The purpose of this study was to determine if the repeatability and pattern of elbow kinematics are affected by changing the relative magnitudes of loads applied to muscles around the elbow in vitro. In eight cadaveric upper extremities, passive and three methods of simulated active elbow flexion were tested with the forearm maintained in both pronation and supination. Passive flexion involved moving the elbow manually through a full arc of motion. Simulated active flexion used a custom designed loading system to generate elbow motion by applying loads to various tendons via pneumatic actuators. Three different simulated active loading protocols, with loading ratios based on muscle activity and physiologic cross-sectional area, were tested. Testing was performed initially on an intact elbow, and then an unstable elbow model created by transection of the lateral collateral ligament (i.e. the radial and lateral ulnar collateral ligaments). An electromagnetic tracking device was used to measure rotation of the ulna relative to the humerus. Varus–valgus angulation and internal–external rotation were less repeatable during passive flexion than simulated active flexion, regardless of the loading ratio used, in both the intact (p<0.05) and unstable (p<0.05) elbows. Throughout the arc of flexion, the motion pathways were similar for the three simulated active motion protocols employed in this study (p>0.05). The pathways followed during passive motion were different from those generated with simulated active motion, especially in the unstable elbow with the forearm supinated (p<0.001). These results suggest that using simulated active motion rather than manual passive motion can improve the repeatability of elbow kinematics generated in the laboratory, and that a wide range of muscle loading ratios may produce similar kinematic output.