In an experiment that directly manipulated grassland plant species richness and composition, decreased plant species richness (''diversity'') increased pathogen load (the percentage of leaf area infected by species-specific foliar pathogens across the plant community) in 1998. Pathogen load was almost three times greater in the average mono- culture than in the average plot planted with 24 grassland plant species, an approximately natural diversity. Eleven individual diseases increased in severity (percentage of leaf area infected by a single disease) at lower plant species richness, and severity of only one disease was positively correlated with diversity. For 10 of the 11 diseases whose severity was negatively related to diversity, disease severity was positively correlated with host abun- dance, and in six of these cases, species diversity had no effect on disease severity after controlling for the effects of host abundance. These results suggest that increased abun- dances of individual host species at lower species diversity increased disease transmission and severity. In 1996 and 1997, similar results for a smaller number of diseases sampled were found in this experiment and another similar one. Although the effect of diversity on disease was highly significant, considerable variance in pathogen load remained among plots of a given diversity level. Much of this residual variance was explained by community characteristics that were a function of the species composition of the communities (the identity of species present vs. those lost). Specifically, communities that lost less disease- prone species had higher pathogen loads; this effect explained more variance in pathogen load than did diversity. Also, communities that lost the species dominant at high diversity had higher pathogen loads, presumably because relaxed competition allowed greater in- creases in host abundances, but this effect was weak. Among plant species, disease prone- ness appeared to be determined more by regional than local processes, because it was better correlated with frequency of the plant species' populations across the region than with local abundance or frequency across the state. In total, our results support the hypothesis that decreased species diversity will increase foliar pathogen load if this increases host abun- dance and, therefore, disease transmission. Additionally, changes in community character- istics determined by species composition will strongly influence pathogen load.