We have minimized methodological errors in the isotope dilution technique by using stable isotope, [6,6-2H2]glucose, thus avoiding the problem of contamination of tritiated glucose tracers and, by maintaining a constant plasma tracer enrichment have reduced error due to mixing transients. Using these modifications we have calculated hepatic glucose production in 20 patients with non-insulin-dependent diabetes mellitus during low (1 mU kg-1 min-1) and high (8 mU kg-1 min-1) dose insulin infusions. Mean fasting hepatic glucose production was 14.2 +/- 0.8 mumol kg-1 min-1. This suppressed by only 68% to 4.6 +/- 0.8 mumol kg-1 min-1 during the low-dose insulin infusion (plasma insulin 0.85 +/- 0.05 nmol l-1) and did not suppress further during the high-dose insulin infusion (plasma insulin 14.55 +/- 0.83 nmol l-1). Hepatic glucose production was significantly higher than zero throughout the study. Thus, we have found that minimization of known errors in the isotope dilution technique results in physiologically plausible and significantly positive values for hepatic glucose production indicating that the liver is resistant to insulin in patients with non-insulin-dependent diabetes mellitus.