Mixed-data Morphometric Classification of Seima–Turbino Stone Projectiles: Hafting Templates and Inter-site Variability across Bronze Age Eurasia

Abstract Stone projectiles are the most abundant but least systematically classified artefacts in the Seima–Turbino phenomenon, a Bronze Age network spanning northern Eurasia. This study develops a reproducible, image-based morphometric workflow for 420 complete specimens from eight Seima–Turbino sites and two comparative assemblages. Closed outlines extracted from published site-report illustrations were analysed with Elliptical Fourier Analysis (EFA; 24 harmonics), local curvature metrics, four categorical technological attributes and mixed-data principal component analysis (PCAmix), then partitioned with Partitioning Around Medoids (PAM) clustering and evaluated through decision-tree classification. The resulting classification identifies eleven morpho-technological types, recoverable with 84.3% cross-validated accuracy, organised around two hafting templates: tanged and non-tanged. Site-level disparity and size variability distinguish tightly standardised assemblages from broader repertoires, suggesting different degrees of template control, production scale and depositional aggregation. Inter-site morphological distances show no significant decay with geographic distance across approximately 2,000 km, while a provisional short-range negative autocorrelation suggests that projectile forms circulated through selective interaction pathways rather than simple isotropic diffusion. Cross-reference with published ancient DNA data further indicates that the most widespread type occurs among individuals of three distinct ancestries at Rostovka. The study provides a transferable framework for analysing legacy lithic datasets where direct remeasurement is impractical.