Power scaling relationships between body mass and organismal traits are fundamental to biology. Compilations of mammalian masses and basal metabolic rates date back over a century and are used both to support and to assail the universal quarter‐power scaling invoked by the metabolic theory of ecology. However, the slope of this interspecific allometry is typically estimated without accounting for intraspecific variation in body mass or phylogenetic constraints on metabolism. We returned to the original literature and culled nearly all unique measurements of body mass and basal metabolism for 695 mammal species and (1) phylogenetically corrected the data using the fullest available phylogeny, (2) applied several different regression analyses, (3) resampled regressions by drawing randomly selected species from each of the polytomies in the phylogenetic hypothesis at each iteration, and (4) ran these same analyses independently on separate clades. Overall, 95% confidence intervals of slope estimates frequently did not include 0.75, and clade‐specific slopes varied from 0.5 to 0.85, depending on the clade and regression model. Our approach reveals that the choice of analytical model has a systematic influence on the estimated allometry, but irrespective of the model applied, we find little support for a universal metabolic rate–body mass scaling relationship.