We performed quantitative computed tomography in vitro on the first and third lumbar vertebrae in human cadavera using a dibasic potassium phosphate phantom for calibration. The quantitative computed-tomography numbers exhibited a significant positive correlation (R2 = 0.89, p less than 0.0001) with direct measurements of the apparent density of the vertebral trabecular bone. We also conducted uniaxial compression tests to failure of the vertebral bodies after removal of the posterior elements, and found that vertebral compressive strength was also correlated at a high level of significance (R2 = 0.82, p less than 0.0001) with direct measurement of the trabecular apparent density. These findings suggested the possibility that the quantitative computed-tomography values might be directly predictive of vertebral compressive strength. However, when we correlated the quantitative computed-tomography values directly with vertebral compressive strength, the results (R2 = 0.46, p less than 0.061) were suggestive but not quite significant. All vertebral bodies failed by compression of the end-plate, suggesting only a modest structural role for the cortical shell under these loading conditions. This was confirmed by comparing the compressive load to failure of twenty additional pairs of vertebrae that were tested with and without an intact vertebral cortex. Removal of the cortex was associated with approximately 10 per cent reduction in vertebral load to failure.