Future bridge management systems will be based on simple stochastic models predicting the residual strength of structural elements. The current deterministic management systems are not effective in optimizing e.g. the life cycle cost of a bridge or a system of bridges. A number of important factors are so uncertain that they cannot be modelled satisfactorily by deterministic techniques. In recent years several researchers have studied stochastic modelling of e.g. deterioration, inspection, reliability, and repair but real applications are only reported in a few cases.