Figure 2.

The saturating assembly rule in combination with nearest neighbour competition resulted in communities with species diversity most similar to that of the real communities. Three rules (a–c) relating body size difference to colonisation probability were used to simulate community assembly. Community trajectories in terms of Simpson's diversity are presented over the course of 25 community assembly cycles, with 1000 replicates per rule combination. Panels (f) and (g) show behaviour under the assembly rules assuming that only the most similar sized species interact (nearest neighbour competition) and (h) and (i) all resident species interact (diffuse competition). The results of the null model (e) are the same for nearest neighbour and diffuse competition (by definition) and hence are presented only once. Panel (d) shows Bayes factors for comparisons of fits to the observed diversity of wild communities of the four different size-based assembly models with the null model.