Estimating Sex of South African Crania Using 3D-ID

Abstract

The freely available software program 3D-ID estimates sex and population affinity using geometric morphometric methods and a large reference database (N > 2000). However, the 2018 version lacked representation of South African populations. This study evaluated the accuracy of 3D-ID (2018 version) for sex estimation in the three largest South African samples. A total of 450 crania (75 males and 75 females per population) from South Africans of African descent (SAA), South Africans of Multi-populations (SAM), and South Africans of European descent (SAE) were digitized using a Microscribe 3DX digitizer. Landmark data were analysed in 3D-ID using both Full (all landmarks) and Reduced (five landmarks excluded) configurations, with assessments using both shape and form (shape + size) variables. When populations were combined, females consistently yielded higher classification accuracies than males. Using the Reduced configuration, sex estimation accuracies for shape and form were higher for SAA females (82.7% and 89.3%) and SAE females (88.0% and 89.3%), and moderate to high for SAE males (76.0% and 85.3%). In contrast, SAA males showed lower accuracies (44.0–69.3%), while SAM individuals generally fell below 75%. Overall, 3D-ID (2018 version) demonstrated moderate to high sex estimation accuracy. However, validation of the newly incorporated South African reference data (3D-ID 2024 version) is recommended before forensic application.