The relationship of cortical bone mineral density (BMD), and geometry to bone strength has been well documented. In this study, we used peripheral quantitative computerized tomography (pQCT) to acquire trabecular BMD and high-resolution images of trabeculae from specimens to determine their relationship with biomechanical properties. Fifty-eight human cubic trabecular bone specimens, including 26 from the vertebral bodies, were scanned in water and air. Trabecular structure was quantitated using software developed with Advanced Visual Systems interfaced on a Sun/Sparc Workstation. BMD was also obtained using a whole-body computerized tomography scanner (QCT). Nondestructive testing of the specimens was performed to assess their elastic modulus. QCT and pQCT measurements of BMD of specimens in water were strongly correlated (r2 = 0.95, p < 0.0001), with a slope (0.96) statistically not significantly different from 1. Strong correlations were found between pQCT measurements of specimens in water and in air, for BMD (r2 = 0.96, p < 0.0001), and for apparent trabecular structural parameters (r2 = 0.89-0.93, p < 0.0001). Correlations were moderate between BMD and apparent trabecular structural parameters (r2 = 0.37-0.64, p < 0.0001). Precision as coefficient of variation (CV) and standardized coefficient of variation (SCV) for these measurements was < 5%. For the vertebral specimens, the correlation was higher between elastic modulus and BMD (r2 = 0.76,p < 0.0001) than between elastic modulus and apparent trabecular structural parameters (r2 = 0.58-0.72, p < 0.0001), while the addition of apparent trabecular nodes and branches to BMD in a multivariate regression model significantly increased the correlation with the elastic modulus (r2 = 0.86, p < 0.01). Thus, pQCT can comparably and reproducibly measure trabecular bone mineral in water or air, and trabecular structure can be quantitated from pQCT images. The combination of volumetric BMD with trabecular structural parameters rather than either alone improves the prediction of biomechanical properties. Such a noninvasive approach may be useful for the preclinical study of osteoporosis.