Four rapid glucose injections of 5 g each were administered to normal young adult subjects before, during, and after an infusion of glucose. After the first glucose pulse, insulin responses measured immunologically reached a peak between 3 and 5 min and rapidly returned to base line. A short glucose infusion of 300 mg/min decreased the rapid insulin response to a second glucose pulse (- 58%), but after a longer infusion (20 hr) the acute insulin response to a third pulse was restored to normal. Stopping the infusion was followed by return of glucose and insulin levels to prestudy base line within 1 hr, but a fourth glucose pulse was followed by a supernormal acute insulin response (+ 200%). Other observations during these studies showed that a short glucose infusion of either 100 mg/min or 300 mg/min produced a parallel rise in glucose and insulin, but continuation of the infusion for 20 hr was associated with a “paradoxical” fall in glucose and continued rise in insulin. These observations are considered incompatible with a simple linear model often used to describe the relation between plasma glucose and serum insulin. Instead, a two pool system—one for acute insulin release, and the other a time-dependent compartment for long term insulin responses—is suggested.