We present a compilation of endocranial volumes (ECV) for 176 non-human primate species based on individual data collected from 3813 museum specimens, at least 88% being wild-caught. In combination with body mass data from wild individuals, strong correlations between endocranial volume and body mass within taxonomic groups were found. Errors attributable to different techniques for measuring cranial capacity were negligible and unbiased. The overall slopes for regressions of log ECV on log body mass in primates are 0.773 for least-squares regression and 0.793 for reduced major axis regression. The least-squares slope is reduced to 0.565 when independent contrasts are substituted for species means (branch lengths from molecular studies). A common slope of 0.646 is obtained with logged species means when grade shifts between major groups are taken into account using ANCOVA. In addition to providing a comprehensive and reliable database for comparative analyses of primate brain size, we show that the scaling relationship between brain mass and ECV does not differ significantly from isometry in primates. We also demonstrate that ECV does not differ substantially between captive and wild samples of the same species. ECV may be a more reliable indicator of brain size than brain mass, because considerably larger samples can be collected to better represent the full range of intraspecific variation. We also provide support for the maternal energy hypothesis by showing that basal metabolic rate (BMR) and gestation period are both positively correlated with brain size in primates, after controlling for the influence of body mass and potential effects of phylogenetic relatedness.