The role of shoulder muscles during passive, active, and resistive phases of shoulder rehabilitation exercises was investigated in ten normal subjects with no history of shoulder pathology. Using the scapular plane as a reference, three-dimensional motion of the shoulder was recorded with a computer-aided motion analysis system (VICON) to determine total shoulder elevation. Simultaneously, electromyographic data were acquired on nine shoulder muscles while performing the three phases of shoulder rehabilitation exercises as described by Neer. Fine wire intramuscular electrodes were placed in the following muscles: trapezius, serratus anterior, deltoid (anterior, middle, and posterior separately), supraspinatus, infraspinatus, biceps, and latissimus dorsi. Phase I (passive) exercises performed in the supine position showed the least electromyography (EMG) activity. There was a gradation of EMG activity as one progressed from Phase I (passive) to Phase II (active) to Phase III (resistive) shoulder exercises. Isometric exercises and Phase III resistive exercises showed high levels of activity in the rotator cuff and deltoid muscles. Supine Phase I exercises should be considered in the early postoperative period after shoulder surgery to achieve maximum motion while minimizing shoulder muscle activity. Progression to Phase II and Phase III exercises may proceed as soft tissue and bony healing permit. Phase III exercises performed with an elastic band should provide a satisfactory method to strengthen these muscles.