Changes in soil microbial community structure due to improvement are often attributed to concurrent shifts in floristic community composition. The bacterial and fungal communities of unimproved and semi-improved (as determined by floristic classification) grassland soils were studied at five upland sites on similar geological substrata using both broad-scale (microbial activity and fungal biomass) and molecular [terminal restriction fragment length polymorphism (TRFLP), automated ribosomal intergenic spacer analysis (ARISA)] approaches. It was hypothesized that microbial community structure would be similar in soils from the same grassland type, and that grassland vegetation classifications could thus be used as predictors of microbial community structure. Microbial community measurements varied widely according to both site and grassland type, and trends in the effect of grassland improvement differed between sites. These results were consistent with those from similar studies, and indicated that floristic community composition was not a stable predictor of microbial community structure across sites. This may indicate a lack of correlation between grassland plant composition and soil microbial community structure, or that differences in soil chemistry between sites had larger impacts on soil microbial populations than plant-related effects.