Sex estimation in forensic anthropology: a test of the Klales et al. (2012) method with implications of asymmetry

Abstract

A sample of 204 American individuals was examined to assess the accuracy and reliability of the three non-metric traits described by Phenice (1969) and revised by Klales et al. (2012) for assigning sex. In addition, the bilateral stability of the three non-metric traits was assessed to determine if asymmetrical expression of the traits compromises the classification accuracy of the revised method, since a prior study found that application of Phenice’s original technique yielded low classification accuracy when applied to the right innominate. Klales and colleagues claimed that expansion of the classification system from a dichotomous present/absent scale into five character states and the incorporation of logistic regression based on posterior probabilities vastly improves the accuracy rates for correct sex identification over the original method. Validity of the method developed by Klales and colleagues has not been tested by an external observer on a modern sample of American individuals (individuals who have died within the last 50 years). The current study tests the reliability and validity of Klales et al.’s (2012) technique for assigning sex of both the left and right innominate. Validity was tested using the sample of innominates 204 individuals from the William Bass Skeletal Collection housed at the University of Tennessee, Knoxville. Intra- and interobserver agreement was evaluated for Klales and colleagues’ method. Intra-observer and interobserver agreement was statistically evaluated with Cohen’s weighted kappa and the intra-class correlation coefficient. A series of Wilcoxon matched-pairs signed-ranks tests were used to evaluate statistical differences in the trait scores between the left and right innominates. Results show that the Klales et al. (2012) technique yields moderate to high levels of intra- and inter-observer agreement and yields correct sex identifications among individuals of known-sex in 93.6% of cases when all three traits are combined. Accuracy of correct sex identification was further increased to 99% by re-calibrating the logistical regression equation to fit the sample obtained from the William Bass Skeletal Collection. A Wilcoxon matched-pairs signed-ranks test revealed a statistically significant difference in trait scores of the ventral arc between European and African Americans; however, this difference does not compromise the accuracy of the method for correct identification of sex in known-sex individuals.