OBJECTIVE: To study the effect of accuracy on the clinical outcome of local steroid injections to the shoulder. METHODS: 37 patients with shoulder symptoms of at least two months' duration received local injections of a mixture of triamcinolone and radiographic contrast material using a standardised technique. Radiographs of the joint were taken immediately afterwards. Details of the patients' symptoms (assessed by visual analogue scales) and range of movement at the joint were obtained before and two weeks after the injection. At follow up the patients were also assessed by means of a five point global rating scale of maximum and current benefit. RESULTS: 14 of the 38 procedures (37%) were judged to be accurately placed: four of the 14 attempted subacromial injections (29%) and 10 of the 24 attempted glenohumeral injections (42%). There were significant differences in relation to outcome between the accurately placed and the inaccurately placed groups. CONCLUSIONS: Accuracy of steroid placement by injection in patients with shoulder symptoms may significantly affect the clinical outcome.