Isokinetic dynamometry is used in the assessment and rehabilitation of shoulder function in tennis players. The aim of this study was to validate a newly installed Biodex III Isokinetic Dynamometer for internal and external shoulder rotation. A non-injured male performed this movement which replicates an integral section of the tennis service action (Cohen et al., 1994). Three maximal trials were performed at six randomly assigned angular velocities (0.52, 1.05, 1.57, 2.09, 2.62 and 3.14 rad x s(-1)) with 90 s rest between each set. Trials were recorded using an on-line motion analysis system and kinematic data were obtained. Kinematic angular displacement of the lever arm was slightly greater (> or = 0.01 rad x s(-1)) than for the isokinetic dynamometer at all test angular velocities. Mean angular velocities from the kinematic data were almost identical to those from the isokinetic dynamometer but less than the target values, and this difference (e.g. 2.55 rad x s(-1) at 3.14 rad x s(-1)) was greater at higher angular velocities owing to the greater acceleration phases required. Peak angular velocity was similar on the isokinetic dynamometer to the target values, but substantially less than those from the kinematic data (e.g. 0.68 rad x s(-1) at 0.52 rad x s(-1)). This suggests that the isokinetic data are over-smoothed and may mask important information. In summary, criterion validity of this isokinetic dynamometer is supported for displacement and mean angular velocity, but not for peak angular velocity. Shoulder rotations in tennis are often at angular velocities greater than the Biodex can reproduce, but for the purpose of monitoring the shoulder strength and range of motion of tennis players and providing safe functional rehabilitation, the use of slower angular velocities is acceptable.