Exposure and experience are a confounded nuisance in research on driver behaviour

Road accidents are usually the outcome of multicausal interface problems. Statistical controls therefore need to be equally complex, if accident analyses are to produce understanding of causal factors and lead to correction of contributory errors in driver behaviour. Crude control of accident data for distance travelled, time of day, etc. may identify high-risk groups. However, such controls could produce misleading results from research on individual differences in liability to error, because driving is basically self-paced and purposeful. Certain individuals will thus exhibit characteristically raised levels of risk exposure within the driving task, often as a result of pressures extrinsic to the traffic system. If the errors made by these individuals are to be identified, accident data must be corrected for such self-induced risk exposure, instead of this factor merely being used to “explain” accidents. This need seems most acute in experimental studies, where self-imposed demands and purposes of driving change over time and effects of exposure and experience are thus statistically confounded in accident data. However, control for self-induced risk exposure also seems important in studies of professional drivers' accidents and its neglect calls into question one of the basic assumptions of the “induced exposure” method of interpreting accident data.