In the archive of the Inquisition at Udine, Carlo Ginzburg found a run of cases from Friuli beginning in 1575 in which the defendants said something the inquisitors were not expecting and could not use. They were not witches, they insisted. They were benandanti, those who go well. Born with the caul, they were called out on certain nights, went forth in spirit while their bodies lay as if asleep, and fought the witches in the fields with stalks of fennel. If they won, the harvest was good. If they lost, it failed.
The inquisitors had a category ready and it did not fit. Their questions kept steering towards the sabbath, the pact, the renunciation of the faith. The defendants kept insisting that they fought for Christ and against the witches, which is the opposite claim. Over roughly fifty years of intermittent questioning the pressure told. The answers shifted, the vocabulary of the interrogators seeped into the testimony, and by the 1630s some benandanti were describing themselves in the terms the court had been offering all along.
That slow assimilation is the finding, and it is the clearest documented example of the process historians suspect lies behind confessions everywhere. The interrogation did not uncover a belief. It replaced one.
Ginzburg's method, reading the record for what the defendant said before the questioner reshaped it, has been taken up across the field, well beyond witchcraft.