Craniometric determination of population affinity in South Africans

A vital aspect of skeletal analysis is the determination of population affinity of an unknown individual. The aim of this paper is to develop discriminant function formulae to determine race from craniometric dimensions of South African blacks and whites. Skeletons used in this study came from the Universities of the Witwatersrand and Pretoria. The sample is composed of 53 white males and 53 white females and 45 black males and 45 black females. Using 13 standard cranial and 4 mandibular dimensions, average accuracies of 98% were obtained from the crania, which were much more discriminatory than the mandibles (74% males, 87% females). When a "leave-one-out classification" technique was applied to the sample to measure accuracy of multivariate classification, this accuracy was about the same as obtained from the multivariate function. A posterior probability of 0.80 or more was found in as much as 96% of the sample. Stepwise discriminant function formulae for incomplete remains (vault and face) were also derived. Prediction accuracy was considerably lower when North American based formulae were tested on the South Africans, indicating significant craniometric differences between these populations.