Abstract Gram-negative septicaemia is being recognized (is an increasingly common cause of mortality in hospital practice. In this paper the value of a simple and rapid test which can be used to detect the presence of circulating endotoxins is assessed in experimental and clinical situations. The test is based on the ability of Gram-negative endotoxins to coagulate a protein moiety obtained from the horseshoe crab (Limulus polyphemus). The sensitivity of a highly purified lysate when measured in vitro showed a positive Limulus test with as little as 510 μg/ml of endotoxin. In experimental animals we were able to detect endotoxin in the portal circulation only after the endotoxin was injected into the peritoneal cavity or when the bowel was made ischaemic or perforated, but never spontaneously. The clinical studies were performed in 50 surgical patients to determine if the Limulus test was useful in detecting endotoxaemia. Seventeen patients were diagnosed clinically as having local infection, and Gram-negative organisms were cultured from various sources in 16. Fifty-one (77per cent) out of 66 Limulus tests performed during the episode of infection were positive, while only 1 (3 per cent) of the tests was positive once the infection had been adequately treated. Four patients in this group had persistent endotoxaemia, and all died. In the 8 patients with generalized peritonitis, 16 (73 per cent) out of 17 Limulus tests performed were positive at the time of peritonitis. In 25 patients with no clinical evidence of infection or peritonitis, 19 (18 per cent) out of 125 Limulus tests were positive. The patients were also screened by the 125I-fibrinogen test for the presence of deep vein thrombosis, which was found in 13 (26 per cent). Thrombosis was detected in only 3 (19 per cent) of 16 patients in whom the Limulus test was negative throughout the investigation period. In contrast, 10 (29 per cent) of 34 patients who had positive Limulus tests developed thrombosis. In 6 the disease was bilateral, 5 had extensive thrombi which involved the major veins and 1 of these subsequently had a massive pulmonary embolism. The implications of these findings and the use of the Limulus test in clinical practice are discussed.