Matching research goals and methods in system safety engineering

System safety research has a weakness in terms of evaluation. In particular, the field has a poor track record for applying appropriate research methods to the questions we wish to answer, and for evaluating the answers in a convincing way. Given changes in government and public attitudes to research funding, this will be an increasing concern in the near future. In this paper, we define a taxonomy of knowledge types which illustrates the problem and helps to solve it by matching types of questions with appropriate methods. We present prioritised suggestions for further action.