Motion of the shoulder is integral to the sport of swimming; to date, accurate kinematic measurement of this motion is poorly reported in the literature. To fully understand the implications of shoulder movement patterns on issues such as performance or injury in swimming a suitable tool for the measurement of this movement must be developed. The system involved six cameras, three above water and three below water, which were focussed on a 2-metre calibration cube floating in the pool (one metre above water and one metre below water). This cube supplied the two calibration spaces required to digitise points above and below water simultaneously. An 85cm calibration cube was moved within the larger calibration space with points above and below water. This smaller cube was filmed, then analysed to determine the validity and accuracy of the three-dimensional analysis system. Five elite swimmers who were appropriately marked were videotaped swimming five sub-maximal non-breathing trials through the calibrated space. This video was then manually digitised to create a model which gave measures of humeral elevation, the plane of elevation, rotation and elbow flexion throughout the swim stroke. One trial of one subject was analysed nine times to ascertain the amount of error introduced by the digitising procedure (reliability). The four subjects` five trials and one subject`s four trials were subsequently analysed to determine the variability between repeated trials (repeatability). The data supplied by the repeatability trial formed the basis of the description of the pattern of motion of the humerus with respect to the thorax during the freestyle swim stroke. Results The system developed for this study proved to be valid and accurate (Standard error of measurement, SEM 0.6º-0.8º). The reliability was good with average SEM`s ranging from 4º- 12º and the coefficients of multiple correlation (CMC) ranging from 0.95- 0.99 for the different kinematic variables. There was a slight but expected degradation of these values for the repeatability trial with the average SEM`s ranging from 9º- 17º and CMC`s from 0.85- 0.96. These analyses suggest that this method of three-dimensional kinematic analysis, although sensitive enough to determine differences from side-to-side in a single swimmer, did generate significant errors. The subsequent examination of these kinematic graphs needs to take these errors into consideration when interpreting the results. Discussion Finally, the developed system supplied graphs that described the kinematic pattern of the humerus with respect to the thorax during the freestyle swim stroke. This information adds to and enhances the understanding of the body of literature on the biomechanics of swimming. It also supplies a tool to be used for ongoing research in the area of threedimensional analysis of motion of swimming.