Habitat filtering models predict ecologically similar plant species to jointly colonize sites due to comparable environmental characteristics leading to an aggregated pattern of species spatial occurrence. Models focused on interspecific competition expect species with similar ecological requirements to be spatially segregated. While both models are corroborated by field work, few empirical studies have tried to infer under which habitat conditions these patterns of co-occurrence prevail. Here we use an exceptional data set on central European pine forest understorey plant communities to assess the change in community structure along gradients of soil productivity and heterogeneity. We found all understorey communities to be significantly nested. The degree of segregation increased with increasing soil Ca and Mg content, as well as with increasing pH, nutrient availability, and moisture. However, variability in soil characteristics did not significantly influence the pattern of species co-occurrence. We also found an intimate link between productivity, species richness, and species segregation making any causal inference challenging. These results point to possible misinterpretations and pitfalls in studies on community assembly. Finally our results demonstrate that managed forests provide a natural experiment of understorey community assembly under controlled conditions, an experiment that deserves further attention.