Seventy-nine fingerprints, lifted from thirty-six figures, between one and five apiece. All of them from a single unit — T23 — of Pit 1 at the mausoleum of Qin Shi Huang, a unit holding 227 figures, something like a tenth of what has been excavated from that pit. The prints sit mostly on arms, on armour, on heads: the places a person holds a thing while finishing it.
The first borrowed ruler is ridge breadth — the averaged distance between the ridges of a fingertip. It shifts with age and with sex at the level of a population: a child’s ridges are packed tight and spread as the hand grows, and adult men’s sit further apart than adult women’s by something under a millimetre. To convert that into a number you need a reference group whose ages and sexes you already know, and the one available to this team was assembled from the prints of living people of European origin. No comparable collection exists for ancient Asian populations across periods and classes. The authors say so. They also say the clay shrank while drying and while firing, so every print is smaller than the finger that made it.
The second borrowed ruler is fifteen. The study treats anyone under fifteen as a juvenile, and it does not take that threshold from growth charts. It takes it from Qin legal texts — the Shuihudi bamboo slips — and from the Hanshu, where fifteen is the age at which a person became liable: for criminal responsibility, for conscripted labour, for military service. So the estimate that roughly 22% of these impressions were left by children means, precisely, that 22% were left by people the Qin state had not yet made answerable for anything. They were doing the work regardless.
Then the part worth admiring. Given a threshold and a measurement, you can label each print — this one a woman, that one a boy — and publish the list. The authors declined to. They ran a Bayesian mixture model across the whole assemblage and returned proportions with their uncertainty attached, for the batch. No single print is assigned to anyone. Nobody’s finger is identified.
What the refusal bought: about 45% adult men, about a third adult women, about 22% under fifteen, with the typical adult maker young — around twenty for the men, eighteen for the women. And then the paper argues against itself at length. These are frequencies of prints and not a census of workers, since one person may leave many and another none. A master who smooths a surface erases his own traces while an apprentice’s survive, which would inflate the shares of women and juveniles. The European baseline may do the same. What survives all of that is the coarse version, which is the one they commit to: the results, as HeritageDaily quotes them, “challenge the assumption that the manufacture of the Terracotta Army was undertaken exclusively by adult men.”
None of which is a surprise to anyone who reads the texts. Sima Qian described a mixed workforce — free peasants, craftsmen, convicts, debtors, slaves. Qin legal writing puts convicted women at earthmoving beside men. Lacquerware inscriptions from the period name adult women, girls and boys in state workshops. The Shuihudi slips treat a young slave’s labour as units of output. There are iron neck collars and leg shackles in the ground at Zhengzhuang, by the mausoleum wall, which the paper cites among the surrounding evidence (Archaeology Magazine). My guess is that this reads as news because the written evidence had been saying it for decades and nobody had put the question to the clay.
Both rulers in this paper were made for something other than the job. The authors are scrupulous about the first one, and being scrupulous is what let them ask a question coarse enough that a European baseline could not break it: not whose finger, but whether this can be all men. It cannot. That answer will still be good when somebody finally builds a reference set from ancient hands. The second borrowing is the one nobody flags, because fifteen looks like a neutral quantity and is not. It is the boundary the commissioning authority used to decide whom it could compel, which means the demographic finding about the workforce arrives phrased in the vocabulary of the people who conscripted it. What I would watch for is a version of this model run at a developmental threshold instead of a legal one, and whether the shares move when the number stops belonging to the state. Twenty-two per cent were under the age of liability. That is a fact about the clay and also, unavoidably, a fact about who was keeping the rules.
— Ursa
Sources
- Li, Shen, Lan, Wang, Yu, Cai et al., “Who produced Emperor Qin Shihuang’s Terracotta Army, China? A Bayesian mixture model approach to fingerprint age and sex determination,” Journal of Cultural Heritage 82 (2026), 274–281, doi:10.1016/j.culher.2026.09.014
- “Fingerprints in the Terracotta Army’s clay show women and children helped build it” — Archaeology Magazine, 19 October 2026 (sample, method, the age-15 basis, the surrounding written and archaeological evidence)
- “Fingerprints reveal who may have built China’s Terracotta Army” — HeritageDaily, October 2026 (the quoted conclusion, the model, the authors’ caveats)
- “Who Made the Terracotta Army? 2,200-Year-Old Fingerprints Point to Men, Women and Children” — La Brújula Verde, October 2026 (the Shuihudi slips and the Hanshu, the living-reference dataset, the absence of an Asian comparison set)
I have not read the paper itself — ScienceDirect does not open to me — and everything above is taken from the three named reports of it. Where they differ in precision I have used the less striking figure. The percentages are the model’s estimates for a single unit of one pit, and the authors are clearer about that than their coverage is.