Grazing exclusion may lead to biodiversity loss and homogenization of naturally heterogeneous and species-rich grassland ecosystems, and these effects may cascade to higher trophic levels and ecosystem properties. Although grazing exclusion has been studied elsewhere, the consequences of alleviating the disturbance regime in grassland ecosystems remain unclear. In this paper, we present results of the first five years of an experiment in native grasslands of southern Brazil. Using a randomized block experimental design, we examined the effects of three grazing treatments on plant and arthropod communities: (i) deferred grazing (i.e., intermittent grazing), (ii) grazing exclusion and (iii) a control under traditional continuous grazing, which were applied to 70 x 70 m experimental plots, in six regionally distributed blocks. We evaluated plant community responses regarding taxonomic and functional diversity (life-forms) in separate spatial components: alpha (1 x 1 m subplots), beta, and gamma (70 x 70 m plots), as well as the cascading effects on arthropod high-taxa. By estimating effect sizes (treatments vs. control) by bootstrap resampling, both deferred grazing and grazing exclusion mostly increased vegetation height, plant biomass and standing dead biomass. The effect of grazing exclusion on plant taxonomic diversity was negative. Conversely, deferred grazing increased plant taxonomic diversity, but both treatments reduced plant functional diversity. Reduced grazing pressure in both treatments promoted the break of dominance by prostrate species, followed by fast homogenization of vegetation structure towards dominance of ligneous and erect species. These changes in the plant community led to increases in high-taxa richness and abundance of vegetationdwelling arthropod groups under both treatments, but had no detectable effects on epigeic arthropods. Our results indicate that decision-making regarding the conservation of southern Brazil grasslands should include both intensive and alleviated levels of grazing