ABSTRACTAim Species richness patterns along elevational gradients have been documented extensively. Yet, the implications of differences in how the data are compiled are seldom explored. We investigate the effect of grain size on the richness–elevation relationship. Grain size varies among the principal methods used to collect or aggregate species occurrences: localized sites, elevational ‘bins’ and interpolation of species ranges. Assumptions of sampling and species distributions also vary among these methods. Methodology can influence the pattern that is perceived and comparability of results. We compare patterns from all three methods explicitly using the same suite of observations, based on museum records and field surveys of non‐flying small mammals. Our assessment is enhanced by comparing patterns resulting from each method for each of six adjacent mountain ranges.Location Utah, North America.Methods We document elevational species richness patterns using generalized linear models (GLMs), comparing the general shape of the trend as well as curvature, location and magnitude of peak richness across methods, both within and among gradients. We also introduce a new procedure to test for richness peaks using site‐based occurrences.Results We find a general congruence of the richness–elevation relationship, depicting a hump‐shaped pattern with a second‐order polynomial GLM showing a significant fit to nearly all gradient‐methodology combinations. However, underlying characteristics of the trend may vary with grain size. As grain size coarsens, maximum species richness increases and elevation of the mode slightly decreases. Results for curvature vary, but degree of curvature tends to increase as grain size coarsens. The richness–elevation patterns are independent of sampling effects.Main conclusions The perceived elevational diversity pattern for small mammals along these mountain ranges is not scale‐dependent. Differences in how the data are compiled are not reflected in major differences in patterns, even when local samples are neither uniformly spaced nor sampled with the same intensity. This result lends confidence to the assertion that patterns documented in similar studies with different methodologies and for which sampling is sufficiently comprehensive are good indicators of diversity. However, consistency of results from more than one compilation method may help to address issues of scale‐dependence, more so when these comparisons are made explicit.
