Authors of several studies of the spatial distributions of microorganisms have shown strong geographical patterns and stressed the importance of considering the spatial component explicitly when studying assemblage - environment relationships. The processes underlying the patterns are still under debate because it is difficult to separate the unique roles of dispersal limitation and mass effects from spatially structured variation in environment. We analyzed correlations between assemblage dissimilarity and geographical and environmental distances in a large French diatom database, subdivided into regions, years, and different water-quality levels, with multiple regression on distance matrices (MRM) and partial Mantel correlograms. Before we applied MRM, we identified the strongest environmental predictors with the BIO-ENV procedure, which selects the best predictors after testing correlations between distance matrices including every possible set of variables. Environmental control of assemblages was stronger than spatial factors in explaining assemblage patterns, but purely spatial patterns also were significant at the national scale and within some regions. When we averaged environmental and biological data over 3 y, environmental variables accounted for more variability in assemblage structure than relationships estimated with data from a single year. Assemblages in mountainous regions showed particularly strong spatial patterns, perhaps because dispersal barriers hinder the exchange of colonists across sites. The strong spatial structure in the diatom data leads us to encourage researchers to divide ecoregions into smaller areas, especially in mountainous landscapes, when studying assemblage - environment relationships. We also recommend the use of averaged biological and environmental data when developing biotypologies of biotic assemblages for environmental assessment and conservation.