Predictive use of modern reference osteological collections for disentangling the shape of Eurasian equid cheek teeth and metapodials in archaeological material

Abstract

AbstractEquids have shaped past Eurasian societies in many ways. This applies in particular to domestic horses, donkeys, and their hybrids. Key to documenting modes of exploitation and cultural trajectories in past societies is the correct taxonomic classification of tooth and bone specimens found in archaeological sites. However, close osteomorphological resemblance of wild and domestic equids and their economically valuable hybrids, that is, mules and hinnies, complicates the identification of intentionally fragmented or naturally damaged archaeological specimens. Here, we apply geometric morphometrics (GM) to mandibular teeth and metapodials, two skeletal elements commonly found in archaeological collections and known for their diagnostic properties using traditional morphometric methods. We registered a statistically representative set of 2D and 3D coordinates on mandibular teeth (P3, P4, M1, and M2) and metapodials of 92 domestic horses (Equus caballus Linnaeus, 1758), 70 domestic donkeys (Equus asinus Linnaeus, 1758), 30 hybrids, and 63 Asiatic wild asses (Equus hemionus Pallas, 1775). Taxonomic classification of these 255 specimens considered both shape and form, applying linear discriminant analysis, k‐nearest neighbors algorithm, and artificial neural networks to seven combinations of taxa. We obtained correct classifications in over 87% and 80%, respectively, of the premolars and molars and in over 93% and 89%, respectively, of the metacarpals and metatarsals. This modern dataset was then used to classify equid specimens from three archaeological sites in the Middle East already analyzed morphologically. Taking into account the past zoogeography of wild equids and the historical distribution of their domesticated descendants and hybrids, the GM approach presented in this study offers the possibility to morphologically classify archaeological equids with far greater certainty than has been the case so far.