Technical Note: Reliability of Pöch's facial shape classification system: A morphometric approach

A standardized vocabulary of facial characteristics is essential in forensic anthropology for identifying both living and deceased individuals. Pöch's facial shape classification system is widely used for this purpose, but its reliability has not been critically evaluated. This is particularly important as the system was originally developed based on European male populations, and prior research suggests that classification systems can be prone to observer inconsistencies. In this study, we evaluate the representativeness and reliability of Pöch's system in describing face shape variation within a Mexican sample (n = 973). First, using geometric morphometrics and principal component analysis, we evaluated how well the system's categories capture the observed facial diversity. Sixteen observers then classified 60 randomly selected faces twice using Pöch's system, and we computed Fleiss' Kappa and Cohen's Kappa to assess intra- and inter-observer agreement. The level of consensus among observers was also measured by calculating the proportion of agreement for each facial shape. Our findings revealed significant morphological redundancy among most of Pöch's categories, poor agreement within (mean Cohen's Kappa = 0.203) and between (mean Fleiss' Kappa = 0.112) observers, and only the round and oval categories showed high levels of consensus. Given the system's poor performance, we conducted a post-hoc k-means clustering analysis of the Procrustes coordinates to explore a more representative classification scheme. The resulting four facial shape categories could inform the development of more representative and reliable classification systems.