A reassessment of single long bone dimensions to estimate sex in adult South African skeletal remains

Abstract

The use of postcraniometrics has been popularised in forensic anthropology to find landmark measurements and methods that provide high classification rates, high reliability, and repeatability in estimating demographic parameters. In 2000, Loth and İşcan utilised bony landmarks from the humerus, tibia and femur to develop univariate sectioning points to estimate sex in South African Black and White populations. The accuracy of the Loth and İşcan (2000) method has, however, not been validated. This study, therefore, aimed to validate the accuracy of the Loth and İşcan (2000) sectioning points on a Black and White South African sample. Given the small sample size of the Loth and İşcan (2000) study, this study further aimed at developing new sectioning points using a much larger sample of South African Black and White individuals, as well as adding sectioning points for South African Coloured individuals. In addition to creating new sectioning points, this study aimed to develop multivariate formulae utilising only the measurements defined in Loth and İşcan (2000) by employing flexible discriminant analysis (FDA) and stepwise discriminant function analyses (DFA). Applying the Loth and İşcan (2000) sectioning points to the current sample reduced accuracy to 80,75% for Black South Africans and 78,63% for White South Africans. Tibial distal breadth (TDB) misclassified all Black females and correctly classified only 2,00% of White females. The newly developed sectioning points outperformed the original ones. In multivariate analysis, both FDA and DFA performed well, with FDA performing slightly better. FDA accuracies ranged from 87,4% to 90,9%, while DFA ranged from 86,9% to 91,25%, with the highest accuracy for Black South Africans. Overall, the Loth and İşcan (2000) method showed reduced accuracy when applied to a larger South African sample, particularly for females. This study enhanced sex estimation by refining sectioning points and introducing multivariate formulae, achieving classification rates of 88,50–94,00% for Black, 84,50–90,50% for White, and 84,20–90,79% for Coloured South Africans.