There is at present general agreement, based upon abundant in vitro and in tvvo experimental evidence, that insulin exerts an immediate and profound effect on glucose utilization by muscle and adipose tissue (1-5). On the other hand, a direct hepatic action of insulin is disclaimed by most investigators since innumerable studies both in vitro and in vivo have failed to demonstrate clearly a consistent and reproducible effect of insulin upon hepatic glucose metabolism (5-12).Recent observations from this laboratory suggested that the site (13) and the rate (14) of insulin administration are important factors in eliciting a hepatic effect. However, in these as in all previous in vivo studies (11,(15)(16)(17)(18)(19), the conclusions regarding the hepatic action of insulin were inferential, since the net balance of glucose across the liver was not ascertained by direct measurement.Any experiments designed to elucidate the effect of insulin on hepatic carbohydrate metabolism should fulfill the following prerequisites:1. Hepatic, in contrast to splanchnic, glucose metabolism should be measured. In view of the marked sensitivity of the extrahepatic splanchnic tissues to insulin (1-5, 11, 12), separation of the effects of insulin on these tissues from its effects on the liver is mandatory. This, in turn, re-* This work was supported in part by a grant from the Upjohn Company and in part by the Dallas Heart Association. (23,24), and glucagon (25, 26), may obscure a hepatic effect of insulin by the marked increase in hepatic glucose release which characterizes the action of these counterregulatory hormones (27)(28)(29)(30).The present experiments were designed to meet these prerequisites in the following manner. First, dogs with complete end-to-side portacaval shunts were studied. This operation completely separates the liver from the remainder of the splanchnic tissues and thereby permits the measurement of hepatic rather than splanchnic glucose metabolism. Second, measures were taken to minimize or prevent the counter-regulatory mechanisms to hypoglycemia. In one group of experiments, insulin was administered by slow infusion which produced a very gradual modest decline in arterial glucose concentration. In another group, the hypoglycemic stimulus was further reduced in magnitude and duration by administering glucose after insulin infusion had been started. Finally hypoglycemia was prevented by using diabetic dogs with fasting hyperglycemia. 507