Quantitative CT (QCT) studies of trabecular vertebral bone tissue have been carried out in vitro on a GE CT/T 9800 scanner. Results of both single energy (SE) 80 kVp and dual energy (DE) 80/140 kVp QCT data are compared with chemical mineral analysis to determine accuracy. We examined 62 vertebral specimens, from 28 cadavers (19 male and 9 female with an age range of 19-93 years, mean = 60.4). Averaging the results of all vertebral bodies of the same individual for SEQCT versus ashweight, we found a correlation coefficient (r) of 0.94 (p less than 0.0001), a standard error of the estimate (SEE) of 12.2 mg/cm3 (calibrated to K2HPO4), with a coefficient of variation (CV) of 13.2% and an average underestimation of bone mineral content of 18.7 mg/cm3. The corresponding DEQCT results were r = 0.98 (p less than 0.0001), SEE = 7.4 mg/cm3, CV = 7.0%, and an average underestimation of 4.9 mg/cm3. The SE and DE results are correlated with r = 0.98 (p less than 0.0001), SEE = 8.0 mg/cm3, and CV = 8.7%. From our SEQCT data and the results of the chemical analysis of bone mineral and fat content we calculated a fat sensitivity of 7.7 mg/cm3 K2HPO4 per 100 mg/cm3 fat change for our scanner. Using an average fat variability of 87.5 mg/cm3, this leads to a fat-related uncertainty for the normative SEQCT data of 6.7 mg/cm3, which is far lower than the normal biological variation of 29.4 mg/cm3. Using tabulated normative data on fat content versus age and versus mineral content of 188 vertebral specimens from five collaborating centers, we derived a correction algorithm for QCT measurement that reduces our average underestimation to 0.88 mg/cm3 with an SEE of 12.1 mg/cm3. Hence, this correction procedure can be used to estimate the fat corrected absolute mineral density for research purposes or for scanners with high fat sensitivity. For the GE CT/T 9800 scanner, with a relatively low fat to mineral sensitivity at 80 kVp, the correction procedure is generally not recommended for clinical studies since it minimizes the average fat induced error but does not reduce the residual, partially fat related uncertainty. Finally, since the fat related uncertainty is small compared to biological variation, the correlation is high between SEQCT and DEQCT, and the radiation dose is lower and the precision higher for SEQCT, we suggest that most clinical diagnostic studies using the GE CT/T 9800 scanner for bone mineral determination employ SEQCT at 80 kVp.