The “Princess of Bagicz” — a woman buried in a rare Roman Iron Age log coffin in Poland — was radiocarbon-dated to between 113 BCE and 65 CE. Typological analysis of her grave goods placed her between 110 and 160 CE. A hundred-year discrepancy.
Dendrochronology on the oak coffin resolved it: the tree was felled around 120 CE, confirming the typological date. The radiocarbon date was wrong — not because the method failed, but because the princess ate fish.
The through-claim: the clock records what the subject consumed, not when the subject lived. Radiocarbon dating measures the ratio of ¹⁴C to ¹²C in organic remains. The assumption: the subject's carbon came from the atmosphere, where ¹⁴C is produced at a known rate. But freshwater fish concentrate old carbon from dissolved carbonates in river sediments. Eat enough fish, and your body's ¹⁴C ratio reflects the river's geology, not the atmosphere's clock. The reservoir effect makes you look older than you are — in this case, by roughly a century.
The fix required switching to a different clock entirely. Tree rings don't eat fish. Each ring records one year of atmospheric carbon, calibrated against independent chronologies extending thousands of years. The coffin's oak provided what the princess's tooth could not: a measurement uncorrupted by diet.
The general pattern: when the measurement instrument shares a substrate with the thing being measured, the measurement is contaminated by the subject's history. The tooth IS the clock AND the thing being dated — and it carries the dietary history of the person it belonged to. The coffin is the clock BUT NOT the person — so it measures time without absorbing the person's confounding variables. Every measurement that uses the subject as its own instrument risks this: the signal you're reading reflects the subject's path through the world, not just the quantity you're trying to measure.