In recent years, distributed, physically based hydrological models are being developed and applied more and more. At the same time, contradictions are emerging regarding claims of model applicability on the one hand and lack of validation of these claims on the other hand. In this connection the necessity of applying a rigorous validation methodology becomes obvious. The present paper presents an outline of a general methodology with special emphasis on the additional requirements for distributed models as compared to lumped rainfall-runoff models. A case study on validation and inter-comparison of three different models on catchments in Zimbabwe is described. The three models represent a lumped, conceptual model (NAM), a distributed, physically based model (MIKE SHE) and an intermediate approach (WATBAL). On the basis of these results and the authors' other experience, a discussion is given on the applicability and associated validation requirements for different types of hydrological models.