The deuterated water method is used extensively to measure gluconeogenesis in humans. This method assumes negligible exchange of the lower three carbons of fructose 6-phsophate via transaldolase exchange since this exchange will result in enrichment of carbon 5 of glucose in the absence of net gluconeogenesis. The present studies tested this assumption. ²H₂O and acetaminophen were ingested and [1-¹³C]acetate infused in 11 nondiabetic subjects after a 16-h fast. Plasma and urinary glucuronide enrichments were measured using nuclear magnetic resonance spectroscopy before and during a 0.35 mU·kg FFM⁻¹·min⁻¹ insulin infusion. Rates of endogenous glucose production measured with [3-³H]- and [6,6-²H₂]glucose did not differ either before (14.0 ± 0.7 vs. 13.8 ± 0.7 μmol·kg⁻¹·min⁻¹) or during the clamp (10.4 ± 0.9 vs. 10.9 ± 0.7 μmol·kg⁻¹·min⁻¹), consistent with equilibration and quantitative removal of tritium during triose isomerase exchange. Plasma [3-¹³C] glucose-to-[4-¹³C]glucose and urinary [3-¹³C] glucuronide-to-[4-¹³C]glucuronide ratios were <1.0 (P < 0.001) in all subjects both before (0.66 ± 0.04 and 0.60 ± 0.04) and during (059 ± 0.05 and 0.56 ± 0.06) the insulin infusion, respectively, indicating that ∼35-45% of the labeling of the 5th carbon of glucose by deuterium was due to transaldolase exchange rather than gluconeogenesis. When corrected for transaldolase exchange, rates of gluconeogenesis were lower (P < 0.001) and glycogenolysis higher (P < 0.001) than uncorrected rates both before and during the insulin infusion. In conclusion, assuming negligible dilution by glycerol and near-complete triose isomerase equilibration, these data provide strong experimental evidence that transaldolase exchange occurs in humans, resulting in an overestimate of gluconeogenesis and an underestimate of glycogenolysis when measured with the ²H₂O method. Use of appropriate ¹³C tracers provides a means of correcting for transaldolase exchange.