Gadi Herzlinger, Yoel Rak, Patricia Smith, Uzy Smilansky
Mandibular ramus morphology is central to paleoanthropology, yet many quantitative workflows emphasize whole-shape summaries and do not explicitly report the intrinsic geometry of particular anatomical curves and surfaces. We present an automated pipeline that extracts key features of the notch-condyle complex from three-dimensional surface meshes, positions specimens in anatomically defined reference frames, and outputs standardized object-level measurements suitable for reproducible comparison and downstream statistics. We demonstrate the approach on a mixed sample of Pleistocene and Holocene hominin mandibles/rami. Correlation structure and principal component analysis are used to summarize covariation, and permutation-based tests evaluate whether predefined groups differ in multivariate space. Overall, the descriptors show limited redundancy, with only a small subset exhibiting moderate to strong coupling, indicating that the pipeline captures multiple, partly independent aspects of local ramus geometry. The principal component analysis indicates that variability is distributed across several dimensions rather than dominated by a single axis. In the first two components, Holocene specimens form a relatively compact cluster, whereas Pleistocene specimens are more broadly dispersed, with substantial overlap among fossil groups. Despite this overlap, permutation-based testing in reduced principal component space detects a significant effect of group membership, consistent with a conservative Holocene-Pleistocene shift combined with marked within-group dispersion in the fossil sample.