Three methods by which to determine absolute total cerebral hemoglobin concentration (tHb in micromol/L) by near-infrared spectrophotometry (NIRS) have evolved: (1) tHbo, requiring oxygenation changes and arterial oxygen saturation measurements as a reference using a relative NIRS algorithm, (2) tHbg, using a geometrical multidistance principle and (3) tHbgo, a combination of both. The aim of this study was to compare the three methods quantitatively. Sixteen clinically stable preterm infants with a mean gestational age of 29.6 (range of 25.1-36.4) weeks, birthweight of 1386 (680-2820) g and a postnatal age of 2.5 (0.5-6) days, who needed supplemental oxygen, were enrolled. The mean+/-standard deviation tHbg was 150.2+/-41.8 micromol/L (range of 61.6-228.9 micromol/L), the tHbo was 62.1+/-27.2 micromol/L (26.0-110.8 micromol/L) and the tHbgo was 89.3+/-45.6 micromol/L (26.5-195.9 micromol/L). The correlation coefficient among the three methods were tHbg and tHbgo r=0.736; tHbo and tHbgo r=0.938; tHbg and tHbo r=0.598. A multiple regression with variable selection by Mellow's C(p) showed, that tHbg was correlated to the birthweight, the postnatal age, the heart rate and the pCO2 (r(2)=0.588), tHbo and tHbgo were associated with the hemoglobin concentration in the blood, the mean arterial blood pressure and the pCO2 (r(2)=0.493 and 0.406, respectively). The three methods (tHbg, tHbo, and tHbgo) give systematically different tHb readings and large intersubject variability.