Many biological theories have been proposed to explain latitudinal and elevational gradients of species richness, but only recently have theories been proposed that suggest that these patterns may be due solely to geographic constraints. These null models predict mid-domain peaks in species richness as a consequence of geometric patterns resulting from overlapping species ranges between 2 geographic boundaries. Desert rodents exhibit a marked mid-domain peak in species richness for boundaries defined by the latitudinal extent of North American deserts (19°N–45°N). Empirical patterns are compared with predictions of 2 null models: an analytical–stochastic model and the binomial model. Empirical species richness occurs almost entirely within 95% prediction curves of the analytical–stochastic model. Observed species richness is highly correlated with predictions of the binomial model (r² = 93%) but does not generally occur within 95% confidence intervals, in part because empirical range-size distributions differ from predicted distributions. Other diversity theories, species–area relationships, productivity gradients, latitudinal gradients, and Rapoport’s rule are evaluated; none is consistent with empirical patterns. These results demonstrate that the mid-domain effect is a consequence of overlap of variably sized ranges within a bounded region for both ecologically defined hard boundaries and boundaries determined by the shape of the earth. The significant deviations from null-model predictions become the biological points of interest: skewed or localized (or both) pulses or depressions of species richness. Deviations in the present analysis demonstrated a localized pulse in richness caused by a local hard boundary, the Baja peninsula.