Background. Studies in animals have suggested that intravenous vasopressin is associated with better vital-organ perfusion and resuscitation rates than is epinephrine in the treatment of cardiac arrest. We did a randomised comparison of vasopressin with epinephrine in patients with ventricular fibrillation in out-of-hospital cardiac arrest. Methods. 40 patients in ventricular fibrillation resistant to electrical defibrillation were prospectively and randomly assigned epinephrine (1 mg intravenously; n = 20) or vasopressin (40 U intravenously; n = 20) as primary drug therapy for cardiac arrest. The endpoints of this double-blind study were successful resuscitation (hospital admission), survival for 24 h, survival to hospital discharge, acid neurological outcome (Glasgow coma scale). Analyses were by intention to treat. Findings. Seven (35%) patients in the epinephrine group and 14 (70%) in the vasopressin group survived to hospital admission (p = 0.06). At 24 h, four (20%) epinephrine-treated patients and 12 (60%) vasopressin-treated patients were alive (p = 0.02). Three (15%) patients in the epinephrine group and eight (40%) in the vasopressin group survived to hospital discharge (p = 0.16). Neurological outcomes were similar (mean Glasgow coma score at hospital discharge 10.7 [SE 3.8] vs 11.7 [1.6], p = 0.78). Interpretation. In this preliminary study, a significantly larger proportion of patients treated with vasopressin than of those treated with epinephrine were resuscitated successfully from out-of-hospital ventricular fibrillation and survived for 24 h. Based upon these findings, larger multicentre studies of vasopressin in the treatment of cardiac arrest are needed.