Arthroscopy and examination under anaesthesia were performed for 328 consecutive knee injuries with haemarthrosis. These examinations were grouped according to a modified classification (Jackson and Abe, 1972) into very useful (117/328, 36 per cent), useful (98/328, 30 per cent) and useless (113/328, 34 per cent) categories. The probability of arthroscopy being useful was estimated mathematically. The factors which made this procedure useful were knee pain on exertion before the injury (P = 0.0561), the mechanism of the injury (P less than 0.0001) and the clinical stability of the patella (P = 0.0007). On arrival in the emergency department it was first decided whether the leg should be mobilized, immobilized in a plaster cast, operated on or, if a definitive diagnosis could not be reached, arthroscopy was deemed necessary. This resulted in the treatment following arthroscopy, and examination under anaesthesia, being altered from conservative to operative (P less than 0.0001). Results suggest that arthroscopy and examination under anaesthesia should always be considered to help in the diagnosis of acute injury haemarthrosis of the knee especially after a valgus or varus strain.