Abstract The future likelihood of a hazardous event is predicted from its frequency per unit of past time or by aggregating the influence of the variables that determine it. It is assumed that accurate scientific measurement provides the best estimate, and that public perceptions, if discrepant, are ‘irrational’. Certainly, perception is a constructive process, dependent on partial information, selected again at the receptor stage and distorted in memory store. However, so-called ‘objective’ assessments generally subsume a limited range of variables, discount interactions and are biased to variables that are readily quantifiable for recording or future modelling. Extrapolation from the past rashly assumes stable continuity. In assessing ‘seriousness’ as distinct from probability, it may be argued that only the public can judge, since social and moral values are the ultimate criteria. Clearly, both objective and perceived assessments must be considered and eventually reconciled. The problems of measurement and a quest for lawfulness in the perception of risk are discussed.