Forty-eight patients with frozen shoulder for less than six months were assigned at random to receive three shoulder injections into the subacromial bursa or glenohumeral joint at weekly intervals. The treatment groups were (1) intra-articular methylprednisolone and lidocaine, (2) intrabursal methylprednisolone and lidocaine, (3) intra-articular lidocaine, (4) intrabursal lidocaine. The same physical therapy program was carried out for all patients. Assessments of pain and range of motion were performed by a physical therapist who was uninformed about the nature of the injection therapy. There was no significant difference in outcome between intrabursal injection and intra-articular injection. Injection of steroid with lidocaine had no advantage over lidocaine alone in restoring shoulder motion, but partial, transient pain relief occurred in two thirds of the steroid-treated patients.