This paper describes how fault tree analysis may be used to calculate the probability that people will detect fires in hospitals. In particular, study is made of the contribution of patients and varying staff numbers. Their contribution is based on new data gathered by shadowing staff and assessing wards. This data showed that staff spent 75 per cent of their time undertaking duties that would facilitate the detection of fires, in locations where they could observe more than 30 per cent of the ward. This provides a greater understanding why the historical data indicates that staff observation of fire is so effective. It is concluded that fault tree analysis offers a useful tool which can inform decisions regarding the cost/benefit of fire protection measures.