Runoff and infiltration were investigated on abandoned fields of patchy vegetation in semi-arid Spain during 15 months of natural rainfall and by rainfall simulations. The aim was to ascertain sources and sinks of runoff and the effects of soils and plant cover. Soils of the catena developed from mica schists of the upper hillslopes, fan deposits of the lower hillslopes, and an alluvial terrace at the bottom. Runoff from natural events were from three sets of three pairs each of 10 × 2 m runoff plots. The pairs of each set had different densities of plant cover; the sets were vegetated with tussock grass, Stipa tenacissima, a shrub, Anthyllis cytisoides, and a bush, Retama sphaerocarpa. Nineteen natural rainfall events of intensities up to 18 mm/h produced 400 mm of rain during the study period. Because the rainfall threshold for runoff production was about 20 mm, only eight events produced runoff. The rainfall simulations used a sprinkler that produced 50 mm/h of rain for 30 minutes; runoff was recorded each minute in 0.24 m2 bounded plots.The depth and structure of the soil mantle provide the main controls on runoff rates, which are lowest on the lower fan deposits and highest on the thin upslope soils. The river-bank terrace, with a surface covered by crusts and mosses, also yields relatively high runoff. In general, vegetation density varies inversely with runoff. Nevertheless, shrub and bush litter favor runoff, as does a particular spatial distribution of individual plants on the hillslope. Settling of the upper few centimeters of soils of the alluvial fan following cessation of cultivation 15 to 40 years ago has produced a near-surface compacted layer favoring shallow subsurface runoff. Apparently contradictory results between runoff plots and rainfall simulations are the result of differing processes.