Pattern expressions of morphoscopic cranial traits vary across populations with classification accuracies being highly dependent on the reference collection to which unknown skulls are compared. Despite recent developments in population-specific standards for South Africans, researchers have not addressed the accuracy of morphological methods. Several studies demonstrate differences in sexual dimorphism between South Africans and North Americans, warranting a need to re-evaluate sex estimation methods in South Africa. The purposes of this study were to test the reliability and accuracy of the Walker (2008) method and to examine patterns of sexual dimorphism among South Africans. A total of 245 modern Black and White South African male and female crania from the Pretoria Bone Collection, University of Pretoria, were scored using the Walker (2008) methodology. Cohen's kappa was used to evaluate reliability of the method, and percent correct assessed validity of the method. Logistic regression was utilised to create modified population-specific formulae. Inter- and intra-observer agreement was moderate to excellent (0.60-0.90), except for the mental eminence (0.40). The percent correct results for sex were 80% or higher for combinations of glabella, mastoid and menton and between 68% and 73% for menton, mastoid, orbital and nuchal margin using logistic equations of Walker (2008). White males had the highest (94-97%) and White females had the lowest (31-62%) percent correct. The low accuracies obtained when using Walker's (2008) equations emphasised the need for population-specific sex estimation models. Modified formulae for South Africans were created, yielding higher classification rates (84-93%) than when North American standards were employed.