To retrospectively evaluate the accuracy of magnetic resonance (MR) arthrography in the classification of anteroinferior labroligamentous injuries by using arthroscopy as the reference standard. Ethical committee approval and informed consent were obtained. MR arthrograms obtained in 205 patients, including a study group of 104 patients (74 male and 30 female; mean age, 28.2 years) with arthroscopically proved labroligamentous injuries and a control group of 101 patients (65 male and 36 female; mean age, 31.4 years) with intact labroligamentous complex, were reviewed in random order. MR arthrograms were analyzed for the presence and type (Bankart, anterior labral periosteal sleeve avulsion [ALPSA], Perthes, glenolabral articular disruption [GLAD], or nonclassifiable lesion) of labroligamentous injuries by two radiologists in consensus. Results were compared with arthroscopic findings. Sensitivity, specificity, accuracy, and corresponding 95% confidence intervals for the detection and classification of anteroinferior labroligamentous lesions with MR arthrography were calculated. At arthroscopy, 104 anteroinferior labroligamentous lesions were diagnosed, including 44 Bankart lesions, 22 ALPSA lesions, 12 Perthes lesions, and three GLAD lesions. Twenty-three labral lesions were nonclassifiable at arthroscopy, all of which occurred after a history of chronic instability. Nineteen (83%) of these 23 lesions were also nonclassifiable at MR arthrography. With arthroscopy used as the reference standard, labroligamentous lesions were detected and correctly classified at MR arthrography with sensitivities of 88% and 77%, specificities of 91% and 91%, and accuracies of 89% and 84%, respectively. Bankart, ALPSA, and Perthes lesions were correctly classified in 80%, 77%, and 50% of cases, respectively. The three GLAD lesions were all correctly assessed. MR arthrography is accurate in enabling classification of acute and chronic anteroinferior labroligamentous injuries, although correct interpretation of Perthes lesions remains difficult.