The selection of a joint coordinate system affects the outcome of motion pathways. We developed coordinate systems for the ulna and humerus, which are generated from upper extremity motion. These Motion-Derived Coordinate Systems (CS) were compared to traditional Anatomy-Derived CS created using surface digitizations of anatomical features. Within-subject repeatability of creating MotionDerived CS was quantified. In vitro elbow flexion was generated in the gravity-dependent position using an active upper extremity motion simulator. Kinematic pathways of those motions were calculated in terms of valgus angulation and internal rotation of the ulna relative to the humerus, using both CS. The method of creating Motion-Derived CS was highly repeatable-less than 0.5 mm and 18 for all coordinate directions measured. Inter-subject variability of active flexion pathways was reduced with Motion-Derived CS compared to AnatomyDerived CS ( p < 0.05). The decrease in inter-subject kinematic variability when using Motion-Derived CS may increase the statistical power of biomechanical studies and allow for reduced sample sizes. This minimally invasive method, which also determines the elbow flexion and forearm rotation axes and center of the capitellum, may also be applicable in computer-navigated surgery of the upper limb. ß