Although many models of accident causation have been developed and have contributed to improving our understanding of how accidents occur, relatively few published studies have evaluated such models with accident data. An accident causation model for the railway industry, proposed earlier by the authors, was reviewed, and 80 railway accident investigation reports from the UK were analyzed to evaluate the model's usefulness and examine the presence of any significant correlations between the components of the model. Overall, it was proved that our model is useful in explaining how a railway accident/incident or near miss occurs and that every component of the model is essential. Human failures, technical failures, and external intrusions were all observed in about one half of the accidents. Human responses were observed in most cases, and responses by protective systems were also reported in many cases. The frequencies of other components such as feedback loops were not negligible. The analysis also revealed several interesting relationships between the components, some of which have implications for preventing or reducing the number of railway accidents and incidents. The results from this study can be transferred to other safety-critical domains such as aviation, maritime, and medicine. (C) 2013 Elsevier Ltd. All rights reserved.