Estimating Ancestry Among South African Ethnic Groups

Abstract

The objective of this research project is to assess craniometric differences among the socially defined black South African groups, namely Zulu, Sotho, Pedi, Venda, Tshwane, Tsonga, Swazi, Xhosa, and Ndebele. Current ancestry estimation methods pool these groups into the broad category of black South Africans. This general description of the population as ‘black South Africans’ may be problematic as various ethnic groups comprise this broad classification. The refinement of self-identification based on ethnicity may improve biological profiles, possibly improving the identification of missing persons from their skeletal remains.

A total of 365 male, adult crania of black South Africans were selected from the Pretoria Bone Collection, the University of Pretoria, and the Raymond A. Dart Collection, at the University of Witwatersrand. Eighty-five standard cranial landmarks were collected using the 3Skull programme and a Microscribe G2 digitizer (Ousley, 2004). The technical error of measurement (TEM) displayed great intra- and inter-observer agreement. An analysis of variance (ANOVA) was conducted, and twenty-three measurements were found to be statistically significant. A Tukey’s Honest Significant Difference (HSD) post hoc test was conducted using the statistically significant cranial measurements and demonstrated that the midface and occipital bones had the most intergroup differences, while the lambda-subtense fraction (OCF) had the most post-hoc pairwise comparisons. The most prevalent inter-group difference was observed between the Swazi and Sotho group. Discriminant Function Analysis (DFA) assessed relationships between size among the groups. A stepwise selection was used to obtain the variables that were best at separating the different groups in the different DFA models. The groups were tested individually, based on geographical location and historical linguistic lineage clusters. The skull was subdivided into the cranial vault, cranial base and splanchnocranium. The various DFA models had overall model classification accuracies that were greater than chance, but their percentages were not high enough to be used for classification purposes in a forensic setting. Clustering the different groups based on their geographical location and historical linguistic lineages resulted in higher overall DFA model classification accuracies than when the groups were assessed as separate groups. The use of historical linguistic lineages may possibly be an alternative manner to refine the black South African classification.