Estimating Body Mass In Biological Anthropology: An Evaluation Using Three-Dimensional Computed Tomography

Estimates of body mass are essential to biological anthropology research. The primary source for such estimates is skeletal morphology, and several predictive equations have been developed for cranial and postcranial material. These equations are widely used, but a number of factors suggest that they may not be as reliable as they are generally assumed to be. In particular, reference samples are often small and analyses frequently employ indirect measurements, specimens without accompanying body mass values, or mean data. In addition, tests of the equations have rarely involved external validation with samples of known mass.This project addressed these issues through three studies, using a large sample of modern humans for which both body masses and skeletal measurements were available. The sample consisted of Swiss forensic cases whose skeletal measurements were reconstructed from whole-body computed tomography scans. The first study compared the accuracy of three sets of commonly employed cranial equations. The second assessed published postcranial equations and compared the results to previous evaluations that had used less robust test samples. Several expectations regarding the performance of the equations were also tested. The third study employed the same sample to develop and test new regression equations for estimating mass from cranial and postcranial variables. The study was designed to compare the relative utility of the cranial and postcranial equations and to test the effect of variable choice, statistical method, and evaluation criteria on estimation competence. Results suggest that body mass estimates should be used more cautiously than is usually the case. Overall, cranial equations did not estimate mass accurately. Several that have been deemed to be reliable in previous studies, did not perform well. Postcranial equations estimated mass more accurately, but not consistently. They also did not necessarily perform in accordance with statements in the literature. Deriving new equations using a known reference sample improved estimation competence compared to previous studies, but accuracy rates remained relatively low. Key assumptions about the best criteria to use for evaluating predictive competence were not supported. Further research may explain these discrepancies, but until then, estimates generated with currently published equations should be treated as “ballpark figures”.