% against 13.4%.

What this changes

Three practical things follow.

First, the number most people quote for themselves is wrong in both directions. Jomon ancestry in a mainland Japanese person is not half and not a trace: it is around a sixth to a fifth of the genome, and it varies by region, from about 13% in Kansai to about 19% in Tohoku, and 28.5% in Okinawa.

Second, the Jomon inheritance is much more visible on the Y chromosome than in the genome as a whole. The lineage D-M55 is carried by roughly a third of Japanese men, reaching about 88% in Ainu samples and about 38% in Okinawa, and it is almost absent everywhere else on earth, at a few percent in Korea and rare elsewhere. A Japanese man with that lineage has an unbroken male line back to the hunter-gatherers of the Ice Age archipelago, whatever the rest of his genome says.

Third, the idea of Japan as a uniquely homogeneous nation, which has done a great deal of political work over the last century, does not survive contact with its own ancient DNA. The archipelago is the product of at least three populations, one of which spent fifteen thousand years in near isolation, arriving at different times and mixing at different rates, with Hokkaido and the Ryukyus on very different schedules from the centre. That is not an unusual story. It is the ordinary human story, which is the point.

Methods note

All coordinates are Global25 scaled coordinates. Population averages were taken in priority from the Global25 ancient averages, then the Global25 modern averages, then the Moriopoulos 2026 collection, then the dated individual ancient file. The three source model uses Japan_Jomon.SG (the IK002 genome, dated about 832 BC, 81% coverage) as the Jomon pole, China_WLR_BA_o as the Northeast Asian pole and Han_Shandong as the East Asian pole; separations are 0.303 between Jomon and Northeast Asian, 0.199 between Jomon and East Asian and 0.217 between the two continental poles, giving a condition number of 7.0 for every fit. NNLS fits use a sum to one constraint applied as a weighted extra equation, weight 1000, followed by normalisation. The Jomon pole was varied across seven alternatives (range 19.1 to 19.8% for modern Japanese) and the continental poles across eight combinations (range 16.5 to 20.0%). Spike-in validation crossed five continental hosts with three Jomon sources at 0, 2, 5, 10, 15, 20 and 30%. Yayoi and Okhotsk individuals were modelled one by one, all at coverage above 97% except the Kuma-Nishioda average, whose fit distance of 0.0808 is the worst in the article and whose figure should be treated as indicative only. Published proportions, dates, effective population sizes and haplogroup frequencies are taken from the cited studies and are not recomputed here.

Global25 coordinates used in this article

The Jomon, Yayoi, Kofun, Okhotsk and Ryukyu genomes used above, the modern populations and the three sources. They can be pasted directly into the Vahaduo tool.

