MYS0CZKZ4KE5BD4·Mystery · object

Who were the Denisovans? The Harbin "Dragon Man" cranium

Thilo Parg · CC BY-SA 3.0

First evidence: Mitochondrial genome of a finger bone from Denisova Cave, Altai, 2010⁠[1]
The cranium: Harbin cranium HBSM2018-000018(A), Geoscience Museum of Hebei GEO University⁠[8]
Found: Reportedly 1933, building the Dongjiang Bridge over the Songhua River, Harbin⁠[6]
Age: At least 146,000 years (U-series on the bone); source layer 138-309 ka⁠[6]
Named: Homo longi, Ji and Ni 2021⁠[8]
Identified as Denisovan: 18 June 2025: 95 proteins with three Denisovan variants; mtDNA from dental calculus⁠[9]
Latest: Xujiayao proteins show the Denisovan COL1A2-996K variant (Nature, 21 October 2026)⁠[12]
Status: Partly explained
Questions: 2 of 4 answered

People Qiaomei FuPalaeogeneticist, IVPP Beijing; led the Harbin and Xujiayao molecular work Ji QiangPalaeontologist, Hebei GEO University; named Homo longi Ni XijunPalaeontologist, IVPP and Hebei GEO University; co-named Homo longi Svante PääboGeneticist, MPI-EVA; senior author of the 2010 Denisovan papers

Catalog keyMYS0CZKZ4KE5BD4

In 2010 a fragment of a child's finger bone from Denisova Cave in the Altai Mountains yielded a mitochondrial genome unlike that of any human or Neanderthal⁠[1]. Its nuclear genome showed a sister group of Neanderthals, which the authors called Denisovans, and showed that they had interbred with the ancestors of living Melanesians⁠[2]. Denisovan DNA was later traced in Papuans, in East and South Asians, and in the high-altitude variant of EPAS1 carried by Tibetans⁠[3]⁠[4]⁠[5]. Yet the people themselves were known from a few teeth and scraps of bone. A population that had left its DNA across Asia and Oceania had no face.

Meanwhile East Asia held a set of large, robust crania that fitted no named species well: Dali, Jinniushan, Hualongdong, Xujiayao, and the most complete of all, a cranium reportedly found in 1933 during the building of a bridge in Harbin and first described in 2021⁠[6]⁠[7]. Its describers named it Homo longi, the "Dragon Man", and placed it in a lineage they saw as the sister group of Homo sapiens⁠[8]. Several researchers suspected these were the Denisovans, but no DNA or protein linked them.

The link came in 2025. Qiaomei Fu and colleagues recovered 95 proteins from the Harbin cranium's petrous bone, with three amino acid variants found only in Denisovans, and, separately, Denisovan mitochondrial DNA from its dental calculus⁠[9]⁠[10]. In September and October 2026 proteins identified five Denisovan specimens at Bianfu Cave in Yunnan and showed that the Xujiayao fossils from northern China were Denisovan too; a comparison of the crania then found a common Denisovan morphology shared by Harbin, Xujiayao, Jinniushan, Dali and Hualongdong⁠[11]⁠[12]. The Denisovans have a face, and it was already in the museum drawers. Of the 4 questions on this site, 2 are answered. The most likely answer to “How deep are the roots of the Denisovan lineage, and how many distinct Denisovan populations were there?”, not yet confirmed: the lineage goes back about a million years and comprised several deeply diverged regional populations,….

Key questionsset by the editors

How deep are the roots of the Denisovan lineage, and how many distinct Denisovan populations were there?Most likely: the lineage goes back about a million years and comprised several deeply diverged regional populations,…

Which fossils are Denisovans, and so what did Denisovans look like?Answered: the large, robust Middle and Late Pleistocene crania of East Asia (Harbin, Xujiayao and their allies) are…

All 4 questions

The object

The Harbin cranium is almost complete and undistorted, though it lacks the mandible. It is massive, larger than any other known archaic human cranium, long and low, without the globular braincase of modern humans, with an endocranial capacity of about 1,420 ml, in the range of modern humans and Neanderthals⁠[8]. It is said to have been found in 1933 when the Dongjiang Bridge was built over the Songhua River and kept out of sight for decades; its history is described by its describers as "long, difficult, and confused"⁠[6]. It entered the collection of the Geoscience Museum of Hebei GEO University as HBSM2018-000018(A)⁠[8].

Because the find spot was lost, the age had to be reconstructed. The cranium's trace elements, rare-earth pattern and strontium isotope ratio match sediments from a core drilled near the bridge and place it in the upper Upper Huangshan Formation, dated by luminescence to 138-309 ka; ten uranium-series analyses on the bone give a minimum age of 146 ka⁠[6].

