We estimated and tested variability of seagrass leaf-associated epifaunal assemblages at a range of scales. Sampling was performed in 36 seagrass (Zostera marina) meadows within three regions along the Swedish west coast following a hierarchical design (samples separated by 10 s m, km or 100 km). Results showed strongest variability (43-81%) at the intermediate amongst-meadow (km) scale using biomass of functional categories, while considering taxa composition the within-meadow (10 s m) scale contributed most to variability (60%). Using functional categories, we found that embayment exposure and seagrass shoot density were the most important predictor variables explaining part of the variability in biomass of suspension feeders (bivalves and barnacles) and grazers. In contrast, variability in epifaunal taxa composition was predicted mainly by sediment chemistry, substratum coverage and geographical positioning. Our findings suggest that models to develop predictive power and mechanistic understanding should focus on variables and processes varying at small and intermediate scales rather than those varying at larger scales.