This study uses metrical characteristics of the patella to derive population specific equations for sex determination in South Africa. Six measurements were taken from 120 normal and undamaged left patellae in a sample of known race, equally distributed by sex, obtained from the Raymond A Dart collection of human skeletons. These data were subjected to discriminant analysis. Maximum height and maximum breadth were selected in the stepwise analysis with an average accuracy of 83% in correct sex classification. Four functions were derived from the direct analysis with a range of average accuracy between 77% and 85%. While most individual variables showed high misclassification rates and may not be useful on their own, maximum height and maximum breadth have an average accuracy of 85 and 79%, respectively.