Using the covariation of extant hominoid upper and lower jaws to predict dental arcades of extinct hominins

Upper and lower jaws are well represented in the fossil record of mammals and are frequently used to diagnose species. Some hominin species are only known by either their maxillary or mandibular morphology, and in this study, we explore the possibility of predicting their complementary dental arcade shape to aid the recognition of conspecific specimens in the fossil record. To this end, we apply multiple multivariate regression to analyze 3D landmark coordinates collected on associated upper and lower dental arcades of extant Homo, Pan, Gorilla, Pongo, and Hylobates. We first study the extant patterns of variation in dental arcade shape and quantify how accurate predictions of complementary arcades are. Then we explore applications of this extant framework for interpreting the fossil record based on two fossil hominin specimens with associated upper and lower jaws, KNM-WT 15000 (Homo erectus sensu lato) and Sts 52 (Australopithecus africanus), as well as two non-associated specimens of Paranthropus boisei, the maxilla of OH 5 and the Peninj mandible. We find that the shape differences between the predictions and the original fossil specimens are in the range of variation within genera or species and therefore are consistent with their known affinity. Our approach can provide a reference against which intraspecific variation of extinct species can be assessed. We show that our method predicts arcade shapes reliably even if the target shape is not represented in the reference sample. We find that in extant hominoids, the amount of within-taxon variation in dental arcade shape often overlaps with the amount of between-taxon shape variation. This implies that whereas a large difference in dental arcade shape between two individuals typically suggests that they belong to different species or even genera, a small shape difference does not necessarily imply conspecificity.