An inductive computer algorithm (ID3) was applied to a bootstrapped training set/test set design in which the model was developed from 70 percent of the sample and tested on the remaining 30 percent. Relative accuracy of sex classification was evaluated for seven subsets of 31 morphological features of the adult os coxae. Using 115 ossa coxarum selected from the Terry Collection, a selected suite of the three most consistently diagnostic attributes averaged 93.1 percent correct classification of individuals by sex over 10 trials. Attribute suites developed collaboratively with three well known skeletal experts averaged 87.8, 91.3, and 89.6 percent correct. The full set of 31 attributes averaged 90.0 percent accuracy. The study demonstrates a small set of three criteria, selected and ordered by ID3, that is more accurate than other combinations, and suggests that ID3 is a useful approach for developing identification systems. Tables, references