Grime's (1998)“mass-ratio” hypothesis holds that ecosystem processes depend in the short term on functional properties of dominant plants and in the longer term on how resident species influence the recruitment of dominants. The latter of these effects may be especially important among early-successional species in disturbed ecosystems, but experimental tests are few. We removed two groups of early-successional species, an annual forb Gutierrezia dracunculoides (DC.) S. F. Blake and annual species (mostly grasses) that complete growth early in the growing season [early-season (ES) species], from a heavily-grazed grassland in central Texas, USA dominated by a C4 perennial grass. We sought to determine effects of annuals on grassland functioning [productivity, water balance, soil and plant nitrogen (N)] and composition. Removals did not impact N retention in the soil/plant system during the two years of this study, but removing ES annuals increased the amount of water between 30 and 120 cm in the soil profile early in each growing season. Production and N accumulation by vegetation declined following the removal of ES annuals in approximate proportion to the contribution of annuals to aboveground biomass and N, consistent with the mass-ratio hypothesis. By the second year, production and N uptake by initially sub-dominant species increased to fully compensate for the loss of annuals. These results are consistent with the view that ecosystem functions are more strongly linked to species attributes than to diversity per se. Longer-term effects of annuals on grassland composition were evident in a dramatic increase in biomass of perennial forbs after annuals were removed. Because perennial forbs differ from the dominant grass in this grassland in traits that influence ecosystem functioning, ES annuals may affect grassland functioning more by regulating the composition of vegetation than by directly affecting process rates.