Five different rainfall - runoff models including three sophisticated conceptual models (Sacramento, Stanford and Monash Models), a simple conceptual model (Boughton Model) and a black-box or purely mathematical method involving a recursive time series alogirithm were tested on three catchments in the south-west region of Western Australia. A variety of techniques were used to assess the performance of each of the models which were calibrated on about six years of data and then tested on a further six or seven years of extra data. As well as the objective comparison of the models, a subjective assessment of the user and computation aspects of model calibration is presented. All of the models except the Boughton Model performed in a satisfactory manner with the Sacramento Model and Time Series Method being given overall recommendation.