At underground or railway station complexes there are generally alternative routes which passengers can use when faced with congestion at escalators, stair-cases, etc. When dealing with these types of situation, the whole pedestrian network likely to be affected by improvements should be modelled, and alternative schemes evaluated, in the same way as for a highway network. Halcrow Fox and Associates (HFA) have developed such techniques on a number of studies in London in recent years. This paper describes the procedures followed, and how these have changed over time, particularly with the advent of the latest congested assignment modelling techniques. The results obtained in recent studies at two of the most important stations in London are presented, and procedures for evaluating pedestrian networks are outlined. Finally, shortcomings which remain in the models are discussed and suggestions made for possible future developments.