The purpose of this study was to assess the sensitivity of magnetic resonance imaging in determining the presence of articular cartilage injuries of the knee with arthroscopy as the standard for comparison. Forty-nine articular cartilage lesions were documented in 28 knees (27 patients) by arthroscopy. There were 22 men and 5 women with an average age of 29 years. Multiplanar magnetic resonance imaging was performed with spin echo and gradient-refocused acquisition in a steady state pulse technique. All of the knees had magnetic resonance imaging done within 4 weeks prior to arthroscopy. The magnetic resonance images were interpreted before arthroscopy and interpreted again after the results of arthroscopy were known to better define the potential learning curve for evaluating chondral lesions and to identify the technical limits of the existing imaging protocol/software. For full-thickness articular cartilage lesions, the prearthroscopy sensitivity of magnetic resonance imaging was 41% (12/29) and the postarthroscopy sensitivity was 83% (24/29). For partial-thickness chondral injury, the prearthroscopy sensitivity of magnetic resonance imaging was 15% (3/20) and the postarthroscopy sensitivity was 55% (11/20). The presence of an intraarticular effusion assisted the detection of chondral lesions because of an "arthrogram" effect. As a noninvasive method of evaluating articular cartilage and despite experienced interpretation and the benefit of retrospective analysis, both the prearthroscopy and the postarthroscopy sensitivity of magnetic resonance imaging was low using the imaging parameters described. Injury to articular cartilage is a frequent cause of knee pain and knee surgery; it is important to note at this time that magnetic resonance imaging cannot reliably exclude the presence of an articular cartilage injury.