The rate at which a plant species is attacked by invertebrate herbivores has been hypothesized to depend on plant species richness, yet empirical evidence is scarce. Current theory predicts higher herbivore damage in monocultures than in species-rich mixtures. We quantified herbivore damage by insects and molluscs to plants in experimental plots established in 2002 from a species pool of 60 species of Central European Arrhenatherum grasslands. Plots differed in plant species richness (1, 2, 4, 8, 16, 60 species), number of functional groups (1, 2, 3, 4), functional group and species composition. We estimated herbivore damage by insects and molluscs at the level of transplanted plant individuals ("phytometer" species Plantago lanceolata, Trifolium pratense, Rumex acetosa) and of the entire plant community during 2003 and 2004. In contrast to previous studies, our design allows specific predictions about the relative contributions of functional diversity, plant functional identity, and species richness in relation to herbivory. Additionally, the phytometer approach is new to biodiversity-herbivory studies, allowing estimates of species-specific herbivory rates within the larger biodiversity-ecosystem functioning context. Herbivory in phytometers and experimental communities tended to increase with plant species richness and the number of plant functional groups, but the effects were rarely significant. Herbivory in phytometers was in some cases positively correlated with community biomass or leaf area index. The most important factor influencing invertebrate herbivory was the presence of particular plant functional groups. Legume (grass) presence strongly increased (decreased) herbivory at the community level. The opposite pattern was found for herbivory in T. pratense phytometers. We conclude that (1) plant species richness is much less important than previously thought and (2) plant functional identity is a much better predictor of invertebrate herbivory in temperate grassland ecosystems.