We compared efficacy of magnetic resonance arthrography (MRA) and magnetic resonance imaging (MRI) for the detection of superior glenoid labral detachment, and especially unstable superior labrum anterior-posterior (SLAP) type 2 lesions. MR imagings and MR arthrograms of 90 shoulders of 90 patients, including 30 shoulders of unstable SLAP 2 lesions, were interpreted by a novice and an expert observer. These results were correlated with arthroscopic findings. For the expert observer, MRI had a sensitivity of 53%, a specificity of 93%, and an accuracy of 80%, which were improved to 60%, 98%, and 85%, respectively, by MRA. These increases were statistically insignificant. For the novice observer, MRI had a sensitivity of 45%, a specificity of 80%, and an accuracy of 64%, which were significantly improved to 76%, 83%, and 80%, respectively, by MRA. In detecting unstable SLAP 2 lesions, MRA proved to be useful to a novice interpreter, but not significantly useful to an experienced interpreter, compared with unenhanced MRI. We advocate MRA as the examination of first choice for detecting unstable SLAP 2 lesions if they are strongly suspected clinically.