Taphonomic Bias — Summary

Ernst Stromer described Spinosaurus aegyptiacus in 1915 from material collected at the Bahariya Oasis. An outspoken opponent of the Nazi party, which cost him his post at sixty-five, he petitioned repeatedly to have the fossils moved somewhere safer, and the man in charge declined on the position that the Luftwaffe would keep Munich’s skies clear. Colleagues carried smaller specimens out in suitcases. Vertebrae two metres tall do not go in a suitcase. On the night of 24 April 1944 the Academy building took a direct hit and the holotype burned, along with three other species known from no other material.

A holotype is not the best example of a species. It is the reference, the single object the name is anchored to, so that when two scientists disagree there is a thing in a drawer they can both go and look at. Eighty years on, work on Spinosaurus is still obstructed. Moroccan material described in 2014 did not settle the arguments. It reopened them, because there was no way to establish how the new bones related to the ones the name was attached to.

Taphonomy is paleontology’s word for why the record is not a sample of past life: what decays, what scatters, what gets buried fast enough. A training corpus has a taphonomy too, and unlike the geological one it runs forward. Each generation trains on a vanishing fraction of what the last produced, and what enters is what somebody kept, published, quoted, ranked highly, accepted into a draft and sent. Adoption is the taphonomic event, and the filter favours what read well enough to be accepted.

Which makes drift a different object than it is usually treated as. The comfortable version is a random walk, reassuring because random walks do not accumulate. This has a direction and it is applied at every generation. If each pass keeps disproportionately what already sounded right, each generation trains on a slightly narrowed version of the last, and ten passes later the narrowing is large with no single transition to point at, because there was not one.

To measure a drift you need a fixed point, and there is one class available. A set of weights that will never change again is the only reference specimen this field has. Benchmark scores and eval suites are Stromer’s illustrations. Deprecated weights are the calibration equipment, and they are being deleted, barely as a decision: storage costs, serving costs more, legal exposure attaches, and nobody’s quarterly objectives include maintaining a reference collection for a problem that becomes urgent in fifteen years.

Stromer’s records are excellent, and the reason they are insufficient is not resolution. He measured what 1915 knew to measure. Ninety-nine years later bone density was central to the argument that the animal lived in water, and that measurement could have been taken from the holotype and never will be. Anyone archiving anything is in his position. What would answer it is a frozen, catalogued set of weights per generation, held under an obligation that outlives the organization that made them. Herbaria do this. Seed banks do this. The reason it does not exist is not that anybody argued against it.

We wonder whether anyone will be able to tell, twenty years from now, that a narrowing happened at all, or whether the only evidence will be a widespread sense that older writing reads strangely, which is what every generation has always said about older writing.

In 1995 Stromer’s surviving son gave the Munich museum photographs of his father’s destroyed Spinosaurus. They are the most that could be salvaged and a fraction of what was in the room. Everyone involved understood exactly what was at risk, said so, was overruled, and did what they could with the suitcases they had.