The official biography of Homo erectus is a story about open ground. For most of a century the species has been drawn striding across savannah and light woodland — a lean, long-legged walker with a handaxe, leaving Africa about 1.9 million years ago, reaching the Caucasus and northern China a couple of hundred thousand years later, and holding out in Java until roughly 108,000 years ago.
When the rainforest closed in, the story went, Homo erectus was finished: dense tropical forest was a habitat only Homo sapiens ever learned to crack. Two papers published in Nature on 18 November 2026 take that sentence apart.
Eight teeth from two caves in southern China, their authors argue, belonged to Homo erectus living under a closed canopy between 324,000 and 184,000 years ago — 100,000 to 200,000 years more recently than the species was thought to have survived anywhere on the Chinese mainland. And for much of that time, the chamber that held them was a porcupine's den.
Baxian and Naxian are sister caves in the karst country of Chongzuo, in the Guangxi Zhuang Autonomous Region — 15 kilometres apart on the same limestone corridor above the meandering Zuojiang River, 13 and 20 kilometres north-west and north of the city, their entrances five to seven metres above the valley floor.
Both end in small rear chambers packed with the same cemented, cataclastic breccia. Together the two deposits have yielded more than 18,000 mammalian fossils; the excavation team was originally hunting something else entirely, the three-metre ape Gigantopithecus. What the breccia also held was nine hominin teeth.
Dating them took five independent clocks — infrared-stimulated luminescence on feldspar and optically stimulated luminescence on quartz from the sediments, uranium-series and coupled uranium-series/electron-spin-resonance dating of teeth, and uranium-series dating of the flowstone that caps the fossil layer. Forty-five radiometric ages came back; forty-three were run through two Bayesian models. The result: Baxian between 324,000 and 184,000 years ago, Naxian between 319,000 and 241,000 — statistically indistinguishable from each other, and a 95 per cent probability that Homo erectus was on this landscape somewhere inside the Baxian window. Until now, no securely dated Homo erectus fossil in mainland China was younger than about 400,000 years.
Identifying the teeth was a separate argument, and it was won on shape before it was won on molecules. Scanned by micro-CT, the buried surface beneath the enamel cap — the enamel–dentine junction — carries a wrinkled, branching pattern that is a signature of Chinese Middle Pleistocene Homo erectus, and the Guangxi teeth show it, in simplified form, alongside short square upper molars, asymmetric lower premolars with a protruding distolingual corner, and a shelf-like protostylid.
Geometric morphometrics put seven of the nine unambiguously inside the Homo erectus range and statistically apart from Neanderthals, Denisovans and modern humans. An eighth, the lower premolar NX-A2-56, fell outside every comparison group as a statistical outlier, but kept the plesiomorphic talonid and the vertical wrinkling of the enamel–dentine junction that mark it as erectus-like rather than something new. Eight, then, of nine.
The ninth argument is the one that will be quoted for years, because it did not exist as a method a decade ago. Enamel is the hardest tissue a human body makes, and it is stuffed with proteins — amelogenin, ameloblastin, enamelin — that survive in cold and even in warm caves long after DNA has dissolved. Five of the teeth gave up original enamel proteomes, with 556 to 808 amino-acid positions read per tooth.
At position 253 of ameloblastin, four of them carry glycine where almost every other hominin carries alanine, the residue of a single codon change on chromosome 4. That substitution is absent from the Neanderthal and Denisovan genomes, from the great apes, from the 1.77-million-year-old Homo erectus of Dmanisi, and from all 1,461,678 living people in the gnomAD database. It has turned up in exactly one other place on Earth: six Chinese Homo erectus teeth around 400,000 years old from Zhoukoudian, Sunjiadong and Hexian, published earlier in 2026.
But at a second position, 273, the Guangxi teeth carry the ancestral methionine while those six carry the derived valine — enough, the authors argue, to place them on their own branch: a population of Homo erectus that no one had ever sampled molecularly. The same extracts also sexed the owners. Four teeth carried amelogenin-Y peptides and belonged to males; the fifth, a premolar, did not, and belonged to a female.
Then there is the tooth that does not fit. PA1796, an upper second molar from Baxian, has a short, rhomboidal enamel–dentine junction sitting between Homo erectus and Neanderthals, taller dentine horns, simpler trigon and talon basins, a more developed and lingually expanded talon — and none of the Homo erectus protein marker. It is not a modern human, not a Neanderthal, and not comfortably an erectus.
