Despite the dangers, many drivers continue to take risks when driving. This paper outlines an explanation of this risk-taking behavior and the failure of numerous fear-arousing messages to change the manner in which many people drive. Being an occupant in a car and learning to drive may be seen as analogous to systemic desensitization and flooding procedures, in which fear is extinguished and/or a response inconsistent with fear is learned. Once this procedure is complete the fear response is unlikely to be reinstated by messages pointing out possible dangers on the road, because the situations in which the messages are received are usually inappropriate, and many believe that they are superior drivers and therefore not at risk. News presentation of the huge road toll and multiple fatality crashes may only confirm to many people that they are better than average drivers since so many other people have been killed or seriously injured, and they, the superior drivers, have not. A prediction of this account is that driving confidence will increase with increasing age, through the greater on-road fear-reducing experience and increased exposure to the road toll. This prediction was examined in surveys of 2,963 Australian drivers, conducted as part of the evaluation of random breath testing. Respondents were asked to rate their ability as drivers compared with average, and to rate their ability to drive under the influence of alcohol. The overconfidence observed in Canada, Sweden, New Zealand, and the United States was identified in Australian drivers. The predicted increase in confidence with increasing age was supported up to the age of 40 years, after which confidence changed little. Confidence in ability to drive after consuming alcohol increased steadily with age. Finally, it was predicted that the introduction of random breath testing and the associated media campaign partly about the effects of alcohol (which was successful in reducing the road toll) would decrease confidence in ability to drive under the influence of alcohol. Comparison of survey data before and after the introduction of random breath testing did not support this prediction.