1. Metacommunity theory predicts that increasing patch size and patch connectivity can alter local species diversity by affecting either colonization rates, extinction rates or both. Although species’ dispersal abilities or ‘dispersal mode’ (e.g. gravity-, wind- or animal-dispersed seeds) can mediate the effects of patch size and connectivity on diversity, these important factors are frequently overlooked in empirical metacommunity work. 2. We use a natural metacommunity of aspen stands within a grassland matrix to determine whether dispersal mode alters the influence of stand size and connectivity on understorey plant diversity. We sampled the same area in each patch, controlled for the presence of matrix species in aspen stands, and tested for the effects of size, connectivity and dispersal mode on metacommunity richness. Because dispersal groups responded differently to patch size and connectivity, we created a null model and assessed ungulate activity to explore whether competitive dynamics or herbivory were driving diversity patterns. 3. Animal-dispersed species and species with no dispersal aid had higher diversity per unit area in larger stands, likely because large stands can both support larger populations that are less prone to extinction and may also attract seed-dispersing animals such as birds and small mammals that are sensitive to edge effects. Consistent with other empirical work, we found a positive relationship between diversity and connectivity for wind-dispersed species. However, we detected a negative effect of stand connectivity on the diversity of species with no dispersal aid, possibly due to the presence of other highly competitive species groups dominating well-connected patches, as our null model results suggest. We found no evidence for higher ungulate activity in highly connected patches, suggesting that herbivory may not be driving the decline in diversity of plants with no dispersal aid. 4. Synthesis. Overall, we see a positive effect of stand area on diversity for most groups despite sampling equal area in all stands, which is a prediction of metacommunity theory that is normally overlooked. Our results demonstrate the importance of considering variation in the dispersal modes of focal species for explaining the diversity patterns of natural metacommunities.