The glucose metabolic rate of the human brain can be measured with labeled deoxyglucose, using positron emission tomography, provided certain conditions are fulfilled. The original method assumed irreversible trapping of deoxyglucose metabolites in brain during the experimental period, and it further requires that a conversion factor between deoxyglucose and glucose, the "lumped constant," be known for the brain regions of interest. We examined the assumption of irreversible trapping of fluorodeoxyglucose metabolites in brain of four patients in 365 normal and 4 recently infarcted regions. The average net, steady-state rate of fluorodeoxyglucose (KD) accumulation in normal regions of the four patients was 0.025 ml g-1 min-1. We also examined the variability of the lumped constant. We first confirmed that methylglucose is not phosphorylated in the human brain. We then estimated the lumped constant from the regional distribution of labeled methylglucose in brain. The average (virtual) volume of distribution of labeled methylglucose in the normal regions was 0.46 ml g-1 and was the same in both gray and white matter structures. The average brain glucose content corresponding to this value was 1.3 mumol g-1, assuming a Michaelis constant (Kt) of 3.7 mM for glucose transport across the blood-brain barrier. The lumped constant varied insignificantly between 0.4 and 0.5 in most regions, with an overall average of 0.44. It did not vary significantly between the patients and was the same in gray and white matter structures, but was inversely related to the calculated metabolic rate. This observation indicates that metabolic rates calculated with a fixed lumped constant (e.g., 0.40) would be slightly underestimated at high metabolic rates and slightly overestimated at low metabolic rates. The average glucose metabolic rates of the 365 normal regions, in which gray matter regions prevailed by 20:1, was 32 mumol 100 g-1 min-1. The average glucose phosphorylation rate in white matter was 20 mumol 100 g-1 min-1 with a lumped constant of 0.45. In the recently infarcted areas, the lumped constants varied from 0.37 to 2.83, corresponding to glucose metabolic rates varying from 2 to 18 mumol 100 g-1 min-1. Two infarct types were identified. In one type, the phosphorylation-limited type, glucose content and the lumped constant were close to normal (1 mumol g-1 and 0.40, respectively). In the other, the transport/flow-limited type, the glucose content was low (0.2 mumol g-1), and the lumped constant in excess of unity. The evidence from the present study upholds the model of Sokoloff et al. in every detail.