A three-dimensional (3-D) biomechanical model was used to determine upper extremity kinematics of 16 male subjects with low-level paraplegia while performing wheelchair propulsion (WCP). A six-camera VICON motion analysis system was used to acquire the coordinate data of ten anatomic markers. Joint axes for the wrist and elbow were defined along with the planes of motion for the upper arm (humerus) and trunk. The group's mean and standard deviation profiles were graphed for eight of the nine rotations measured during WCP. Variability in the intercycle and intersubject movement patterns were calculated using the root mean square standard deviation (RMS sigma) and the coefficient of variation (CV). Motion pattern similarities were quantified using the coefficient of multiple correlation (CMC). The intercycle (Nc > or = 6) motion patterns of individual subjects were highly consistent, similar, and repeatable during WCP. This was confirmed by low CVc values (3-31%), high CMCc values (0.724-0.996) and RMS sigma c values below 3.2 degrees. For the group, mean values of the propulsion velocity, cadence, and propulsion cycle duration were 89.7 m/min, 66.1 pushes/min, and 0.96 s, respectively. Humeral plane and rotation showed large excursions (76.1-81.6 degrees), while trunk lean and forearm carrying angle displayed relatively small ranges of motion (5.5-10.9 degrees). The intersubject (N3 = 16) motion patterns were less similar compared to individual intercycle patterns. This was evidenced by higher CVc values (12-128%) and lower CMC3 values (0.418-0.935). Intersubject humeral patterns were the most consistent while trunk lean was the least consistent. Intersubject root mean square standard deviations (RMS sigma c) were more than three times the corresponding intercycle values for all nine rotations.