Experiments were conducted in the Patagonian steppe in southern South America to test the following hypotheses: (a) grasses take up most of the water from the upper layers of the soil and utilize frequent and short-duration pulses of water availability; (b) shrubs, on the contrary, take up most of the water from the lower layers of the soil and utilize infrequent and long-duration pulses of water availability. Grasses and shrubs were removed selectively and the performance of plants and the availability of soil resources were monitored. Results supported the overall hypothesis that grasses and shrubs in the Patagonian steppe use mainly different resources. Removal of shrubs did not alter grass production but removal of grasses resulted in a small increase in shrub production which was mediated by an increase in deep soil water and in shrub leaf water potential. The efficiency of utilization of resources freed by grass removal was approximately 25%. Shrubs used water exclusively from lower soil layers. Grasses took up most of the water from upper layers but they were also capable of absorbing water from deep layers. This pattern of water partitioning along with the lack of response in leaf nitrogen to the removal treatments suggested that shrubs may be at a disadvantage to grasses with respect to nutrient capture and led to questions about the role of nutrient recirculation, leaching, and nitrogen fixation in the steppe.