OBJECTIVE : To determine how often patients with surgically proven labral tears have labral signal on shoulder MR arthrography (MRA) that is not equal to gadolinium or fluid on T1- and T2-weighted images, respectively. METHODS : Consecutive patients with surgical repair of a SLAP or Bankart labral tear within 95 days of an MRA were included. Using cartilage signal as an internal reference, two musculoskeletal (MSK) radiologists retrospectively categorized labral signal as T1-hyperintense, T1-gadolinium, T2-hyperintense, or T2-fluid. In patients without T1-gadolinium or T2-fluid labral signal, secondary findings such as the orientation, extent, shape, and width of the abnormal signal was recorded. Statistical analyses were performed using Fisher's test and ANOVA. RESULTS : Sixty-one labral tears (36 SLAP and 25 Bankart) in 54 patients (mean age, 30.7; F:M 8:46) met the inclusion criteria. In 67% and 76% of SLAP and Bankart labral tears, T1-gadolinium signal was present (p = 0.43). T2-fluid signal was present in 50% and 92% of these same labral tears (p = 0.001). The absence of T1-gadolinium or T2-fluid signal was more common in SLAP tears (33%) compared to Bankart tears (8%) (p = 0.02). In the SLAP cases, at least two secondary findings of a SLAP tear were present in 92% (11/12). CONCLUSIONS : Lack of surfacing T1-gadolinium or T2-fluid labral signal is unusual in Bankart tears but relatively common in SLAP tears. However, a SLAP tear was diagnosed in 92% of these 12 cases when two secondary findings were present.