The design of the present study of the kinetics of insulin in man combines experimental features which obviate two of the major problems in previous insulin studies. (a) The use of radioiodinated insulin as a tracer has been shown to be inappropriate since its metabolism differs markedly from that of the native hormone. Therefore porcine insulin was administered by procedures which raised insulin levels in arterial plasma into the upper physiologic range. Hypoglycemia was prevented by adjusting the rate of an intravenous infusion of glucose in order to control the blood glucose concentration (the glucose-clamp technique). (b) Estimation of a single biological half-time of insulin after pulse injection of the hormone has been shown to be inappropriate since plasma insulin disappearance curves are multiexponential. Therefore the SAAM 25 computer program was used in order to define the parameters of a three compartment insulin model. The combined insulin mass of the three compartments (expressed as plasma equivalent volume) is equal to inulin space (15.7% body wt). Compartment 1 is apparently the plasma space (4.5%). The other two compartments are extra-vascular; compartment 2 is small (1.7%) and equilibrates rapidly with plasma, and compartment 3 is large (9.5%) and equilibrates slowly with plasma. The SAAM 25 program can simulate the buildup and decay of insulin in compartments 2 and 3 which cannot be assayed directly. Insulin in compartment 3 was found to correlate remarkably with the time-course of the servo-controlled glucose infusion. Under conditions of a steady-state arterial glucose level, glucose infusion is a measure of glucose utilization. We conclude that compartment 3 insulin (rather than plasma insulin) is a more direct determinant of glucose utilization. We suggest that the combined use of glucose-clamp and kinetic-modeling techniques should aid in the delineation of pathophysiologic states affecting glucose and insulin metabolism.