PURPOSE: To assess the accuracy of shoulder infiltrations in the subacromial bursa (SAB) by a posterior or an anteromedial approach. Magnetic resonance imaging (MRI) and clinical outcome were used for evaluation. TYPE OF STUDY: A prospective randomized study. METHODS: Thirty-three patients (22 women, 11 men; average age, 46 years; range, 25 to 64 years) with clinical signs of subacromial impingement were infiltrated with a mixture of bupivacaine, methylprednisolone, and gadolinium-DTPA directly followed by MRI to determine the actual site of injection. The SAB was randomly infiltrated posteriorly (n = 17) or anteromedially (n = 16). Injection confidence of the surgeon and body-mass index of the patient were recorded. Follow-up consisted of the Constant Score, Simple Shoulder Test, and visual analog scale score for pain taken within 24 hours and 6 weeks after infiltration. RESULTS: Thirteen injections (76%) were in the SAB with a posterior approach and 10 (69%) with an anteromedial approach. Many surrounding structures were hit as well, especially the rotator cuff. A positive correlation between the injection confidence of the orthopaedic surgeon and the MRI was found in 66%. Only injection of the SAB alone resulted in a significant decrease of the pain (P = .004) and an increase in the functional scores. Injection in the bursa and rotator cuff muscle showed a significant increase in pain (P = .032) but no change in clinical scores. The body mass index had no influence on the scores. CONCLUSIONS: Injections in the SAB are inaccurate, despite the confident feeling of the clinician. The finding that many different structures can be infiltrated with 1 injection can create both false-positive and -negative results. LEVEL OF EVIDENCE: Level II.