This study compares the effects of three modes of isokinetic resistance training at the shoulder--concentric, eccentric, and a combination of both concentric and eccentric, with a group that received no training at all. Twenty-eight healthy volunteers (male and females), 18 to 36 years of age, with no history of shoulder pathology, were randomly assigned to one of four groups; concentric training, eccentric training, a combination of both concentric and eccentric training, or control (no training). Testing and training of the dominant shoulder was performed on an isokinetic dynamometer. All subjects were pretested and post-tested both concentrically and eccentrically for humeral internal/external rotation and abduction at speeds of 60 degrees/sec and 120 degrees/sec. Each training session consisted of a total of twelve sets of ten maximal repetitions, and was repeated three times a week for four weeks. The absolute and percent difference in peak force and peak torque for each group between the pretest and post-test was calculated for each combination of position, mode, and speed. A significant difference between the concentric/eccentric group and the eccentric group was found for abduction (p < 0.05). The eccentric group showed a significantly greater increase from the pretest to post-test for absolute differences in peak force and peak torque compared to the concentric/eccentric group (p < 0.05).