Abstract
We have explored accident data from British Rail in the period from 1946 through 2005. Ourhypothesis has been that safety is improved through learning from experience. Based on a quantitative analysis this hypothesis is tested for the data using a simple regression model. We have discussed the model and its limitations, benefits and possible improvements.We have also explored our findings based on qualitative theory from the field of organisational learning, and have suggested key issues to be explored to improve safety and resilience during changes, such as safety cases, standardisation of training, unambiguous communication and sharing of incidents and mitigating actions.