Jomon_Initial_Shikoku_JpKa6904_BC6870,0.007968,-0.332078,-0.059585,0.013889,0.040007,0.009761,-0.002115,-0.001385,0.021066,0.015490,-0.046118,-0.006294,0.009960,-0.006881,-0.024158,-0.010872,0.008345,0.014316,0.010182,-0.007754,0.060893,-0.044391,0.003697,0.003494,-0.096039
Jomon_Honshu_Early_average,0.004325,-0.343249,-0.053627,0.009561,0.040685,0.010151,0.001974,-0.001939,0.021598,0.017385,-0.049074,-0.005605,0.006541,-0.006771,-0.021091,-0.016309,0.003390,0.010515,0.004600,-0.008254,0.068654,-0.046864,0.009687,0.008314,-0.100279
Jomon_Honshu_MidLate_average,0.003415,-0.343249,-0.057699,0.011628,0.040469,0.010040,0.001762,-0.001615,0.024338,0.020228,-0.047742,-0.008018,0.009589,0.001445,-0.023141,-0.012463,0.004042,0.006651,0.007793,-0.009005,0.062453,-0.044392,0.010538,0.005000,-0.102685
Jomon_Funadomari_Late_FUN23_BC1805,0.009106,-0.342233,-0.055814,0.009044,0.033545,-0.001116,-0.003995,0.000923,0.021884,0.015672,-0.049691,-0.002847,0.000743,-0.007982,-0.024565,-0.016176,-0.000652,0.008108,0.005782,-0.003252,0.067381,-0.042413,0.010106,0.006507,-0.101907
Jomon_IK002_BC832,0.011382,-0.338171,-0.050911,0.014535,0.040315,0.009203,-0.006110,-0.001615,0.020657,0.017312,-0.052939,-0.001649,0.009217,-0.010459,-0.023344,-0.012596,0.004433,0.010515,0.007165,-0.013757,0.071748,-0.042166,0.015159,0.010724,-0.098434
Okinawa_Nagabaka_BC800_average,0.010813,-0.338679,-0.049025,0.010497,0.038315,0.011435,-0.008460,-0.000116,0.026077,0.021868,-0.047336,-0.003072,0.010258,-0.013555,-0.018662,-0.014453,0.008214,0.010388,0.011313,-0.010818,0.066383,-0.048595,0.009182,0.005724,-0.105021
Yayoi_Shomura_Iki_BC345,0.019350,-0.414336,-0.001508,-0.034561,0.032621,0.011992,-0.005405,0.001846,0.017589,0.013850,-0.063656,-0.006594,0.004162,0.002477,-0.014115,-0.008618,-0.001825,0.005068,0.006285,-0.006128,0.043922,-0.025472,0.008258,0.000482,-0.057719
Yayoi_Neshiko16_Hirado_BC281,0.005691,-0.342233,-0.050157,0.005491,0.042162,0.020359,-0.006110,-0.004615,0.021066,0.020228,-0.049204,-0.005095,0.009514,-0.007432,-0.024294,-0.020154,0.002608,0.009375,0.009930,-0.006253,0.066508,-0.048596,0.008504,0.011568,-0.096758
Yayoi_Neshiko13_Hirado_BC270,0.013659,-0.342233,-0.042992,0.012274,0.035083,0.016176,-0.005170,-0.001615,0.018612,0.020410,-0.050016,-0.000150,0.005798,-0.014038,-0.022530,-0.015778,0.005476,0.007601,0.001760,-0.005002,0.064137,-0.049956,0.012571,0.010363,-0.096878
Yayoi_Neshiko11_Hirado_BC146,0.014797,-0.389963,-0.013576,-0.025517,0.027082,0.005578,0.003055,0.000231,0.012885,0.009112,-0.060084,-0.004796,0.017691,-0.005918,-0.012893,-0.008884,0.002477,0.005574,-0.003520,-0.002501,0.046418,-0.032026,0.004683,0.004097,-0.075921
Yayoi_KumaNishioda_Kyushu_n2,0.006829,-0.432107,0.019233,-0.058301,0.024928,-0.000140,-0.011751,0.020538,-0.009102,0.021868,-0.064387,0.011165,0.003494,-0.018304,0.009365,-0.010806,0.007888,0.009122,-0.034944,-0.010568,0.035125,-0.019413,-0.031798,0.026510,-0.050056
Kofun_Honshu_AD571_625_average,0.018591,-0.443786,0.010685,-0.056417,0.033032,0.014409,0.000235,0.001846,-0.004431,0.006136,-0.068636,-0.006394,0.011645,-0.004679,-0.007645,-0.005922,-0.003260,0.003209,0.001173,-0.008087,0.028990,-0.015828,0.002711,0.002932,-0.035446
Okhotsk_Hokkaido_AD1107_NAT002,0.028456,-0.418398,0.067882,-0.022933,-0.071090,-0.035419,-0.003760,0.008538,0.012476,0.017859,-0.030367,0.002098,-0.001784,-0.004266,-0.017644,-0.012994,0.002868,0.011909,0.025517,0.015507,0.024083,-0.045504,-0.017378,0.006386,-0.014969
Okinawa_Nagabaka_historic_AD1800_average,0.013659,-0.414336,0.005280,-0.042098,0.039802,0.005671,0.001958,0.003539,0.001840,0.012392,-0.063765,-0.006294,0.009663,-0.001743,-0.012984,-0.005878,0.006041,0.006123,0.000042,-0.004710,0.031985,-0.021474,0.007354,0.004780,-0.043270
Ainu_Sakhalin_modern_n1,0.011382,-0.380824,-0.011314,-0.006460,-0.007386,-0.008367,-0.004935,0.009923,0.019839,0.023508,-0.038161,-0.005245,0.002825,0.007707,-0.044381,-0.007425,0.009779,0.010642,0.012570,-0.001501,0.049288,-0.037838,-0.001849,0.004338,-0.055444
Ryukyuan_modern,0.019245,-0.424293,-0.003756,-0.040277,0.027719,0.017559,-0.001636,-0.002556,0.000671,0.013275,-0.084309,-0.010223,0.015069,-0.013042,-0.022376,-0.010988,0.003031,0.002549,0.001722,-0.000126,0.028647,-0.004347,0.008855,0.009556,-0.006030
Japanese_modern_n476,0.018943,-0.442662,0.013179,-0.055994,0.034415,0.009253,0.002746,0.002571,-0.004892,0.009949,-0.074367,-0.008738,0.009687,-0.004181,-0.011531,-0.006185,-0.000776,0.003902,0.001682,-0.007967,0.027601,-0.016179,0.007458,0.003097,-0.039300
Source_NortheastAsian_China_WLR_BA_o,0.019350,-0.468159,0.095789,-0.039729,-0.088632,-0.050200,0.013161,0.026076,0.006954,0.020410,-0.039623,0.004496,-0.008325,0.021469,0.005429,-0.002387,-0.002868,-0.003547,0.013073,0.033266,-0.030197,-0.030542,-0.051641,-0.005663,-0.003832
Source_EastAsian_Han_Shandong,0.022132,-0.448639,0.005363,-0.060114,0.051531,0.021258,0.006189,0.000436,-0.010726,0.003564,-0.073490,-0.008742,0.010291,-0.008762,-0.007842,-0.000044,0.001434,0.001337,-0.001201,-0.008449,0.010967,0.006608,0.011613,0.001058,-0.000492
Reference_Korean,0.022622,-0.455486,0.018157,-0.064129,0.038828,0.011603,0.005097,0.003500,-0.009523,0.006633,-0.075044,-0.008652,0.010128,-0.005809,-0.009008,-0.002497,0.000891,0.001879,-0.000168,-0.007308,0.016234,-0.004364,0.007405,0.000994,-0.017493
Reference_Ulchi_Amur,0.029025,-0.442365,0.082250,-0.035013,-0.065151,-0.042921,0.011022,0.013707,0.011678,0.017604,-0.021581,0.002623,-0.002170,0.000468,-0.008849,-0.006895,0.001278,0.009730,0.022286,0.014770,0.018592,-0.039037,-0.013323,0.007941,-0.005401

