Spiders were sampled at Mpala Research Centre, Laikipia, Kenya by pitfall-trapping and sweep-netting from May 2001 to July 2002, at a Kenyan Long-term Exclosure Experiment. The aim was to establish species composition, checklist and examine spider responses to disturbances caused by cattle, megaherbivores (giraffe and elephants) and mesoherbivores (other ungulates) by looking at three levels of resolution, namely the overall community, guilds and individual species. This is the first controlled replicated experimental study on the effects on invertebrates (spiders) by different land uses (access by large herbivores). A total of 10,487 individuals from 132 species belonging to 30 families were recorded. The family Salticidae had the highest number of species (24), followed by Gnaphosidae (20), Araneidae and Lycosidae (15 each), Theridiidae and Thomisidae (8 each) and Zodariidae (4). Most of the other families had fewer than 4 species. Throughout the study period, species not previously sampled emerged after rainfall peaks. Exclosure treatments affected plant cover, spider diversity and total species mainly through the effects of cattle, whose presence significantly reduced relative vegetation cover. An increase in vegetation cover significantly increased the diversity, total species and species evenness of the overall spider community (total samples data set). Megaherbivores and mesoherbivores had no effects on overall spider diversity. Relative vegetation cover explained approximately 20-30% of variation in community diversity, species richness and species evenness. At the guild level of resolution, the exclosure treatments had no significant effects on diversity, species richness and species evenness of web builders, plant wanderers and ground wanderers. Plant wanderers were significantly and positively correlated with relative vegetation cover, which explained 17% of variation in their diversity. Six individual species responded strongly and in contrasting ways to the same environmental variables, indicating that this level was more sensitive to environmental changes than guilds or the overall spider community. Spider diversity, relative vegetation cover and rainfall varied at a temporal scale of months and not at a spatial scale of hundreds of metres. Only species diversity and species richness from sweep-netting samples and total species from pitfall-trapping varied significantly at a spatial scale of hundreds of metres. Ordination analysis revealed that sweep-netting samples were a better indicator of grazing impacts than pitfalltrapping or combined samples and grouped to reflect cattle grazing, non-cattle grazing and to a small extent the control treatments. Other ordination analyses showed that only samples from sweep-netting and not from pitfall-trapping, were spatially partitioned at a scale of hundreds of metres. This study concludes that the spider fauna of black cotton soil habitats is rich and useful for environmental monitoring and that monitoring of several individual species as indicator of grazing impacts in savanna could be useful and relatively easy.