One of the most generalized conclusions arising from studies analyzing the ecological variation of energy metabolism in endotherms is the apparent negative correlation between ambient temperature and mass-independent basal metabolic rate (residual BMR). As a consequence, ambient temperature has been considered the most important external factor driving the evolution of residual BMR. It is not clear, however, if this relationship is size dependent, and artifacts such as the biased sampling of body masses in physiological data sets could cause us to overstate the ubiquity of the relationship. Accordingly, here we used published data on body mass (mb), BMR, and annual mean temperature (Tmean) for 458 mammal species (and/or subspecies) to examine the size-dependence of the relationship between temperature and BMR. We found a significant interaction between mb and Tmean, such that the effect of Tmean on residual BMR decreases as a function of mb. In line with this, the amount of residual variance in BMR explained by Tmean decreased with mb, from 20 - 30% at body sizes < 100 g to almost zero at body size > 1,000 g. These data suggest that our current understanding of the importance of broad-scale variation in ambient temperature as a driver of metabolic evolution in endotherms probably is affected by the large number of small species in both nature and physiological datasets.