We evaluated the validity of the traditional method of assessment of the speed of insulin action during a euglycemic-hyperinsulinemic clamp. We first estimated the error of Steele's model on glucose uptake in these experimental conditions. We tested the appropriateness of estimating the half-time of insulin action by expressing the glucose flux changes as a percent of the maximal change (normalization on a 0% to 100% scale). For this purpose, we performed a 390-minute euglycemic-hyperinsulinemic (2 mU.min-1.kg-1) clamp in five chronically catheterized conscious dogs. We used [3-3H]glucose to assess glucose kinetics. We used a novel analysis based on a circulatory model, which allowed us to overcome the limitations of compartmental analysis. We found that the primary effect of insulin (increased from 12.3 +/- 1.6 to 104 +/- 15 microU/mL) was to increase the whole-body fractional extraction of glucose (3.0% +/- 0.3% to 18% +/- 2%). Insulin did not alter the mean whole-body artery-vein transit time (3.1 +/- 0.2 v 2.9 +/- 0.4 minutes). In contrast to the assumptions of the Steele model, which assumes that glucose uptake and rate of appearance (Ra) are equal during the clamp, during the initial 30 minutes of the clamp the increase in glucose uptake preceded (by approximately 4 minutes) the increase in Ra. Thus, during this period uptake exceeded Ra by about 15%. The maximal difference between Ra and uptake (1 to 1.5 mg.min-1.kg-1) occurred approximately 15 minutes after insulin infusion. Finally, to estimate the half-time of the insulin signal that controls glucose uptake and production, we accounted for the nonlinear relationship between insulin concentration and glucose uptake and production. We found that the traditional normalization of the glucose flux changes on a 0% to 100% scale underestimated the half-time of onset of the insulin signal that controls glucose uptake (half-time, 20 v 54 minutes) and glucose production (half-time, 25 v 39 minutes). Accounting for the nonlinearity of the dose-response curves may thus be of crucial importance in the evaluation of the onset and offset of insulin action.