Evapotranspiration is one of the most difficult terms of the water balance to estimate and model, because of its complex links with atmospheric, hydrological, and ecological processes. In this paper we show how a switching procedure for the boundary conditions at the soil surface achieves a good description of evapotranspiration in a catchment-scale process-based hydrological model. The switching algorithm relies on one parameter, a threshold soil water pressure head (ψmin), to distinguish between atmosphere-controlled and soil-limited evapotranspiration. We successfully applied the model to a small water-limited catchment in southwestern Victoria, Australia, mainly used as pasture and where an extensive hydrological data set is available. Our simulation results show that the model is capable of reproducing satisfactorily the hydrological regime of the catchment without the need for a detailed multiparameter calibration. Specifically, the observed daily flow hydrographs, typical of ephemeral streams, are reproduced satisfactorily for both the calibration and validation phases. Water table levels proved to be more difficult to match, even though the overall groundwater dynamics are well captured by the model. The comparison between measured and observed evapotranspiration rates demonstrates the capability of ψmin to describe the conversion of potential evaporative demand into actual evapotranspiration. The effect of ψmin on the components of the catchment water balance and the model numerical performance were investigated for two different soil types through a sensitivity analysis. The modeled reduction of evapotranspiration with decreasing soil water potential is shown to be analogous to the commonly adopted Feddes formulation of water stress. The results show that the boundary condition-switching algorithm, with a proper choice of ψmin and soil retention curves, can represent a simple and effective way to account for the impacts exerted on the catchment hydrological response by shallow rooted vegetation.