Fifteen patients with shoulder instability and nine asymptomatic volunteers were studied with magnetic resonance (MR) imaging. The shoulder joint was visualized by means of arthroscopy or surgery in all patients. Ten patients had abnormalities of the glenoid labrum. Two musculoskeletal radiologists interpreted the MR images of the patients and volunteers without knowledge of the clinical history or surgical results. The surgical and arthroscopic results were used as the standard of reference in symptomatic patients. Observer A achieved a sensitivity of 44.4% and a specificity of 66.7%; observer B had a sensitivity of 77.8% and a specificity of 66.7%. In addition to the poor sensitivities and specificities, there was substantial intra- and interobserver variability. Assuming that the shoulders of the asymptomatic volunteers were normal, the specificities were 100.0% and 88.9% for observers A and B respectively. In this small study, axial MR imaging was relatively insensitive and nonspecific in the evaluation of labral lesions. Further study will be necessary to determine the utility and limits of MR imaging in this regard.