The AI Journal

Between 1208 and 1235, the clergy of Várad, in the Hungarian kingdom, kept a register of every ordeal held before the shrine of Saint Ladislaus. Three hundred and eight accused men and women were sent to carry a bar of iron, heated in the church fire, a set distance in their bare hand. Their arm was then bandaged and sealed, and inspected three days later for how the burn had healed. A hundred of those three hundred and eight cases never reached the iron at all: the parties settled, or the accused confessed, once the ordeal had been scheduled but before it took place. Of the two hundred and eight who did carry the iron, a hundred and thirty were pronounced clean. Genuine third-degree burns do not heal cleanly enough, in three days, for five in eight sufferers to pass a priest’s inspection. Something other than the iron’s temperature was deciding these cases.

Várad was not an isolated custom. The same ritual logic, documented across England, France and the German lands by Robert Bartlett’s Trial by Fire and Water, still the standard account of the institution’s mechanics, ran through several variant forms: the hot iron, the scalding cauldron from which a stone or ring had to be retrieved, and the cold-water ordeal, in which a bound defendant was lowered into water that a priest had blessed on the premise that pure water would refuse to receive the guilty. Sinking counted as innocence; floating counted as guilt. The Hungarian register survives in unusual statistical detail, but the institution it documents was the ordinary machinery of proof across much of Latin Christendom for several centuries.

The conventional reading of trial by ordeal treats episodes like Várad’s as evidence of a system that was, at best, theatre for a predetermined verdict, and, at worst, simple fraud dressed up as divine judgment. Peter Leeson’s 2012 reanalysis of the Várad data in the Journal of Law and Economics argues that the theatre was doing real work. Leeson’s model starts from a premise that ordeal scholarship had generally treated as an embarrassment rather than a mechanism: that both the accused and the priests administering the rite sincerely believed a guiltless person would emerge from the ordeal unharmed, because God would not let an innocent burn. Given that belief, a guilty defendant faces a choice quite different from an innocent one. An innocent defendant, confident of vindication, has every reason to proceed to the ordeal rather than accept a costly settlement. A guilty defendant, believing the ordeal will expose the truth, has every reason to confess or settle beforehand rather than risk it. The belief in a “judgment of God” does not need to be true for this sorting to happen. It only needs to be held.

That sorting explains the hundred cases that never reached the iron, but not, on its own, the implausible pass rate among those that did. For that, Leeson’s account turns to the priest rather than the defendant. The clergyman administering the ordeal typically knew more about the case than the bare fact of an accusation: local reputation, the plausibility of the charge, sometimes a private admission extracted under the threat of the ordeal itself. Bartlett documents the latitude built into the ritual at every stage that made this private information usable — the temperature of water judged “boiling” by eye, the exact distance an iron had to be carried, and above all the three-day interval before a wound now sealed under bandages and wax was unwrapped and read as clean or foul by the same priest who had presided over the rite. None of these judgments were mechanical. A cleric persuaded of a defendant’s likely innocence had every opportunity to find the wound healing well; one persuaded of guilt had just as much room to find it festering. The ordeal did not need a rigged thermometer. It needed a discretionary reading, applied by someone with an opinion.

Charles Radding’s account of the ordeal’s eventual disappearance supports this reading of the institution from the other direction. Ordeals did not fade gradually as belief in divine judgment waned over the twelfth century; they ended abruptly, within a generation, after the Fourth Lateran Council’s canon 18, in 1215, forbade priests, deacons and subdeacons from any part in administering or blessing them. Within decades, England had replaced ordeal with jury presentment, and continental practice shifted toward inquisitorial procedure built on witnesses and confession. If ordeals had simply been superstition that persisted because people believed in magic, removing the clergy from the ritual should not, by itself, have ended it; the water and the iron were still there, and belief in divine judgment did not evaporate on the date of a canon. What the council removed was the one actor whose discretion made the system function as a screen rather than a genuine, unmediated trial by combustion. Take the priest’s judgment out of the loop, and an ordeal is no longer a confession machine. It is only a burn.

The strongest objection to this reading does not deny the statistics; it denies that anyone need have been calculating at all. Rebecca Colman’s earlier study of early medieval legal procedure warned specifically against the temptation to read institutions like the ordeal as covert rational-choice mechanisms operated by cynics who privately knew better. The clergy who conducted these rites left no manuals instructing them to cool the iron for a sympathetic defendant, and no confession survives admitting to it. Everything Leeson infers about priestly manipulation is inferred backward from an aggregate pass rate, never observed directly in a single case, and it is entirely possible that individual clerics believed as sincerely in the ordeal’s supernatural mechanism as the defendants did.

That objection narrows the claim rather than defeating it. The screening effect on the accused’s side of the ledger does not require any priest to have manipulated anything: it follows purely from a guilty defendant’s rational fear of a genuinely random or genuinely divine test, which is sufficient by itself to explain why a third of Várad’s cases settled before the iron was heated. The manipulation half of the argument need not be conscious fraud to produce a biased pass rate; a cleric who has heard a defendant’s neighbours vouch for him, and who then reads an ambiguous, half-healed wound three days later, will resolve that ambiguity in the direction his prior belief points, without any sense of having cheated the ritual at all. The ordeal, on this narrower version of the claim, was not a fraud that medieval people were too credulous to see through. It was a genuinely functioning instrument for aggregating dispersed local knowledge about guilt, built out of components — sincere belief, discretionary judgment, a three-day interval for private opinion to settle into public verdict — that required no one involved to have understood, or intended, what the institution as a whole was doing.