It is not known how well skin markers represent the skeletal knee joint motion during running. Hence the purpose of this investigation was to compare the skin marker derived tibiofemoral motion with the skeletal tibiofemoral motion during running. In addition to skin markers attached to the shank and thigh, triads of reflective markers were attached to bone pins inserted into the tibia and femur. Three-dimensional kinematics of the stance phase of five running trials were recorded for three subjects using high-speed cine cameras (200 Hz). The knee motion was expressed in terms of Cardan angles calculated from both the external and skeletal markers. Good agreement was present between the skin and bone marker based knee flexion/extension. For abduction/adduction and internal/external knee rotation, the difference between skeletal and external motion was large compared to the amplitude of these motions. Average errors relative to the range of motion during running stance were 21% for flexion/extension, 63% for internal/external rotation, and 70% for abduction/adduction. The errors were highly subject dependent preventing the realization of a successful correction algorithm. It was concluded that knee rotations other than flexion/extension may be affected with substantial errors when using skin markers.