Although macroscopic geometric architecture is an important determinant of bone strength, there is limited published information relating to the validation of the techniques used in its measurement. This study describes new techniques for assessing geometry at the midfemur using magnetic resonance imaging (MRI) and dual-energy X-ray absorptiometry (DXA) and examines both the repeatability and the accuracy of these and previously described DXA methods. Contiguous transverse MRI (Philips 1.5T) scans of the middle one-third femur were made in 13 subjects, 3 subjects with osteoporosis. Midpoint values for total width (TW), cortical width (CW), total cross-sectional area (TCSA), cortical cross-sectional area (CCSA), and volumes from reconstructed three-dimensional (3D) images (total volume [TV] and cortical volume [CVol]) were derived. Midpoint TW and CW also were determined using DXA (Lunar V3.6, lumbar software) by visual and automated edge detection analysis. Repeatability was assessed on scans made on two occasions and then analyzed twice by two independent observers (blinded), with intra- and interobserver repeatability expressed as the CV (CV +/- SD). Accuracy was examined by comparing MRI and DXA measurements of venison bone (and Perspex phantom for MRI), against "gold standard" measures made by vernier caliper (width), photographic image digitization (area) and water displacement (volume). Agreement between methods was analyzed using mean differences (MD +/- SD%). MRI CVs ranged from 0.5 +/- 0.5% (TV) to 3.1 +/- 3.1% (CW) for intraobserver and 0.55 +/- 0.5% (TV) to 3.6 +/- 3.6% (CW) for interobserver repeatability. DXA results ranged from 1.6 +/- 1.5% (TW) to 4.4 +/- 4.5% (CW) for intraobserver and 3.8 +/- 3.8% (TW) to 8.3 +/- 8.1% (CW) for interobserver variation. MRI accuracy was excellent for TV (3.3 +/- 6.4%), CVol (3.5 +/- 4.0%), TCSA (1.8 +/- 2.6%), and CCSA (1.6 +/- 4.2%) but not TW (4.1 +/- 1.4%) or CW (16.4 +/14.9%). DXA results were TW (6.8 +/- 2.7%) and CW (16.4 +/- 17.0%). MRI measures of geometric parameters of the midfemur are highly accurate and repeatable, even in osteoporosis. Both MRI and DXA techniques have limited value in determining cortical width. MRI may prove valuable in the assessment of surface-specific bone accrual and resorption responses to disease, therapy, and variations in mechanical loading.