Aim To determine how well variation in median body size of avian assemblages is predicted by (1) the environmental models usually employed in analyses of Bergmann's rule and (2) random sampling from the regional body size frequency distribution. If body size frequency distributions of local assemblages represent a random sample of a regional frequency distribution, then geographical variation in body sizes of assemblages might be a consequence of the determinants of spatial variation in species richness rather than direct influences on body size per se.
Location Southern Africa.
Methods Median body masses (as a measure of body size) of avian assemblages were calculated for quarter-degree grid cells across South Africa and Lesotho. The relationship between median body mass and four environmental variables (minimum and maximum monthly temperatures, precipitation and seasonality in the normalized difference vegetation index, as a measure of seasonality in productivity) was examined using general linear models first without taking spatial autocorrelation into account, and then accounting for it by fitting an exponential spatial covariance structure. Model fit was assessed using the Akaike information criterion and Akaike weights. At each species richness value, random assemblages were sampled by either drawing species randomly from the regional body mass frequency distribution, or drawing species from the regional body mass frequency distribution with a probability proportional to their geographical distribution in the area. The ability of randomizations to predict actual body masses was examined using two-tailed Fisher exact tests.
Results Seasonality in productivity was the only environmental variable that remained a significant predictor of body mass variation in spatially explicit models, though the positive relationship was weak. When species richness was included in the models it remained the only significant predictor of size variation. Randomizations predicted median body mass poorly at low species richness, but well at high richness.
Main conclusions Environmental models that have previously been proposed explain little of the variation in body mass across avian assemblages in South Africa. However, much of the variation in the median mass of assemblages could be predicted by randomly drawing species from the regional body mass frequency distribution, particularly using randomizations in which all species were drawn from the regional body mass frequency distribution with equal probability and at high species richness values. This outcome emphasizes the need to consider null expectations in investigations of the geographical variation in body size together with the probable environmental mechanisms underlying spatial variation in average size. Moreover, it suggests that in the South African avifauna, spatial variation in the body sizes of assemblages may be determined indirectly by the factors that influence geographical variation in species richness.