Abstract. Concern over the accelerating loss of biodiversity has stimulated renewed interest in relationships among species richness, species composition, and the functional properties of ecosystems. Mechanistically, the degree of functional differentiation or complementarity among individual species determines the form of such relationships and is thus important to distinguishing among alternative hypotheses for the effects of diversity on ecosystem processes. Although a growing number of studies have reported relationships between plant diversity and ecosystem processes, few have explicitly addressed how functional diversity at higher trophic levels influences ecosystem processes. We used mesocosm experiments to test the impacts of three herbivorous crustacean species (Gammarus mucronatus, Idotea baltica, and Erichsonella attenuata) on plant biomass accumulation, relative dominance of plant functional groups, and herbivore secondary production in beds of eelgrass (Zostera marina), a dominant feature of naturally low-diversity estuaries throughout the northern hemisphere. By establishing treatments with all possible combinations of the three grazer species, we tested the degree of functional redundancy among grazers and their relative impacts on productivity.Grazer species composition strongly influenced eelgrass biomass accumulation and grazer secondary production, whereas none of the processes we studied was clearly related to grazer species richness over the narrow range (0-3 species) studied. In fact, all three measured ecosystem processes-epiphyte grazing, and eelgrass and grazer biomass accumulation-reached highest values in particular single-species treatments. Experimental deletions of individual species from the otherwise-intact assemblage confirmed that the three grazer species were functionally redundant in impacting epiphyte accumulation, whereas secondary production was sensitive to deletion of G. mucronatus, indicating its unique, nonredundant role in influencing this variable. In the field, seasonal abundance patterns differed markedly among the dominant grazer species, suggesting that complementary grazer phenologies may reduce total variance in grazing pressure on an annual basis. Our results show that even superficially similar grazer species can differ in both sign and magnitude of impacts on ecosystem processes and emphasize that one must be cautious in assuming redundancy when assigning species to functional groups.