Twenty-six patients with Grade III acromioclavicular joint separations were evaluated to determine the outcomes of nonoperative and operative management. Evaluation consisted of a detailed functional questionnaire, physical examination, and comprehensive isokinetic strength assessment. The patients were divided into two groups: operative (n = 16) and nonoperative (n = 10). Operative management consisted of coracoclavicular stabilization with heavy suture material and with nine of the sixteen patients treatment also consisted of coracoacromial ligament transfer and lateral clavicle resection. Nonoperative management consisted of short-term immobilization with early range of motion and rehabilitation. The two groups were similar in all characteristics except mean age: 30.7 years for the operative group and 49.6 years for the nonoperative group. Follow-up evaluation was performed an average of 32.9 months after either injury (nonoperative group) or surgery. Our results indicated that nonoperative management was superior to operative management with respect to time to return to work (0.8 months vs. 2.6 months), time to return to athletics (3.5 months vs. 6.4 months) and time of immobilization (2.7 weeks vs. 6.2 weeks). However, operative management was superior to nonoperative management in the following parameters: time to attain completely pain-free status, the patient's subjective impression of pain, range of motion, functional limitations, cosmesis, and long-term satisfaction. There were no significant differences between the two groups with respect to shoulder range of motion, manual muscle testing, or neurovascular findings. Isokinetic strength testing of the involved shoulder, expressed as a percentage of the uninvolved shoulder, showed no significant differences in peak torque, total work, or total power between the operative and nonoperative groups. However, comparison of the involved to the uninvolved extremity within each group did reveal a trend toward decreased peak torque, work, and power for abduction in the involved extremity regardless of the treatment used. These findings reached statistical significance only for power at the slower testing speed (60 degrees/sec). There was also a significant decrease in power in the involved extremity for external rotation at the faster speed (120 degrees/sec) in the nonoperative group. Finally, the absolute values for peak torque, work, and power were significantly greater for all motions tested in the operative group as compared to the nonoperative group. This may reflect the difference in age between the two groups. Based upon the patients studied, there are benefits to both nonoperative and operative methods of treatment of Grade III acromioclavicular separations. Recovery of strength did not differ between the two groups and therefore should be viewed as a less important factor in patient selection for operative versus nonoperative management. Careful patient selection should remain an important aspect of treatment for this controversial injury.