Shoulder shape directly impacts forelimb function by contributing to glenohumeral (GH) range-of-motion (ROM). However, identifying traits that contribute most to ROM and visualizing how they do so remains challenging, ultimately limiting our ability to reconstruct function and behaviour in fossil species. To address these limitations, we developed an
in silico
proximity-driven model to simulate and visualize three-dimensional (3D) GH rotations in living primate species with diverse locomotor profiles, identify those shapes that are most predictive of ROM using geometric morphometrics, and apply subsequent insights to interpret function and behaviour in the fossil hominin
Australopithecus sediba
. We found that ROM metrics that incorporated 3D rotations best discriminated locomotor groups, and the magnitude of ROM (mobility) was decoupled from the anatomical location of ROM (e.g. high abduction versus low abduction). Morphological traits that enhanced mobility were decoupled from those that enabled overhead positions, and all non-human apes possessed the latter but not necessarily the former. Model simulation in
A. sediba
predicted high mobility and a ROM centred at lower abduction levels than in living apes but higher than in modern humans. Together these results identify novel form-to-function relationships in the shoulder and enhance visualization tools to reconstruct past function and behaviour.