Current diagnosis of wrist joint problems is hampered by poor understanding of wrist joint functioning. To gain insight into the relationship between function and anatomy we aim at building a statistical shape model of the wrist. As a basis for the statistical model we intend to use CT scans of different individuals that are mutually registered to a common pose. Unconstrained registration of individual bones does not lead to an anatomically feasible arrangement of the carpal bones. We propose to use a constrained 3D registration, where bone movements are based on bone geometry and joint kinematics. Constrained co-registration of wrists from different individuals in different postures demonstrates that our method finds anatomically feasible wrists in the same posture that are therefore a sound basis for a statistical shape model.