Abduction strength forms an important component of shoulder function assessment tools such as the Constant-Murley score. However, there is no consistency in literature with regards to the method used for its assessment. An extensive review of the literature failed to reveal any evidence to support any particular method. The purpose of this study was to compare the reliability of a single measurement with that of the maximum and average of three consecutive measurements. Sixty healthy volunteers participated in this study. Three measurements of shoulder abduction strength were taken on two occasions separated by 2 weeks. For each shoulder, two single (first) measurement values, two average values and two maximum values were obtained. For each method, 95 % limits of agreement were calculated. Both methods average and maximum have higher reliability than a single measurement. Furthermore, the reliability of the maximum is slightly higher than that of the average.