Hidden Information in the Fossil Record: Using Discriminant Analysis on Isolated Postcrania

Presenter: Brianna McHorse (Biology)

Mentor: Samantha Hopkins

Facilitator: Chris Moe

Postcrania (non-skull or teeth bones) are often preserved in the fossil record but, unless found with teeth or skulls, are rarely identified beyond the family level. As a result, they offer a potentially untapped resource for studies of extinct diversity. Discriminant statistical analyses of linear measurements on these postcranial bones show remarkably high identification success rates for many mammal types, including antilocaprids (pronghorn), camelids (camels and llamas), and equids (horses). The approach we use is ideal, as it captures more subtle bone-shape variation than examining scatterplots of measurements but is more straightforward than three-dimensional morphometric methods. Further, applying Bayesian methods to the established discriminant analysis can allow integration of multiple skeletal elements, e.g., phalanges (fingers), astragali (ankles), and metapodials (hand and foot bones). We test this new method on a known, artificially created assemblage of modern cervid (deer), camelid, and antilocaprid postcranial bones. In a mixed training set of four bone types, we achieved identification success rates ranging from 87.5% to 100%. Our method is simple but has the potential to quickly and significantly improve knowledge of the hoofed mammal ecology at several postcrania-rich fossil sites. We focus on hoofed mammals, but the method should transfer well to other mammalian groups, shedding light on hidden diversity and improving any studies that rely on identification.