Assessing the reliability of virtual reconstruction of mandibles

Abstract

Mandibular morphological variation is often used to examine various aspects of human palaeobiology. However, fossil and archeological skeletal remains are often fragmented/distorted and so are frequently excluded from studies. This leads to decreased sample sizes and, potentially, to biased results. Thus, it is of interest to restore the original anatomy of incomplete/distorted specimens. Thin plate splines (TPS), commonly used in Geometric Morphometrics (GM), offer the prospect of reconstruction of missing parts and particularly of interest here, missing landmarks. Here, the reliability of TPS based mandibular reconstruction is tested. To that end missing landmarks were simulated in originally complete hemimandibles. TPS was then used to restore the location of simulated missing data and the predicted landmarks were compared to the original ones. Results show that error varies according to the number and location of estimated landmarks. Notwithstanding, estimation error is usually considerably smaller than the morphological differences between individuals from the same species. TPS based reconstruction allows fragmentary mandibles to be used in studies of whole mandibular variation, provided the above mentioned caveats are considered.