Underestimation of glucose turnover rates has been a problem in clamp studies using primed-constant [3-3H]-glucose infusion technique. Due to slow mixing in interstitial compartments concealed specific activity gradients may arise between plasma and interstitial compartments during intravenous unlabelled glucose infusion. Such specific activity gradients, however, can be prevented if plasma specific activity is maintained constant. Two euglycaemic clamp studies (insulin infusion 40 mU m-2 min-1) were performed in six lean normal subjects. Using conventional unlabelled glucose infusates plasma specific activity declined by 74%, tracer determined glucose appearance was smaller than actual glucose infusion rates (317 +/- 11 vs 366 +/- 15 mg m-2 min-1, p < 0.001), and erroneous negative values were calculated for glucose production (- 49 +/- 7 mg m-2 min-1). Average underestimation during the first 2 h correlated with glucose infusion rates (r = 0.88, p < 0.02). In contrast, when plasma specific activity was maintained constant, using appropriately labelled glucose infusates, tracer determined glucose appearance and glucose infusion rates were similar (385 +/- 16 vs 385 +/- 17 mg m-2 min-1), and negative errors for glucose production were avoided. In conclusion, using unlabelled glucose infusates, as in previous studies, suppression of glucose production is overestimated and stimulation of glucose utilization is underestimated. As errors were greater with larger glucose infusions, the mistakes may have been greatest in insulin sensitive control subjects, and smaller in insulin resistant subjects. Therefore, re-evaluation of hepatic insulin sensitivity seems appropriate in diabetes, obesity, and other insulin resistant states.