Variation in computed tomography (CT) image gray-scale and spatial geometry due to specimen orientation, magnification, voxel size, differences in X-ray photon energy and limited field-of-view during the scan, were evaluated in repeated micro-CT scans of iliac crest biopsies and test phantoms. Using the micro-CT scanner on beamline X2B at the Brookhaven National Laboratory's National Synchrotron Light Source, 3-D micro-CT images were generated. They consisted of up to 1024 x 2400(2), 4-microm cubic voxels, each with 16-bit gray-scale. We also reconstructed the images at 16-, 32-, and 48-microm voxel resolution. Scan data were reconstructed from the complete profiles using filtered back-projection and from truncated profiles using profile-extension and with a Local reconstruction algorithm. Three biopsies and one bone-like test phantom were each rescanned at three different times at annual intervals. For the full-data-set reconstructions, the reproducibility of the estimates of mineral content of bone at mean bone opacity value, was +/-28.8, i.e., 2.56%, in a 4-microm cubic voxel at the 95% confidence level. The reproducibility decreased with increased voxel size. The interscan difference in imaged bone volume ranged from 0.86 4-microm 0.64% at 4-microm voxel resolution, and 2.64 4-microm 2.48% at 48 microm.