© 2005 Blackwell Publishing Ltd www.blackwellpublishing.com/geb DOI: 10.1111/j.1466-822x.2005.00189.x 539 Global Ecology and Biogeography, (Global Ecol. Biogeogr.) (2005) 14, 539–547 RESEARCH PAPER Blackwell Publishing, Ltd. Species turnover on elevational gradients in small rodents José Luis Mena* and Ella Vázquez-Domínguez ABSTRACT Aim The elevational gradients of species richness have been well documented; however, studies about species turnover (i.e. beta diversity) in such gradients are limited. We aimed to evaluate the relationship between species turnover and elevation in rodents, and particularly to assess two assumptions at the local scale: (1) species turnover is higher at an intermediate elevation; and (2) turnover decreases at higher elevations as a result of species with larger elevational ranges; and a third assumption at a regional scale; (3) beta diversity changes positively with the size of the elevational range. Location We selected 13 studies from the literature that evaluated the relationship between elevation and species richness in rodents, which include diverse sites around the world. Methods We constructed presence/absence matrices of species occurring at 500-m intervals, in order to measure spatial species turnover, using Whittaker’s (β w ) and Lennon’s (β sim ) indices for pairwise comparisons between each pair of elevation intervals along the gradient (local scale), and for each elevational gradient (regional scale). The relationship between species turnover and elevation at the local scale was evaluated with a LOWESS (locally weighted sums of squares) regression analysis, to determine if the overall trend was linear or curvilinear. The expected distribution of species turnover along the elevational gradients under a mid-domain null model was evaluated for both indices, and the predicted values were statistically compared with the observed species turnover. Regression analysis was used to evaluate the rela- tionship at the regional scale. Results Only one gradient showed spatial autocorrelation. At the local scale, species turnover showed the highest values at an intermediate elevation, which varied between 1000 and 2500 m depending on the particular mountain, for both indices. LOWESS analysis showed a curvilinear trend for most of the gradients, with non- significant regression results. The expected distribution of species turnover by the null mid-domain model showed a U-shaped pattern for β w , while no change (zero values) was predicted with β sim . Observed and predicted species turnover were statistically different. A positive and significant linear relationship was found at the regional scale. Conclusions Our results suggest that the presence of clearly defined species assem- blages is reflected in the variation of turnover with elevation. Accordingly, (1) a com- mon increment of species turnover around mid-elevations along the gradients was found for the rodents studied, whether measured independently (β sim ) or not (β w ), of species richness; (2) the trend was dependent on the size of the gradient: it was evident only for elevational gradients exceeding 1900 m; (3) some gradients did not show an increase in the elevational range of species at higher elevations, contrary to Stevens’ rule expectations; (4) the expected species turnover distribution signific- antly contrasted with the empirical results, confirming that the extent and location of the species’ elevational ranges that overlap along the gradient determine the trend Instituto de Ecología, UNAM, Ciudad Universitaria, Apdo. Postal 70-275, Mexico, DF, 04510, México *Correspondence: José Luis Mena, Apartado Postal 07-0145, Santa Luzmila, Lima 7, Peru. E-mail: jlmena@viabcp.com