The purpose of this study was to measure the contributions of the motions of body segments and joints to racquet head speed during the tennis serve. Nine experienced male players were studied using three-dimensional film analysis. Upper arm twist orientations were calculated with two alternative methods using joint centres and skin-attached markers. The results showed that skin-attached markers could not be used to calculate accurate upper arm twist orientations due to skin movement, and that the use of joint centres produced errors of more than 20 degrees in the upper arm twist orientation when the computed elbow flexion/extension angle exceeded 135 degrees in the final 0.03 s before impact. When there were large errors in the upper arm twist orientation, it was impossible to obtain accurate data for shoulder or elbow joint rotations about any axis. Considering only the contributors that could be measured within our standards of acceptable error, the approximate sequential order of main contributors to racquet speed between maximum knee flexion and impact was: shoulder external rotation, wrist extension, twist rotation of the lower trunk, twist rotation of the upper trunk relative to the lower trunk, shoulder abduction, elbow extension, ulnar deviation rotation, a second twist rotation of the upper trunk relative to the lower trunk, and wrist flexion. The elbow extension and wrist flexion contributions were especially large. Forearm pronation made a brief negative contribution. Computed contributions of shoulder internal rotation, elbow extension and forearm pronation within the final 0.03 s before impact were questionable due to the large degree of elbow extension. Near impact, the combined contribution of shoulder flexion/extension and abduction/adduction rotations to racquet speed was negligible.