In order to optimize non-destructive mechanical testing of trabecular bone specimens, different techniques were analysed, and correlations were established between properties derived from such non-destructive testings and those derived from destructive testing. Non-destructive testing to a fixed percentage of predicted ultimate stress was hampered by inaccuracy of this prediction. Simulation of non-destructive testing conducted to the 'linear' part of the compression curve using a drop of the increase of the stiffness (slope of the compression curve) below a certain value as stop criterion revealed strong correlations ( r : 0.97-0.99) between the stiffness at the stop point and modulus of elasticity derived from destructive testing. However, trabecular damage will probably occur during such testing because high strain values were obtained. Testing to a fixed strain (0.6 per cent) also revealed strong correlation between the stiffness at the 0.6 per cent strain level and modulus of elasticity (r = 0.96) derived from destructive testing. By this technique trabecular damage was avoided and standardized elastic and viscoelastic energies could be obtained. Prediction of modulus of elasticity by indirect methods such as modulus of elasticity of specimens from the opposite symmetric location (r = 0.73), bone mineral concentration (r = 0.72) derived from photon absorptiometry and the CT number (r = 0.78) derived from quantitative computed tomography showed less strong correlations