Long bone morphometrics for human from non-human discrimination

Eight traditional measurements were taken on a sample of 50 human and at least 10 of each of the five animal species; measurements were then analysed using linear discriminant analysis (LDA). The results expectedly indicate enough variation between species to correctly assign an unknown bone as that of a human or non-human, with cross-validated classification accuracy of 95% or better. More importantly, however, the technique also proved to be accurate if only a fragment of the diaphysis is analysed; classification accuracy 63-99%. The results of this study, therefore, outline a forensically useful non-invasive method to distinguish human from animal bones.