Over the past half-century there have been important developments in the measurement of safety of complex technologies, notably aircraft and nuclear safety (Duffey and Saull, 2003; Reason, 1990; Weick and Sutcliffe, 2007). Safety (or its logical inverse, risk) is difficult to measure when it is good; that is, when nothing happens by which to measure the frequency of abnor­mal occurrences. Under these conditions it is natural for humans to con­clude that the risk is very low or zero. The second level worthy of consideration is the frequency of ‘close calls’, or situations that could have resulted in negative consequences had some fortuitous occurrence not intervened. A ‘close call’ is a clear indication of loss of defence in depth, within the safety regime applied to nuclear plant operation. A third level of defence is available through examination of the availability of ‘poised’ or operation-ready safety systems designed to mitigate the consequences of abnormal events. All of these performance indices rely on administrative attention and action by management and by independent safety auditors assigned to ensure that safety-significant events are actually observed and recorded.