The objective of this research is a comparative evaluation of different rainfall-runoff model structures. Comparative model diagnostics facilitates the assessment of strengths and weaknesses of conceptual and physically based modeling approaches. Four differently structured models were compared and analyzed with respect to model uncertainty, parametric and input uncertainty. It can be concluded that catchment runoff is simulated satisfactorily by all models. Physically based model structures do not generally outperform conceptual models but they capture runoff events better which originate from infiltration excess. Conceptual model structures are more flexible in fitting to runoff characteristics in different basins. For all models, systematic deviations from runoff observations provide insight into model structural deficiencies. Model structural uncertainty is comparable to parameter and input uncertainty. Large differences between the four models are detected for simulations of soil moisture and, even more pronounced, for simulations of the runoff components. Soil moisture changes are more dynamically simulated by the physically based model structures, which is in better agreement with observations. Simulated streamflow contributions of overland flow are very low in these models. Conceptual approaches tend to higher portions of overland flow, but allow simulations with different compositions of runoff components depending on parameters. Therefore, observations of runoff components, which are usually not available, could enhance parameter estimation for conceptual models and could assist in hypotheses testing of physically based model.