Abstract: The purpose of this study was to compare computed tomography density (rho(CT)) obtained using typical clinical computed tomography scan parameters to ash density (rho(ash)), for the prediction of densities of femoral head trabecular bone from hip fracture patients. An experimental study was conducted to investigate the relationships between rho(ash) and rho(CT) and between each of these densities and rho(bulk) and rho(dry). Seven human femoral heads from hip fracture patients were computed tomography-scanned ex vivo, and 76 cylindrical trabecular bone specimens were collected. Computed tomography density was computed from computed tomography images by using a calibration Hounsfield units-based equation, whereas rho(bulk), rho(dry) and rho(ash) were determined experimentally. A large variation was found in the mean Hounsfield units of the bone cores (HUcore) with a constant bias from rho(CT) to rho(ash) of 42.5 mg/cm(3). Computed tomography and ash densities were linearly correlated (R-2 = 0.55, p < 0.001). It was demonstrated that rho(ash) provided a good estimate of rho(bulk) (R-2 = 0.78, p < 0.001) and is a strong predictor of rho(dry) (R-2 = 0.99, p < 0.001). In addition, the rho(CT) was linearly related to rho(bulk) (R-2 = 0.43, p < 0.001) and rho(dry) (R-2 = 0.56, p < 0.001). In conclusion, mineral density was an appropriate predictor of rho(bulk) and rho(dry), and rho(CT) was not a surrogate for rho(ash). There were linear relationships between rho(CT) and physical densities; however, following the experimental protocols of this study to determine rho(CT), considerable scatter was present in the rho(CT) relationships.