Body mass estimates are integral to a wide range of inferences in paleoanthropology. Most techniques employ postcranial elements, but predictive equations based on cranial variables have also been developed. Three studies currently provide regression equations for estimating mass from cranial variables, but none of the equations has been tested on samples of known mass. Nor have the equations been compared to each other in terms of performance. Consequently, this study assessed the performance of existing cranial equations using computed tomography scans from a large, documented sample of modern humans of known body mass. Virtual models of the skull were reconstructed and measured using computer software, and the resulting variables were entered into three sets of published regression equations. Estimated and known body masses were then compared. For most equations, prediction errors were high and few individuals were estimated within ±20% of their known mass. Only one equation satisfied the accuracy criteria. In addition, variables that had been previously argued to be good predictors of mass in hominins, including humans, did not estimate mass reliably. These results have important implications for paleoanthropology. In particular, they emphasize the need to develop new equations for estimating fossil hominin body mass from cranial variables.