Tests of the mid-domain hypothesis: A review of the evidence

Geographic variation of species richness is strongly correlated with environmental gradients. However, random arrangement of species distributions within a bounded domain can also theoretically produce richness gradients without underlying environmental gradients. This mid-domain effect (MDE) could serve as the null hypothesis against which to test effects of environmental variables, or as a component of a multivariate explanation of species-richness patterns. Recent reviews have concluded that there is a substantial MDE signature in observed geographical patterns of richness, based on correlations between observed patterns of richness and the predictions of mid-domain models. However, the mid-domain hypothesis makes additional powerful predictions about how richness should vary through space, and about the slope of the relationship between predicted and observed richness. Very few studies have tested these more powerful MDE predictions. Here, we reexamine the published mid-domain literature for agreement between observed patterns of richness and MDE predictions. We find that 50 of 53 published studies of MDEs showed significant deviations from the predictions of mid-domain models. When observed richness is correlated with MDE predictions, there are nearly always strongly collinear environmental gradients (e. g., in the Americas, climatic favorability and MDE-predicted richness are both maximal in the middle). Interpolation in sparsely sampled data can also give rise to spurious, apparently strong, mid-domain effects (e. g., the classic study of the Madagascan rain forest). We conclude that observed broad-scale patterns of species richness are not consistent with the mid-domain hypothesis.