The spatial structure of biotic communities can be shaped by niche-based or stochastic processes, and the importance of both can change through time. Niche-based processes include neighbour interactions, which can change in intensity and quality as communities develop in dependence of environmental conditions. Epiphylls, miniature communities of liverworts, lichens, algae and fungi on leaves, develop only in relatively moist forests, but their leaf-surface habitat is still characterized by moisture stress, especially in more exposed parts of the forest. As neighbours may alleviate moisture stress, we expected that in forest gaps (dryer, brighter), epiphyll communities would show more positive neighbour interactions, resulting in more clustered spatial patterns, than in closed forest, in accordance with the stress-gradient hypothesis. To understand how the processes shaping epiphyll communities change through time and differ between gaps and closed forest, we examined the fine-scale spatial structure within and between epiphyll functional groups (algae, fungi, lichens and liverworts) during 12 months on 60 leaves across 12 sites in a tropical rainforest in Panama. We analysed changes in spatial positions and spatial associations within and between functional groups using spatial point pattern analysis based on repeated photography. Epiphyll densities increased through time for most epiphyll functional groups. However, although epiphyll groups changed their positions through time, the general types of spatial patterns remained similar for most groups: only lichen patterns shifted from mostly random to mostly aggregated through time. We found only random and aggregated spatial patterns within groups, while among groups we additionally found segregated spatial associations, but no consistent temporal trend. Patterns did not differ between gaps and closed forest, thus not supporting the stress-gradient hypothesis. Synthesis. Our results only weakly support the assumption that community-shaping processes change from neutral to niche-based as communities develop. We do provide empirical evidence that epiphyll communities are dynamic and exhibit different spatial association patterns within than among groups. Empirical evidence suggests different interactions and processes within than between groups. Experimental studies and monitoring of individual epiphyll patches are recommended to shed further light on what processes shape life on leaves.