Several methods of varying complexity are available for the measurement of in vivo insulin secretion in man. No study has previously compared these in the same subjects to establish which is the most appropriate for routine use. We have, therefore, compared four methods for measuring insulin secretion in man: Hyperglycaemic clamp (Hy), Minimal model (MIN), shortened intravenous glucose tolerance test (IVGTT) and continuous infusion of glucose with model assessment (C.I.G.M.A.). Seventeen subjects with varying degrees of insulin sensitivity were studied. Seven normal (BMI 22.5 +/- 1.5 kg/m2), five obese (BMI 38 +/- 5 kg/m2) and five NIDDM subjects (BMI 27 +/- 3 kg/m2) were investigated, in a randomised fashion, on separate days. First (PSI) and second phase (PSII) rate constants (MIN); incremental insulin secretion 0-10 mins (Hy delta I) and steady state insulin levels from the last 30 minutes (Hy120-150) from the hyperglycaemic clamp; 3 minute insulin concentration and incremental area under insulin secretion curve 0-10 min (IVGTT) and beta-cell function (%) from C.I.G.M.A. were used as indicators of insulin secretion. Each index of insulin secretion could detect an overall difference between the groups. Insulin secretion in normals and obese was similar but significantly increased compared to NIDDM. In normals PSI correlated with C.I.G.M.A. (Rs = 0.92, p < 0.02) and Hy120-150 (Rs = 0.82, p < 0.05). IVGTT0-10 correlated with PSII (Rs = 0.83, p < 0.05), HY delta I (Rs = 0.84, p < 0.05) and IVGTT3 min (Rs = 1.0, p < 0.001). In obese PSII correlated with C.I.G.M.A. (Rs = 0.91, p < 0.05), Hy delta I (Rs = 1.0, p < 0.02) Hy120-150 (Rs = 0.92, p < 0.05) and IVGTT3 min Rs = 1.0, p < 0.02). In addition Hy delta I also correlated with C.I.G.M.A. (Rs = 0.92, p < 0.05) and IVGTT3 min (Rs = 1.0, p < 0.02). In NIDDM Hy delta I correlated with C.I.G.M.A. (Rs = 0.91, p < 0.005). When all subjects from the three groups were combined, significant positive correlations were obtained between each index of insulin secretion. In conclusion we have demonstrated that: (a) C.I.G.M.A., IVGTT, Minimal model and hyperglycaemic clamp can provide similar overall results for, in vivo, beta-cell function in man. (b) Significant positive correlations were obtained between each index of insulin secretion when all subjects were combined. (c) Using the above methodologies insulin secretion in normal and obese appears similar but significantly increased compared to NIDDM subjects.(ABSTRACT TRUNCATED AT 400 WORDS)