The use of shoulder manipulation in the treatment of frozen shoulder syndrome remains controversial. Opponents cite the risk of dislocation, fracture, nerve palsy, and rotator cuff tearing as limiting the usefulness of manipulation. A retrospective study of 38 shoulder manipulations in 32 patients was performed. These patients were followed for an average time of 58 months. The patients were examined in follow up for combined shoulder range of motion, external and internal rotation strength, and status of the long head of the biceps. Manipulation was performed in all patients by the senior author and supervised physical therapy was begun within 24 hours of the manipulation. The average recovery time was 13 weeks. In this series, 97% of patients had relief of pain and recovery of near complete range of motion, although 8% required a second manipulation to obtain a successful result. Mild weakness to manual muscle testing was present in 5.3% of patients in external rotation and 10.5% of patients in internal rotation. There was no deterioration of shoulder function with time. In fact, most patients improved with passage of time, even more. There was no evidence of biceps tendon rupture or rotator cuff insufficiency at the time of follow up in any of the patients. No fractures, dislocations or nerve palsies were observed, although one patient who had no premanipulation arthrogram was found to have a rotator cuff tear a few months after failed manipulation. Manipulation of the shoulder can therefore be offered to reduce the pain and period of disability in patients who fail conservative treatment of frozen shoulder syndrome.