This study tested the ability of basic skeletal measurements to classify single bony elements into 21 North American faunal species.
Physical anthropologists commonly encounter faunal remains at field sites and in the laboratory, and the majority of remains presented to forensic anthropologists are non-human. Without an extensive zooarchaeological or comparative anatomical background, determining the non-human species from an isolated bone can be challenging. In the current study, 58 length and breadth measurements were defined that could be collected reliably across 21 medium-to-large sized mammalian species (16 genera, 9 families). Data were collected from available humeri, radii, ulnae, femora, tibiae, fibulae, scapulae, ossa coxae, and fused metapodials from 652 individuals (~400 per element). Stepwise discriminant function analyses with a leave-one-out cross-validation were performed to evaluate accuracy in assigning each bone to its appropriate order, family, genus, and species. Species classification accuracies for long bones that were characterized by at least four measurements ranged from 75.8 to 92.0 percent, performing better than elements represented by fewer measurements. The humerus and femur were among the most reliable and higher performing elements, and when the two elements were combined species classification was 90.0 percent. As would be expected, taxa exhibiting greater size and morphological variation (e.g., Bovidae), and those with morphological similarities (e.g., dog vs wolf and coyote) displayed higher classification errors. The high classification rates, despite discriminating 21 species, suggest that a few basic measurements are highly useful for faunal species identification and may be used by physical anthropologists with limited faunal expertise to narrow down potential species. (publisher abstract modified)