Harbin cranium (HBSM2018-000018(A)), holotype of Homo longiFu et al. 2025, 1933. CC BY 4.0Artifact

Map of the continentOpen the full map

+−

Denisova Cave, Altai KraiCave

Within 100 m

Baishiya Karst Cave, Xiahe, GansuCave

Within 1 km

Xujiayao, Nihewan BasinOpen-air site

Within 1 km

Heilongjiang Province, ChinaRegion

A region

Bianfu Cave, Jindun, Heqing County, YunnanCave

Within 1 km

Dongjiang Bridge, Songhua River, HarbinFind spot (reported)

Within 10 km

Penghu Channel, Taiwan StraitSeabed

A region

MappedWithin 1 kmA regionNatural Earth

Timeline

Timeline

Squeezed where little happened. Each mark is a published claim, by how it stands:AcceptedMost likelyStill possibleDisputedRuled out

What is known

Denisovans were a sister group of Neanderthals who lived across eastern Asia from at least about 200,000 years ago, from the Altai to the Tibetan Plateau, Taiwan and Yunnan⁠[2]⁠[4]⁠[13]⁠[14]⁠[11]. Molecularly confirmed fossils now include the Denisova Cave teeth and bones, the Xiahe mandible and a rib from Baishiya Karst Cave, the Penghu 1 mandible, the Harbin cranium, two Xujiayao specimens and five specimens from Bianfu Cave⁠[13]⁠[15]⁠[14]⁠[9]⁠[12]⁠[11].

Xiahe mandibleDongju Zhang, 1980. CC BY-SA 4.0Artifact

Penghu 1 mandibleTaichungJohnny. CC BY-SA 4.0Artifact

Their anatomy, as the confirmed fossils show it, combines traits reminiscent of Homo erectus, such as a thick cranial vault, with some seen in modern humans, such as a strongly arched temporal squama, and some shared with late Neanderthals, such as the structure of the inner ear and a suprainiac fossa. Distinctive to them are a low cranial profile with near-parallel side walls, a massive brow, a flat cheek region, weak facial projection, large teeth and robust jaws⁠[12]⁠[14]. Brains were large; the Harbin endocranial volume is about 1,420 ml⁠[8].

Genetically they were not one population. Living people carry DNA from at least three Denisovan groups, and Oceanians and South Asians share one that diverged deeply from the Altai Denisovans⁠[4]⁠[3]. A 200,000-year-old Denisovan from the Altai belonged to a group later replaced by other Denisovans⁠[4].

What has been tested

Two independent molecular tests were run on the Harbin cranium. Palaeoproteomics of two petrous-bone samples recovered 95 endogenous proteins, the best-preserved hominin proteome yet; of 122 single amino acid polymorphisms informative within hominids, three are derived variants known only in Denisovans, and phylogenetic analysis groups the cranium with Denisova 3⁠[9]. No DNA survived in the tooth or the petrous bone, but mitochondrial DNA was captured from dental calculus; it falls within Denisovan variation, on the branch of the early Denisovans at Denisova Cave⁠[10]. Calculus can contain DNA from other sources, so the agreement of the two tests matters: proteins from the bone itself and DNA from the calculus point the same way⁠[9]⁠[10].

Denisova 3 finger phalanxThilo Parg, 2008. CC BY-SA 3.0Artifact

The same test, the Denisovan-specific COL1A2-996K variant, was applied to the Xujiayao bones and found in both specimens; the variant is present in every Denisovan with protein or genome data and absent in Neanderthals and modern humans, who carry arginine at that site⁠[12]. The Bianfu specimens were found by screening 60,000 fragments by peptide mass fingerprinting before sequencing⁠[11]. The extension of the label to Dali, Jinniushan and Hualongdong rests on shared morphology, not on molecules⁠[12].

The readings

Before 2025 there were three main readings of the East Asian crania. One placed Harbin, Dali, Jinniushan and Hualongdong in a new lineage, Homo longi, sister to H. sapiens, and suspected but could not show that the Denisovans belonged to it⁠[7]⁠[8]. A second split them: H. longi for Harbin, Dali and Jinniushan, and a separate species, Homo juluensis, for the large-brained Xujiayao and Xuchang crania and the Denisovan-identified Xiahe and Penghu jaws⁠[16]. A third treated them as regional archaic H. sapiens or as variants of H. heidelbergensis, with no special link to the Denisovans⁠[12].

The molecular results decide between these. Harbin is Denisovan⁠[9]⁠[10]; Xujiayao, the holotype of H. juluensis, is Denisovan too⁠[12]. The two proposed species are therefore one lineage, and it is the Denisovan one. The Yunxian 2 cranium, about a million years old, has been placed as an early member of the same clade, which would give the lineage roots beyond the Middle Pleistocene⁠[17]. What remains is a question of names and of boundaries: which fossils without molecular data belong, and whether the genetic subgroups correspond to anatomical ones⁠[12]⁠[4].

