Safety science is one of the enduring enlightenment projects, which believes that rationality can create a better, more controllable world. Accidents are not seen as meaningless coincidences, but as failures of risk management, as something that can be improved in the future. The tragedy of safety research is that it has to simultaneously deny and affirm the primacy of human agency. As it has gradually expanded away from the sharp end to see accidents as bureaucratic or administrative in origin, the research keeps supplying linguistic and analytic resources that focus on individual shortcomings in leadership, communication or supervision. This paper concludes that individual human agency is useful to safety work, but not just as an instrument of political or organizational expedience. It is useful because it deeply reflects and reinforces how in the West we understand failure and success. The explanatory power of this discourse is confirmed or taken for granted by safety researchers because it appears so ordinary, self-evident and commonsensical.