ABSTRACT
This paper describes a method of modelling human dependent failures across the subtrees of major fault trees to produce results which fully account for all the different types of interaction which can occur. It overcomes the problems inherent in most current methods of modelling human dependent failures in which the approach is to make adjustments within the context of the fault tree logic. The method which is described constructs the fault tree logic to represent independent failures, and corrects the values of human error probability for dependency at the level of the cutsets.
