The lumbar spine of 14 cadavers was studied both by 153Gd dual photon absorptiometry (DPA) and quantitative computed tomography (QCT) at 96 and 125 kVp. The intact spine and the individual vertebrae were analyzed. After these measurements the ash content of the vertebral body, the posterior elements, and the transverse processes was determined. The fat content of the vertebral body as well as its volume was also measured. With DPA, the bone mineral content (BMC) determined in situ as well as on excised spine specimens correlated highly with the amount of total vertebral ash (r>0.92, SEE<3.2 g). The bone mineral density (BMD, area density) of 3 lumbar vertebrae correlated accurately with the mean ash density of the vertebral body (r>0.81, SEE<0.015 g/cm3). The so-called corpus density and central density determinations were less accurate. No difference in accuracy was found between measurements when using 3 mm and 4.5 mm step intervals. Variations in the distribution of mineral between the vertebral body and the posterior elements contribute to the error in predicting vertebral body mineral with DPA. QCT gave a smaller error when a cylindric portion of the vertebral body with a 20 mm diameter was measured compared with one with a 9 mm diameter, when the dual energy technique was used (p<0.01). With dual energy QCT a correlation was found between a center segment of 3 vertebrae in the lumbar spine and the mean ash density of the vertebral body of r=0.92 (SEE=0.010 g/cm3). Single energy QCT was insignificantly less accurate than dual energy QCT. Only small differences were found between vertebrae with high fat density of the vertebral body when single or dual QCT was used. QCT was more accurate than DPA in the prediction of the mineral density of individual vertebral bodies (p<0.05) but no difference was found when the average values for the lumbar spine were calculated.