Abstract

A warning that has sounded many times without harm stops being treated as a warning. Organisational sociology has two rival accounts of that drift. One line, running through Turner's incubation period, Perrow's normal accidents, Vaughan's normalisation of deviance and Sagan's study of nuclear command and control, treats it as something organisations produce and cannot reliably prevent. The high-reliability school of La Porte and Consolini and of Weick holds that some organisations do resist it, by attending to failure and refusing to simplify. Both sides take for granted a claim about what a run of uneventful experience is worth, and neither states it. This article argues that the claim was stated, and argued against, in Sections IV and V of David Hume's An Enquiry Concerning Human Understanding (1748). Section IV shows that no argument carries the mind from the fact that something has not happened to the expectation that it will not happen; Section V says that custom carries it anyway. Read this way, normalised deviance is custom at work on a run of quiet, and the practices of reliable organisations are devices that keep Hume's suspicion alive. The article tests the reading on two synthetic decision records published by WSTiger, a studio of the RISCON family: a chemical plant set in 1984 whose pressure warnings have become routine, and a trading desk set in 2026 that has raised a leverage limit after each breach. Both records separate a calm record from safety. Neither revisit line names an event that would break the calm; each names what would relax the caution instead. The records are modern synthetic scenes with no outcomes, and the article draws no conclusion about which option is right.