In desert shrubland ecosystems water and nutrients re concentrated beneath shrub canopies in 'resource islands'. Rain falling onto these islands reaches the ground as either stemflow or throughfall and then either infiltrates into the soil or runs off as overland flow. This study investigates the partitioning of rainwater between stemflow and throughfall in the first instance and between infiltration and runoff in the second. Two series of 40 rainfall simulation experiments were performed on 16 creosotebush shrubs in the Jornada Basin, New Mexico. The first series of experiments was designed to measure the surface runoff and was performed with each shrub in its growth position. The second series was designed to measure stemflow reaching the shrub base and was conducted with the shrub suspended above the ground. The experimental data show that once equilibrium is achieved, 16% of the rainfall intercepted by the canopy or 6.7% of the rain falling inside the shrub area (i.e., the area inside the shrubs circumscribing ellipse) is funneled to the shrub base as stemflow. This redistribution of the rainfall by stemflow is a function of the ratio of canopy area (i.e., the area covered by the shrub canopy) to collar area (i.e., a circular area centered on the shrub base), with stemflow rate being positively correlated and throughfall rate being negatively correlated with this ration. The surface runoff rate expressed as a proportion of the rate at which rainwater arrives at a point (i.e., stemflow rate plus throughfall rate) is the runoff coefficient. A multiple regression reveals that 75% of the variance in the runoff coefficient can be explained by three independent variables: the rainfall rate, the ratio of the canopy area to the collar area, and the presence or absence of subcanopy vegetation. Although the last variable is a dummy variable, it accounts for 66.4% of the variance in the runoff coefficient. This suggests that the density and extent of the subcanopy vegetation is the single most important control of the partitioning of rainwater between runoff and infiltration beneath creosotebush. Although these findings pertain to creosotebush, similar findings might be expected for other desert shrubs that generate significant stemflow and have subcanopy vegetation. Copyright © 2003 John Wiley & Sons, Ltd.