Vasopressin is often used to treat variceal hemorrhage. However, its efficacy is uncertain, and its portal hemodynamic effects in this setting are unknown. Eleven patients with alcoholic liver disease and bleeding varices were given vasopressin (0.2 U/min for the first hour, then 0.4 U/min for 24 hours). Portal pressure was monitored using an indwelling hepatic vein balloon catheter. Seventeen patients with variceal bleeding who remained stable over 26 hours of initial treatment with crystalloid and blood products served as a comparison group. Vasopressin infusion (0.2 U/min) produced a significant decrease in wedged hepatic venous pressure, hepatic venous pressure gradient (wedged minus free hepatic venous pressure), and heart rate. Increases in the rate of infusion did not produce further decreases in the parameters measured, but the changes were sustained over the course of the infusion. Hemodynamics remained stable in the control group. Portal pressure did not increase when vasopressin was abruptly discontinued in the 3 patients in whom postinfusion measurements were made. Vasopressin retains its portal hypotensive effects in the setting of variceal hemorrhage. Tachyphylaxis does not develop over 26 hours, and a "rebound" increase in portal pressure probably does not occur when the infusion is discontinued.