Abstract Aim The objective of this study was to comprehensively document and examine the alpha and gamma patterns of species richness in non-volant, small mammals (rodents, shrews and mouse opossums) along a tropical elevational gradient. These data were used to determine the support for existing hypotheses of species richness encompassing mid-domain null models, as well as climatic, and community overlap hypotheses. Location Field studies were conducted along a Caribbean slope of the Río Peñas Blancas watershed in the north-eastern region of Costa Rica between 750 and 1850 m at 10 sampling sites. Methods Species richness and abundances of small mammals were surveyed for four seasons including three temporal replicates at each of five elevational sites: late wet season (2000), early wet season (2001), and dry season (2002), and one spatial replicate at five different sites within the same elevations during the late wet season (2001). Species richness at elevations below 700 m was compiled from specimen records from 23 US national and international collections. Predictions of a null model based solely on geometric constraints were examined using a Monte Carlo simulation program, Mid-Domain Null. Results In 16,900 trap nights, 1561 individuals from 16 species were captured. Both alpha and gamma species richness peaked at mid-elevation between 1000 and 1300 m, with richness declining both at higher and lower elevations. Most of the empirical curves of species richness occur within 95% prediction curves of the mid-domain model, although deviations from the null model exist. Regression of the empirical richness on the null model predictions explained nearly half of the variation observed (r2 = 0.45, P = 0.002). Main conclusions The geometric constraints of montane topography appear to influence the diversity pattern of small mammals, although climatic conditions including an intermediate rainfall and temperature regime, and distance from the persistent cloud cap also are correlated with the pattern of species richness. The predictions of productivity, and community overlap hypotheses are not supported with the empirical data.