U.S. freight railroad accident and hazardous materials release rates have declined substantially since 1980. Ironically, this trend has made the identification and implementation of further safety improvement options more challenging because less empirical information exists on which accident causes present the greatest risks. Consequently, more sophisticated methods are needed to identify the best options for transportation risk reduction. Of particular interest is identifying the principal causes of accidents that can result in a tank car release of hazardous materials, which can harm people, property, and the environment. Because large hazardous materials release accidents are relatively rare, railroads cannot effectively manage safety improvement efforts solely in response to the causes of specific accidents. Instead, a risk-based approach is needed to better understand predictive factors for conditions that can cause a release. Railroad derailment data were analyzed to identify the conditions most likely to lead to a release accident. The objective was to identify proxy variables that can be used as performance measures. The speed of derailment and number of derailed cars highly correlated with hazardous materials releases. Some accident causes are much more likely to lead to release conditions than others. Accident prevention efforts to reduce these causes are more likely to reduce the risk of major railroad hazardous materials release accidents.