The value of selective surgery for abdominal stab wounds was assessed in a prospective study of 226 patients admitted over an 18-month period. Two died soon after admission and before operation, the indications for which were signs of peritoneal irritation or haemorrhage, evisceration of bowel and omental protrusion. Of the 226 patients, 113 were selected for immediate surgery, which was carried out within 3 hours of admission. There were 12 deaths (10-6 per cent). Of the 111 patients initially selected for observation, 47 (42-3 per cent) underwent delayed operation with 4 deaths (8-5 per cent), but in only 1 (2-1 per cent) of these patients could the delay have influenced the adverse result. There were no deaths in the 64 patients treated conservatively. The overall mortality rate of patients who were initially selected for observation was 3-9 per cent. One hundred and fifty patients underwent laparotomy, positive findings justifying operation being present in 134 (89 per cent). Six patients required thoractomy for concomitant thoracic injuries, including 2 patients who required cardiorraphy. Four other patients required combined thoracic and abdominal exploration. The selection of patients for either conservative or operative management was based entirely on clinical criteris; abdominal X-rays were not helpful in the early diagnosis of visceral lesions. The indications for operation were evident within 12 hours of admission in 156 of the 160 patients who were treated surgically. A substantial reduction in the incidence of 'negative laparotomy' can be made by this process of selection, and without an associated increase in the morbidity and mortality rates.