In the last two decades, many rainfall-runoff models have been developed, tested and the results published. Most of them give, as their authors claim, satisfactory results. But, given that the results are good, why do new models continue to be published? And why is the unit hydrograph still the most widely used approach, although it is nearing its 50th anniversary? In the study presented here, models of different complexity have been calibrated to three small basins (1.7, 10.9, 120 km2) with dense networks of recording raingauges (0.6, 1.0, 40 km2/gauge) and reliable stream-gauges. The results show that simple models can give satisfactory results; however neither the simple nor the more complex models tested were free from failure in certain cases, because none of them adequately describe the rainfall-runoff process. In addition, it could not be proved that complex models give better results than simpler ones.