Because of the large water storage capacity of soil relative to the atmosphere, changes in soil moisture storage can significantly affect the regional atmospheric budgets of water and energy on monthly, seasonal and longer time scales. Therefore proper modeling of soil liquid water processes is essential to a correct representation of the climate system. This study focuses on the class of summary models of liquid water fluxes in the vadose, or unsaturated zone of the soil, which are applicable to global or regional climate modeling studies. Fourteen such models are represented in this intercomparison study. Observational data from the HAPEX experiment provide validation. Because only limited observational data were available to constrain these models during their development and validation, the models have evolved very diverse treatments of the relevant processes: the basic Darcian (soil internal) and Hortonian (surface liquid flow) processes, as well as the boundary conditions of baseflow drainage and lateral interflow. The annual total local runoff is systematically underestimated by all but one of the participant models. This is one of the few significant biases between the consensus of participant models and the observations. The modeled runoff, averaged over the 14 models, differs from the budget estimate from observations by about 40%. During the period of runoff generation (late winter and early spring) the average model fails to deplete the soil water store as rapidly as is observed, a result consistent with the underprediction of runoff. One cannot rule out insufficient characterization of the field site soils as a primary cause of these discrepancies. Results suggest that model sources of the discrepancy are about equally likely to be related to the prediction of bare soil evaporation (discussed elsewhere in this issue) as they are to the parameterization of runoff and drainage processes.