In modeling, actual crop transpiration as a function of soil hydraulic conditions is usually estimated from a water content or pressure head dependent reduction function. We compared the performance of the empirical pressure head based reduction function of Feddes (FRF) and a more physically based reduction function using matric flux potential as the main parameter (DRF), both available in the SWAP ecohydrological model. Model performance was evaluated by comparison of SWAP predictions and observed water contents and pressure head values in a field experiment with a common bean (Phaseolus vulgaris L.) crop. For >50 d, no rain occurred and the soil reached very dry conditions with pressure heads in the range −100 to −150 m. The SWAP–DRF-predicted pressure head and water content values were less sensitive to root length density distribution than those predicted by SWAP–FRF. Varying wilting pressure head did not improve predictive performance. Root water uptake distribution with time and depth simulated by SWAP showed very different patterns depending on the reduction function used. Root water uptake estimated by SWAP–FRF showed smooth transitions with time and between layers, whereas SWAP–DRF, highly sensitive to hydraulic conditions, generally predicted uptake to be concentrated at a few depths. The order of magnitude of the pressure head difference between root xylem and root surface based on SWAP–DRF-predicted uptake rates, root length density, and reported values of root conductance was the same as the order of magnitude of the limiting root water pressure head, implying the necessity to include root hydraulic resistance in the DRF.