When a train passes a red aspect, this is called a Signal Passed at Danger event or SPAD. Sometimes it is easy to identify the SPAD cause but in other cases it is unclear why the incident occurred, especially if the system operated as usual and the train driver was trained and experienced just like his or her colleagues. In previous research, train driver deceleration behaviour has been shown to be influenced by frequent exposure in the previous 14 days to less restrictive and visually similar signal aspects in the same location. Previous exposure can contribute to SPAD causation unless the initial insufficient deceleration is corrected in time. Six years of SPAD data and red aspect approaches in the Netherlands was used to test whether previous exposure to yellow:number aspects corresponds with a statistically significant increase in SPAD incidents if there is a small window for correction available to drivers. The permitted track speed and signal distance influence the size of this window. The results provide evidence for previous exposure as a cause for SPADs and details to identify locations with increased SPAD probability. Changes in infrastructure and timetable design or adding safety measures for these locations can prevent future SPADs.

Employee behavior plays an important role in the occurrence and prevention of incidents, affecting safety margins. In this study, we examine the potential impact of incidental learning on human behavior in the presence of variation in task design. Incidental learning is the day-to-day on-the-job learning that occurs unintentionally. This learning influences which behavior (schema) is more likely to be activated in the employee’s brain. We posit that an incorrect schema can be activated and lead to undesired behavior if the employee is often exposed to (visually) similar tasks that require different behavior. In rail transport, there is a risk of trains passing through red signals. The train driver’s behavior plays an important role in preventing these signal passed at danger (SPAD) incidents. In this study we used speed and location data to analyze train driver deceleration behavior during red signal approaches in the Netherlands. The Dutch rail system showed variation in yellow signal aspects and signal distance. An analysis using 19 months of empirical data indicated changes in behavior when the employee had been previously exposed to different behavior requirements in the same location with a similar yellow signal. These results imply that task design can be improved by taking into consideration what an employee is exposed to during other moments of the shift, and not just during the execution of the specific task.

The field of safety and incident prevention is becoming more and more data based. Data can help support decision making for a more productive and safer work environment, but only if the data can be, is and should be trusted. Especially with the advance of more data collection of varying quality, checking and judging the data is an increasingly complex task. Within such tasks, cognitive biases are likely to occur, causing analysists to overestimate the quality of the data and safety experts to base their decisions on data of insufficient quality. Cognitive biases describe generic error tendencies of persons, that arise because people tend to automatically rely on their fast information processing and decision making, rather than their slow, more effortful system. This article describes five biases that were identified in the verification of a safety indicator related to train driving. Suggestions are also given on how to formalize the verification process. If decision makers want correct conclusions, safety experts need good quality data. To make sure insufficient quality data is not used for decision making, a solid verification process needs to be put in place that matches the strengths and limits of human cognition.