The ®rst reports of the criticality accident at JCO seemed an echo of historical accidents in dying rust-belt industries: those who design the equipment and procedures are long gone or far away; those who now run the plant allow critical operations to be redesigned on an ad hoc basis with little interest in long-term operational conditions or the monitoring of processes; the workers themselves are underpaid, under-trained, and underquali®ed for the tasks they have to perform, and they improvise with minimal oversight. There is ample evidence for this perspective in the admirable report of Furuta et al (2000), which is careful, detailed, imaginative and unsparing in its criticism of the industry in general and JCO in particular.The report points out that the model of an`industryin-decay' is too simple to explain this case. The plant managers were proud of their safety record in minimising ordinary' accidents and injuries. The business was in some decline, but not decaying. And although there is evidence of neglect of oversight and insuf®cient training of the workers, economic and industrial factors do not provide a suf®ciently robust explanatory framework. It also follows a course that is not unfamiliar to students of past accidents and errors where the operators have not been obviously at fault for clear violation of operating rules and principles. Structure (the social, political and institutional framework that shapes and controls human action) is invoked to control agency (the capacity of human beings for imagining and taking independent action). Organisational and institutional myths, beliefs and values are identi®ed as causal factors, and better methods and procedures for ensuring rational choice are invoked as remedies. What is overlooked are the implications of noting that those who set into motion the sequence that led to disaster did so through careful and logical choice, unconstrained by time or economic pressures. 1 Without minimising the importance of values and beliefs as explanatory factors, I argue for the importance of understanding the degree to which the situational variables within which the operators acted were framed by the industrial-technical characteristics of this particular case. Intermittence ± the infrequency and non-reproducibility of the process ± was one key to the failure because it affected the human agents directly by interfering with individual and collective learning at regulatory, managerial and operational levels. Marginalisation was another: the work site was small, crowded and far away from the centre of JCO's operations, reinforcing the impression that the process was not only ad hoc and improvisational, but not worthy of particular attention (i.e. to safety).Amid the many suggestions and recommendations, two in particular attracted my attention: ®rst, that employees should always be conscious of the risk they take; and, second, that the`safety myth' that the authors ®nd embedded in the industry be replaced with a less mythical attitude based on a`risk-based assessment o...