Craniometric studies of South Africans yield high accuracies of sex and ancestry classification, but most assess only inter-group variation of Black and White individuals, excluding the highly heterogeneous Colored group, which constitute a significant proportion of the population. This study applied a geometric morphometric approach to the neurocrania of 774 Black, Colored, and White individuals to assess sex and ancestry estimation accuracy based on the detected morphological variation. Accuracies of 70% and 83% were achieved for sex and ancestry, respectively, with ancestry-related variation contributing the largest proportion of overall observed variation. Even when comparing the closely related Black and Colored groups, relatively high accuracies were obtained. It is thus recommended that a similar approach be used to develop a contemporary three-dimensional database, which can be used to objectively, reliably, and accurately classify unknown remains in the South African forensic context.