Marsupials display far less forelimb diversity than placentals, possibly because of the laborious forelimb-powered climb to the pouch performed by most marsupial neonates. This is thought to result in stronger morphological integration (i.e., higher co-variance) within the marsupial forelimb skeleton, and lower integration between marsupial fore- and hind limbs, compared to other mammals. Possible mechanisms for this constraint are a fundamental developmental change in marsupial limb patterning, or alternatively more immediate perinatal biomechanical and metabolic requirements. In the latter case, peramelid marsupials (bandicoots), which have neonates that climb very little, should show lower within-limb and higher between-limb integration, compared to other marsupials. We tested this in four peramelid species and the related bilby, using partial correlation analyses of between-landmark linear measurements of limb bones, and Procrustes-based two-block partial least-squares analysis (2B-PLS) of limb bone shapes using the same landmarks. We find extensive between-limb integration in partial correlation analyses of only bone lengths, consistent with a reduction of a short-term biomechanical/allocation constraint in peramelid forelimbs. However, partial correlations of bone proportions and 2B-PLS reveal extensive shape divergence between correlated bone pairs. This result contradicts expectations of developmental constraints or serial homology, instead suggesting a function-driven integration pattern. Comparing visualizations from cross-species principal components analysis and 2B-PLS, we tentatively identify selection for digging and half-bounding as the main driver of bandicoot limb integration patterning. This calls for further assessments of functional versus developmental limb integration in marsupials with a more strenuous neonatal climb to the pouch.