Clinical decision-making could be difficult when Magnetic resonance imaging (MRI) is used for the diagnosis of knee injuries. We retrospectively studied 565 knee arthroscopies done between 2002 and 2005, 110 of which had suspected ligamentous injuries, evaluated clinically, with MRI and subsequently by arthroscopy. The aim of the study was to know the extent of correlation of clinical, MRI features with arthroscopy and whether MRI could be justifiably used to deny an arthroscopy. All patients with a strongly suggestive history were examined in the clinic by experienced orthopaedic surgeons and MRI was requested. Clinical examination was repeated under anaesthesia by the operating surgeon who not necessarily had examined the patient initially. The clinical and arthroscopy findings were recorded systematically. 3 Radiology consultants of varying musculoskeletal experience reported the MRI films. The clinical and MRI findings were compared with arthroscopy for the extent of correlation. We observed that overall Sensitivity of MRI for meniscal injuries was 73%, being more for medial than lateral and 86% for cruciate ligament injuries. Clinical examination had a sensitivity of 33% and a specificity of 93% for meniscal injuries, sensitivity of 86% and specificity of 100% for cruciate injuries. MRI was not able to demonstrate synovial plicae in 13 knees and chondral defects in 3 knees. 96 Knees, which were normal clinically, were found to have meniscal tears on MRI in 65 and subsequently confirmed by arthroscopy in 39. We conclude that an accurately performed clinical examination with positive signs alone, will be justified for arthroscopy and a negative MRI will not be a sufficient evidence to deny an arthroscopy. Also the reporting will largely depend on the quality of information provided by the clinician, technical factors and the musculoskeletal experience of the person reporting the films.