Published claims

Published claims and their status

Year	Who	Claimed	Status	Reason
2010	Johannes Krause and colleagues⁠[1]	The finger bone from Denisova Cave carries the mitochondrial genome of an unknown hominin, whose mtDNA line split from those of modern humans and Neanderthals about a million years ago.	Accepted	The nuclear genome confirmed a new group, though it showed the population to be sister to Neanderthals, so the deep mtDNA line is a discordance rather than the date of the split⁠[2]
2010	David Reich and colleagues⁠[2]	The Denisovans were a sister group of the Neanderthals who contributed about 4 to 6 per cent of the genomes of present-day Melanesians.	Accepted	The high-coverage genome of 2012 and every later study of living people's ancestry have borne it out⁠[18]⁠[3]
2018	Viviane Slon and colleagues⁠[19]	Denisova 11 was the daughter of a Neanderthal mother and a Denisovan father.	Accepted	The genome shows it directly, and no one has challenged the reading⁠[19]
2019	Fahu Chen and colleagues⁠[13]	The Xiahe mandible from the Tibetan Plateau is Denisovan and at least 160,000 years old.	Accepted	Denisovan mtDNA in dated sediments from the same cave, and a second Denisovan bone from it, support the attribution⁠[20]⁠[15]
2019	Guy Jacobs and colleagues⁠[3]	Papuans carry two distinct, deeply diverged Denisovan ancestries, so there were several Denisovan populations.	Accepted	The 200,000-year-old genome finds at least three Denisovan groups behind living people's ancestry⁠[4]
2021	Ni Xijun and colleagues⁠[7]	The Harbin cranium belongs to a lineage, with Dali, Jinniushan and Hualongdong, that is the sister group of Homo sapiens.	Disputed	Harbin is now Denisovan by its proteins and mtDNA, and the genomes place Denisovans beside Neanderthals, though the 2025 Yunxian analysis repeats the morphological result⁠[9]⁠[2]⁠[17]
2021	Ji Qiang and colleagues⁠[8]	The Harbin cranium is the holotype of a new species, Homo longi.	Still possible	The name is valid and now attaches to a Denisovan, but whether Denisovans should carry a species name is still argued⁠[12]⁠[16]
2022	Fabrice Demeter and colleagues⁠[21]	A molar from Tam Ngu Hao 2 cave in Laos is Denisovan, so the lineage reached tropical South-east Asia.	Still possible	The attribution rests on the tooth's shape; its proteins confirm only the genus Homo⁠[21]
2024	Christopher Bae and Wu Xiujie⁠[16]	The Denisovans belong to a new species, Homo juluensis, separate from Homo longi, which keeps Harbin, Dali and Jinniushan.	Disputed	Both Harbin and Xujiayao, which the scheme puts in different species, carry Denisovan protein variants and share a cranial form⁠[12]
2025	Takumi Tsutaya and colleagues⁠[14]	The Penghu 1 mandible, dredged off Taiwan, is a male Denisovan.	Accepted	Two Denisovan-specific amino-acid variants and a male-specific peptide identify it; its age is less certain⁠[14]
2025	Qiaomei Fu and colleagues⁠[9]	The Harbin cranium's proteins carry Denisovan variants and cluster it with Denisova 3, so it is Denisovan.	Accepted	The mtDNA from its dental calculus points to the same lineage independently⁠[10]
2025	Qiaomei Fu and colleagues⁠[10]	Mitochondrial DNA from the Harbin cranium's dental calculus falls within Denisovan variation, on an early branch.	Accepted	No DNA survived in the bone, so the result rests on the calculus alone, but it agrees with the proteome⁠[9]
2025	Feng Xiaobo and colleagues⁠[17]	The million-year-old Yunxian 2 cranium is an early member of the Homo longi clade, which has roots more than a million years deep.	Still possible	It rests on the morphometrics of a digitally restored, crushed skull, with no molecules to test it⁠[17]
2025	Stéphane Peyrégne and colleagues⁠[4]	A 200,000-year-old Denisovan genome shows an early group replaced by later Denisovans, and at least three Denisovan groups behind living people's ancestry.	Most likely	It fits the deep structure seen in Papuan genomes but is a preprint not yet through peer review⁠[4]⁠[3]
2026	Xing Song and colleagues⁠[12]	Two Xujiayao bones carry a Denisovan-specific protein variant, and Xujiayao, Harbin, Jinniushan, Dali and Hualongdong share one Denisovan cranial form.	Accepted	The protein result is direct; the shared form extends the attribution to Dali, Jinniushan and Hualongdong on morphology alone⁠[12]
2026	Rao Huiyun and colleagues⁠[11]	Five hominin specimens from Bianfu Cave in Yunnan, of about 167,000 to 134,000 years ago, are all Denisovan by their proteins.	Accepted	The proteomic identifications are backed by a separate study of the cave's layers and dates⁠[22]

What would settle it

The identity question is answered. The open edges are narrower. A nuclear genome, or more proteins, from any of Dali, Jinniushan or Hualongdong would test the morphological extension of the label. Proteins or DNA from Yunxian would test whether the clade really reaches back a million years. Genomes from the southern fossils, such as Penghu, Bianfu or Tam Ngu Hao 2, would show whether the deeply diverged Denisovan ancestry in Oceanians and South Asians came from a southern population with its own anatomy. The name is a matter for zoological nomenclature: if the Harbin and Xujiayao fossils are one species, Homo longi, published first in 2021, would normally have priority over H. juluensis, unless the lineage is kept informal.