ABSTRACT
The authors have spent a large part of their working life studying accident and incident reports in a wide range of hazardous domains. All accident report writers have to chop up continuous and interacting sequences of prior events into discrete words, paragraphs, conclusions and recommendations. Even a simple starting narrative produced a wide variety of event trees, with different nodes and different factors represented at each node. While some versions were simply inaccurate, most were perfectly acceptable accounts. Over the past 50 years or so there has been a dramatic widening of the scope of accident investigations across many different hazardous domains. Clearly, concerns over hardware failures have not gone away, but some of the emphasis has shifted to software problems. The authors deal with the seminal accident investigations thematically rather than chronologically, beginning with legal inquiries that had an enormous influence not only on their own thinking but also on the way subsequent accidents are investigated.
