Muscle force values and tendon excursions across the glenohumeral joint during unconstrained glenohumeral abduction in the scapular plane were evaluated with a dynamic shoulder testing apparatus. This evaluation was achieved by simulated rotator cuff and middle deltoid activity applied in four plausible muscle force ratios: (1) equal force to each tendon, (2) 2:3 ratio of force applied to the middle deltoid/supraspinatus tendons, (3) 3:2 ratio of force applied to the middle deltoid/supraspinatus tendons, and (4) zero force applied to the supraspinatus tendon to simulate supraspinatus paralysis. The glenohumeral joint was then moved to 5 degrees, 15 degrees, 30 degrees, 45 degrees, 60 degrees, and maximum glenohumeral abduction while muscle forces, tendon excursions, and glenohumeral joint kinematics were monitored. Full glenohumeral abduction was achieved in all four test conditions. When the muscle force combination favored the middle deltoid, the smallest supraspinatus force was required from 30 degrees to maximum glenohumeral abduction; however, when the supraspinatus was favored, the largest supraspinatus force was necessary to achieve maximum glenohumeral abduction. With simulated supraspinatus paralysis the middle deltoid required the greatest increase in force from 15 degrees through 45 degrees of glenohumeral abduction. These results indicate that muscle efficiency during glenohumeral abduction is highly dependent on the ratio of applied force between the middle deltoid and supraspinatus. A larger contribution of force from the supraspinatus was required near the beginning of motion, whereas the middle deltoid was more important near the end of glenohumeral abduction in the scapular plane. Tendon excursion for the middle deltoid (6.4 +/- 0.2 cm) and supraspinatus (3.8 +/- 0.2 cm) were proportionately larger than those for the subscapularis and infraspinatus. Humeral head translations on the glenoid were less than 2 mm in all four conditions evaluated; therefore the glenohumeral joint behaves kinematically as a "ball-and-socket" articulation during glenohumeral abduction. Simulated supraspinatus paralysis does not change normal joint kinematics and does not prevent full glenohumeral abduction.