Chapter 3 relates the approach taken by accident-proneness as one of the earliest attempts to scientifically investigate the ‘human element’ in safety. It took the emerging sciences of psychology and eugenics, and applied them to explain patterns in industrial data. Accident-proneness built on two patterns that were hard to dispute: (1) Human performance is variable, in ways that can be relevant for accidents, and (2) Some people are involved in more accidents than others, in ways that are unlikely to arise purely by chance. In doing so, it provided an object lesson in the limitations and moral perils of treating human capacities in a reductionist, ‘scientific’ way. And, as the chapter explains, accident proneness is ultimately difficult to prove. Safety outcomes are never purely due to specific individual attributes. If the same activities take place in different contexts (busy times versus low workload, or night versus day), even the same individual may well have different safety outcomes. It can never be determined with certainty how much of the outcome is attributable to the individual, and how much to context.