Clinical features and specific aspects of treatment were evaluated in 612 patients with gram-negative bacteremia observed over a 10 year period. Coagulation abnormalities or thrombocytopenia were observed in 64 per cent of the patients. Evidence of disseminated intravascular coagulation (DIC) was found in approximately 10 per cent of them but was of sufficient severity to be associated with subcutaneous or visceral bleeding in 3 per cent of them. The frequency of coagulation abnormalities, other than DIC, was greater in patients with more severe underlying disease but DIC occurred with similar frequency irrespective of the severity of underyling host disease. Coagulation abnormalities of all types were associated with increased fatality rates. Hypothermia was noted in 13 per cent of the patients at the onset of bacteremia but was transient and was not associated with increased fatality. Failure to mount a febrile response greater than 99.6 degrees F within the first 24 hours of bacteremia was associated with a significant increase in fatality rates. Prior corticosteroid therapy diminished the febrile response to bacteremia. Age, underlying host disease, granulocytopenia, congestive heart failure, diabetes mellitus, renal insufficiency, nosocomial infections, and antecedent treatment with antibiotics, corticosteroids, and antimetabolites significantly increased fatality rates. Appropriate antibiotic treatment reduced the fatality rate of those with bacteremia by approximately one-half among patients in each category of severity of underlying host disease. In addition, it was shown that early appropriate antibiotic therapy also reduced the frequency with which shock developed by one half. Even after development of shock, appropriate antibiotic therapy significantly reduced fatality rates. The use of combinations of antibiotics could not be demonstrated to significantly improve survival rates. Minimal differences in therapeutic efficacy could be demonstrated between individual antibiotics and various combinations of antimicrobials. Shock occurred in approximately 40 per cent of the patients and its frequency was not influenced by the species of etiologic agent. Contrary to previous reports, corticosteroid therapy in patients with shock did not enhance survival and treatment with an average of 4.0 g/day of hydrocortisone or its equivalents was associated with a significant increase in fatality rates.