References

  1. Ancient DNA Cooke NP, Mattiangeli V, Cassidy LM, Okazaki K, Stokes CA, Onbe S, Hatakeyama S, Machida K, Kasai K, Tomioka N, Matsumoto A, Ito M, Kojima Y, Bradley DG, Gakuhari T, Nakagome S (2021). Ancient genomics reveals tripartite origins of Japanese populations. Science Advances 7(38): eabh2419. doi.org/10.1126/sciadv.abh2419
  2. Population genomics Liu X, Koyama S, Tomizuka K, Takata S, Ishikawa Y, Ito S, Kosugi S, Suzuki K, Hikino K, Koido M, Koike Y, Horikoshi M, Gakuhari T, Ikegawa S, Matsuda K, Momozawa Y, Terao C, et al. (2024). Decoding triancestral origins, archaic introgression, and natural selection in the Japanese population by whole-genome sequencing. Science Advances 10(16): eadi8419. doi.org/10.1126/sciadv.adi8419
  3. Ancient DNA Kim J, Mizuno F, Matsushita T, Matsushita M, Naka I, Kurosaki K, Tokanai F, Ueda S, Ohashi J (2026). Ancient genomes reveal early-stage admixture and genetic diversity in the Northwestern Kyushu Yayoi. Scientific Reports 16: 4833. doi.org/10.1038/s41598-026-34996-7
  4. Review Hudson MJ, Nakagome S, Whitman JB (2020). The evolving Japanese: the dual structure hypothesis at 30. Evolutionary Human Sciences 2: e6. The reassessment of Hanihara's 1991 model. doi.org/10.1017/ehs.2020.6
  5. Population genetics Jeong C, Nakagome S, Di Rienzo A (2016). Deep history of East Asian populations revealed through genetic analysis of the Ainu. Genetics 202(1): 261 to 272. doi.org/10.1534/genetics.115.178673
  6. Y chromosome Frequencies of the Y lineage D-M55 (D1a2a) in Japanese, Ainu and Okinawan samples, compiled from Tajima et al. 2004, Hammer et al. 2006, Nonaka et al. 2007 and Sato et al. 2014. en.wikipedia.org/wiki/Haplogroup_D-M55
  7. Press RIKEN (2024). Genome analysis of 3,256 Japanese individuals reveals three ancestral components and their regional distribution. riken.jp
  8. Data Global25 PCA coordinates by Davidski, and the Moriopoulos 2026 collection of Global25 coordinates.