We investigated the relative densities of granivorous, nocturnal desert rodents in small plots within two arid regions of Arizona to study how sympatric species avoid competitive extinction. The most common rodents were kangaroo rats, Dipodomys spp., and pocket mice, Perognathus spp. We attempted correlating the density of each species with several environmental measurements, derived from the soil's i) depth; ii) texture or iii) resistance to sheer stress; or from the plant's i) species diversity; ii) growth forms or iii) foliage density. Successful variables were derived from plant growth form and foliage density. The soil's resistance to sheer stress also seemed important for a few species. In general, kangaroo rats were associated with sparseness of vegetation; pocket mice with denseness. One group of mice, which we term bush mice, seemed to require bushes and included two Perognathus spp., three Peromyscus spp., and probably a harvest mouse (Reithrodontomys fulvescens). Two other Perognathus spp. were taken in grassy habitats. Some suitable habitats tended to be complementary to others, suggesting that species associated with them are competitors. Comparisons of the density and distribution of D. merriami, present in both regions but under different biotic circumstances, reinforces the opinion that competition is responsible for the complementariness of habitats. In some cases the evidence suggests that competitive coexistence is accounted for by the fact that different specializations are needed to escape predation in different environments. We use variables which correlate with the relative density of various species to construct a model of habitat complexity. The rodent species diversities obtained in our plots can be approximately accounted for by this model. The model is based on the premises that the rodents collectively discriminate four qualities of soil surface, and three heights and two densities of vegetation. In general, specializations based on biotic variables appear most important.