The obvious candidate is a Denisovan: Homo longi, the lineage known from the Altai, the Tibetan Plateau, Harbin, the Penghu channel and a cave in Laos, and already present in East Asia in this period. The authors say so — carefully. The evidence is circumstantial, and they know it.
"Every label we put on that ninth tooth is a little bit of speculation," the molecular palaeoanthropologist Frido Welker of the University of Copenhagen told Nature. If it is a Denisovan, the valley at Chongzuo held at least two human lineages at the end of the Middle Pleistocene, not long before the first modern humans appear in the region.
The rainforest claim rests, in the end, less on the hominins than on everything around them. The breccia is dominated by forest animals: orangutans, Asian elephants, the rhinoceros Dicerorhinus, tapirs, muntjac deer, leopards, and the extinct Stegodon and Megalovis — the classic Stegodon–Ailuropoda fauna of southern China, the assemblage in which Homo erectus had never before been found.
Carbon isotopes turn that list into a measurement. Plants under a dense canopy are poor in carbon-13, and the animals that eat them carry the signal into their enamel; converted to plant-equivalent values, the Baxian fauna averages −27.6 per mille and the Naxian fauna −29.4, against a dense-canopy threshold of −27.5. Neither cave shows any trace of C4 grasses.
The frequency of canopy-level values is higher than in equatorial African rainforest and comparable to western Amazonia. Clumped-isotope thermometry on the cave sediments points to humid tropical conditions, wetter and cooler than the region today.
The hominins' own teeth are thinner evidence: only three Homo erectus teeth, all from Naxian, were measured for carbon, and they fall at −15.0 to −14.0 per mille — right on the boundary between closed canopy and open C3 vegetation, exactly where the much larger faunal sample sits, but only three data points.
What the teeth say about movement is the quiet surprise of the study. Strontium isotopes in enamel are fixed by the geology of whatever a child ate and drank, and in the Baxian premolar PA1799 the ratio ranges from 0.7110 to 0.7220 — a spread that cannot come from one hillside. Matched against a strontium isoscape built for the region with machine learning, the highest values point outward from the cave corridor, potentially as far as the lower Red River valley of northern Vietnam.
The two caves, 15 kilometres apart with overlapping dates, look like one population roaming a forested landscape rather than a band pinned to a single valley. All of that is childhood only, frozen in the two to six years when those teeth formed; where these people went as adults, the enamel cannot say.
And the study is careful in a way its headlines were not: one paragraph calls the evidence for rainforest use unequivocal, the next describes Homo erectus as "at minimum, capable of persisting among rainforest fauna." Both are in the paper. The second is the one the data supports.
How they found food in a closed canopy, what plants and animals they took, what tools they carried into the forest — none of that is in nine teeth. What is newly on the table is larger: if Homo erectus could live in the rainforest, then the spread of rainforest was probably not what killed it in Java, and the species' long, slow ending needs a different explanation. Nine teeth from a porcupine's den have reopened a question palaeoanthropology thought it had closed.
📊 KEY FACT-CHECK — what these papers do and do not establish
Established: eight of nine teeth are Homo erectus by dental morphology, geometric morphometrics and enamel proteomics; the deposits date to 324,000–184,000 years ago (Baxian) and 319,000–241,000 (Naxian); this is the youngest securely dated Homo erectus in mainland China; the associated fauna is closed-canopy rainforest; the four proteomically identified individuals carry a previously unsampled Homo erectus enamel variant.
Not established: the species of the ninth tooth, PA1796 (not erectus, not Neanderthal, not modern human; a Denisovan attribution is suggested on circumstantial grounds only); whether Homo erectus lived in the rainforest as opposed to using it — the hominin isotope sample is three teeth; and how far these people ranged as adults, since strontium records childhood alone.
[학술 탐사] 중국 열대우림 호저의 굴에서 나온 이빨 아홉 개
우리가 아는 호모 에렉투스는 늘 '열린 땅'의 인간이었다. 190만 년 전 아프리카를 나와 캅카스와 중국 북부를 거쳐 자바까지 내려갔고, 기다란 다리로 사바나와 성긴 숲을 걸으며 주먹도끼를 들었고, 불을 쓴 최초의 인류로도 일컬어진다. 그리고 숲이 무성해지면 끝이었다—빽빽한 열대우림은 오직 호모 사피엔스만이 뚫어낸 서식지라고, 반세기 가까이 그렇게 가르쳤다. 2026년 11월 18일 《네이처》에 실린 논문 두 편은 그 문장을 정면으로 뒤집는다.
Korean Version: [학술 탐사] 중국 열대우림 호저의 굴에서 나온 이빨 아홉 개
Socko/Ghost