For a century, visitors wept over a mother shielding her baby. In 2024, the DNA came back.
The plaster casts of Pompeii are not bodies. They are voids — hollows left in the hardened ash as the dead decayed, filled with plaster by excavators from the 1860s onward. The technique preserved final postures with photographic fidelity, and postures invite narrative: a figure curled beside a smaller one becomes a mother and child; two forms in an embrace become sisters. Museums, guidebooks, and a century and a half of visitors filled in the relationships, and the relationships hardened into fact by repetition. Nobody could check. Then, in 2024, somebody checked.
A team of geneticists extracted ancient DNA from skeletal fragments embedded in the plaster of several famous casts and published the results in Current Biology. At the House of the Golden Bracelet, the adult wearing the heavy gold band — long presented as a mother who died shielding her child — is genetically male, and unrelated to the child on his lap. The four victims of that house, displayed for decades as a nuclear family, share no close kinship at all. The pair in the celebrated “embrace,” variously narrated as sisters or as mother and daughter, includes at least one genetic male. The strontium and ancestry data added a quieter correction: the tested victims descended largely from recent immigrants from the eastern Mediterranean — Pompeii's dead were a port city's dead, more cosmopolitan than the Roman-family tableau the casts had been arranged to tell.
The study also documented how physically constructed some of the narratives were: several casts had been restored, repositioned, and smoothed by earlier generations of conservators, the plaster itself edited toward the story being told about it. None of this makes the casts less real as records of how people died. It makes them unreliable as records of who those people were to each other — a distinction the ticket-buying public was never offered.
The researchers drew the moral themselves, in the paper and in interviews: the errors were not random, and that is what makes them worth a record. Interpreters had read jewelry as female. They had read an adult holding a child as the child's mother. They had read physical closeness as blood family. Take each guess alone and it sounds reasonable — plenty of adults holding children are their mothers. The problem is that every guess leaned the same way, toward the family structure the interpreters themselves lived inside, and a set of guesses that all lean in one direction isn't observation anymore. It's projection with a museum label. And there is a specific reason the stories then survived a hundred and fifty years unchecked: they matched what visitors already expected a Roman family to look like, and a story that confirms your expectations never triggers the reflex to verify it. That is the same reflex failure running through this whole file, pointed in a new direction. The 2024 study needs no modern parallel from elsewhere, because it is the modern half of its own story: the trap this record documents — narrating the dead in your own image and mistaking it for looking at them — was sprung not on the Romans but on us. It held from the first plaster pour until somebody finally ran the test.
Nineteen centuries of mourners wept at the wrong story. The casts never lied — you saw what you wanted to see. You usually do.