Using Signal Detection Theory to Understand Grade Crossing Warning Time and Motorist Stopping Behavior
Motorist error or poor judgment is a significant causal factor in highway-rail grade crossing collisions. Crashes at grade crossings equipped with warning devices often involve motorists who drive around gates or across railroad tracks while flashing lights are warning them that a train is approaching. This noncompliant behavior may be due to the motorists’ expectations of train arrival time following activation of gates and lights as well as the overall duration of the warning. Because warning times are variable, it is uncertain whether the mean warning duration, the variability of the warning’s duration, or both are influencing motorists’ decisions to disregard the warnings. As a result, signal detection theory was used to model motorists’ stopping behavior at active grade crossings. The key factor in predicting motorist stopping behavior is treating the subjective probability that a train is in the grade crossing as a function of the expected arrival time of the train and this was modeled with Gaussian, Chi-squared and Poisson probability distributions. The probability of stopping predicted from each probability distribution was compared with data collected by Richards and Heathington (1990). The Gaussian model provided the best fit to the data and indicated that warning time variability is the most important factor affecting motorist stopping behavior. Additional data collection to confirm and refine the model is discussed. A theory of motorist behavior at grade crossings, such as signal detection theory, provides a means to critically examine inchoate hypotheses so that they can be more formally stated and vigorously tested. This theory should continue to be developed for evaluating motorist behavior at grade crossings.