A theoretical model of the process of general deterrence is constructed, and tested against data collected as part of an evaluation of the impact of the introduction of random breath testing (RBT) in New South Wales, Australia in December 1982. The model entails a specification of the causal links between police activity and media publicity, and behaviour change. The model goes beyond utility theory in the conceptualisation of the processes whereby an individual may choose between driving after drinking and alternative modes of action. The data were derived from two surveys of the general population conducted within four months of the introduction of RBT, and included a longitudinal component (185 drinking motorists were reinterviewed after six weeks). Despite problems of measurement, the theoretical model was strongly supported for the short-term impact of RBT. It is concluded that deterrence is an unstable process, and that the long-term deterrent impact of measures like RBT depends mainly on the level of continual, visible police enforcement.