Measurement of driver performance, with regard to safety, has traditionally posed great difficulty. While safety is often discussed in terms of risk probabilities, measurement of risk probability is hampered by several factors. First, accidents are relatively rare occurrences. Furthermore, the identifying event (the accident) occurs at the end of the time period which contains the precipitating events. As a result, it is quite difficult to monitor for potential accident scenarios. In addition, accidents are most typically the result of several compounding factors, which makes determination of causality very difficult.The safety state network is a probabilistic model which captures the behavior of a system. Based on a finite Markov network, the safety state network models the human/machine and human/human interactions in a transportation system and forms a framework for capturing and comparing the probabilistic decision patterns of control elements in a transportation system. This paper discusses the theoretical basis of the safety state model, and applications to measuring driver performance.