1. Human‐caused loss of vertebrate and invertebrate animals, defaunation, is increasing, and potentially affects plant community structure of diverse grassland ecosystems world‐wide.
2. We experimentally simulated defaunation using size‐selective fences to progressively exclude large‐, medium‐ and small‐sized mammals, and invertebrates from two subalpine vegetation types in the Swiss National Park (SNP): intensively grazed short‐grass and moderately grazed tall‐grass vegetation. We assessed plant community properties yearly from 2009 to 2013, and examined treatment effects on plant community structure in the two grassland types.
3. In the short‐grass vegetation, the exclusion of large mammals increased total plant biomass, while the exclusion of large and medium‐sized mammals increased total, grass and forb biomass compared to when all animals had access. These increases became stronger when invertebrates were also excluded. The exclusion of all mammals and invertebrates increased biomass of grasses by 205%, forbs by 100% and total plant biomass by 118% compared to when all animals had access, hence enhancing relative biomass of grasses from 43.6% to 60%, changing plant species composition and lowering richness of forbs by 16%, the number of plant families by 13% and family‐level Shannon diversity by 23%. In contrast to these significant community‐level responses found in the short‐grass vegetation, there was no evidence that the size‐selective exclusion of animals altered the plant community structure of the tall‐grass vegetation. The contrasting results were due to the difference in plant community composition prior to our experiment, which were related to differences in quantity and quality of forage and in grazing intensities of herbivores between the two grassland types.
4. Synthesis. Our results showed that different‐sized animals, in particular large mammals and invertebrates, contributed to maintain the plant community structure in the short‐grass vegetation, highlighting the importance of multiple, functionally different animal groups for ecosystem functioning and stability. In contrast to the short‐grass vegetation, we could not detect such a top‐down control by animals in the tall‐grass vegetation. Our results suggest that potential defaunation effects on grassland plant community structure depend on the degree of grazing pressure release and grassland vegetation type.