Recent studies indicate that hydrogen-labeled glucose tracers underestimate glucose turnover in humans under conditions of high flux. The cause of this underestimation is unknown. To determine whether the error is time-, pool-, model-, or insulin-dependent, glucose turnover was measured simultaneously with [6-3H]-, [6,6-2H2]-, and [6-14C]glucose during a 7-h infusion of either insulin (1 mU.kg-1.min-1) or saline. During the insulin infusion, steady-state glucose turnover measured with both [6-3H]glucose (8.0 +/- 0.5 mg.kg-1.min-1) and [6,6-2H2]glucose (7.6 +/- 0.5 mg.kg-1.min-1) was lower (P less than .01) than either the glucose infusion rate required to maintain euglycemia (9.8 +/- 0.7 mg.kg-1.min-1) or glucose turnover determined with [6-14C]glucose and corrected for Cori cycle activity (9.8 +/- 0.7 mg.kg-1.min-1). Consequently "negative" glucose production rates (P less than .01) were obtained with either [6-3H]- or [6,6-2H2]- but not [6-14C]glucose. The difference between turnover estimated with [6-3H]glucose and actual glucose disposal (or 14C glucose flux) did not decrease with time and was not dependent on duration of isotope infusion. During saline infusion, estimates of glucose turnover were similar regardless of the glucose tracer used. High-performance liquid chromatography of the radioactive glucose tracer and plasma revealed the presence of a tritiated nonglucose contaminant. Although the contaminant represented only 1.5% of the radioactivity in the [6-3H]glucose infusate, its clearance was 10-fold less (P less than .001) than that of [6-3H]glucose. This resulted in accumulation in plasma, with the contaminant accounting for 16.6 +/- 2.09 and 10.8 +/- 0.9% of what customarily is assumed to be plasma glucose radioactivity during the insulin or saline infusion, respectively (P less than .01). When corrected for the presence of the contaminant, glucose turnover determined with [6-3H]glucose during insulin infusion (9.5 +/- 0.6 mg.kg-1.min-1) no longer differed from either the glucose infusion rate or that determined with [6-14C]glucose. Therefore, the underestimation of glucose turnover during insulin infusion and negative glucose production rates observed with traditional methods to analyze plasma radioactivity and commercially available tracers is the result of an artifactual increase in [6-3H]glucose specific activity. The etiology of the underestimation of glucose turnover with [6,6-2H2]glucose remains to be determined.