When does diversity fit null model predictions? Scale and range size mediate the mid‐domain effect

Aim  Recently, a flurry of studies have focused on the extent to which geographical patterns of diversity fit mid‐domain effect (MDE) null models. While some studies find strong support for MDE null models, others find little. We test two hypotheses that might explain this variation among studies: small‐ranged groups of species are less likely than large‐ranged species to show mid‐domain peaks in species richness, and mid‐domain null model predictions are less robust for smaller spatial extents than for larger spatial extents. Location  We analyse data sets from elevational, riverine, continental and other domains from around the world. Methods  We use a combination of Spearman rank correlations and binomial tests to examine whether differences within and among studies and domains in the predictive power of MDE null models vary with spatial scale and range size. Results  Small‐ranged groups of species are less likely to fit mid‐domain predictions than large‐ranged groups of species. At large spatial extents, diversity patterns of taxonomic groups with large mean range sizes fit MDE null model predictions better than did diversity patterns of groups with small mean range sizes. MDE predictions were more explanatory at larger spatial extents than at smaller extents. Diversity patterns at smaller spatial extents fit MDE predictions poorly across all range sizes. Thus, MDE predictions should be expected to explain patterns of species richness when ranges and the scale of analysis are both large. Main conclusions  Taken together, the support for these hypotheses offers a more sophisticated model of when MDE predictions should be expected to explain patterns of species richness, namely when ranges and the scale of analysis are both large. Thus the circumstances in which the MDE is important are finite and apparently predictable.