Understanding patterns and mechanisms of variation in the compositional structure of communities across spatial scales is one of the fundamental challenges in ecology and biogeography. In this study, we evaluated the effects of spatial extent (i.e. size of study region) on: 1) whether community composition can be better explained by environmental (i.e. niche-based) or spatial (e.g. dispersal-based) processes ; and 2) how climate and soils contribute to the influence of environment on plant community composition. We surveyed community composition across a network of 398 forest plots spanning a ~4000 m elevational gradient in the Madidi region in northwestern Bolivia. Using redundancy analyses and hierarchical variation partitioning, we disentangled the effects of environmental and spatial predictors on species composition, further decomposing the environmental effect between its climatic and soil components. We repeated analyses for 200 sub-regions ranging in spatial extent from ~250 to ~17 500 km2. Our analyses show a high degree of idiosyncrasy in results that come from different sub-regions. Despite this variability, we were able to identify various important patterns in the structure of tropical plant communities in our study system. First, even though sub-regions varied in size by nearly two orders of magnitude, the total amount of explained variation in community composition was scale independent; at all spatial scales, environment and space accounted for about 25% of the differences in community composition among plots. Second, the measured environmental effect was higher than the spatial effect on average and in the vast majority of sub-regions. This was true regardless of the spatial extent of analysis. Finally, we found that both climatic and soil variables accounted for significant fractions of variation, but climate was always more important than soils. This article is protected by copyright. All rights reserved.