The diagnostic value of arthroscopy was evaluated in 148 patients with acute hemarthrosis and/or instability of the knee. The treatment planned after clinical examination was compared with the treatment given after arthroscopy. Seventy-nine per cent of the patients had ligamentous injuries; 59 per cent of tears were combined with other injuries, and 71 per cent were complete ruptures. Stability testing under anesthesia was most inaccurate for the anterior cruciate ligament, with 13 per cent false positive and 30 per cent false negative results. The planned treatment was altered as a consequence of arthroscopy in 31 per cent of cases. Without arthroscopy, the preoperative diagnoses would have been seriously wrong in 15 per cent of the patients. Twenty per cent of total anterior cruciate ligament ruptures would have been overlooked.