Roux's principle of bone functional adaptation postulates that bone tissue, and particularly trabecular bone tissue, responds to mechanical stimuli by adjusting (modeling) its architecture accordingly. Hence, it predicts that the new modeled trabecular structure is mechanically improved (stiffer and stronger) in line with the habitual in vivo loading direction. While previous studies found indirect evidence to support this theory, direct support was so far unattainable. This is attributed to the fact that each trabecular bone is unique, and that trabecular bone tissue tends to be damaged during mechanical testing. Consequently, a unique modeled trabecular structure can be mechanically tested only along one direction and a comparison to other directions for that specific structure is impossible. To address this issue, we have 3D printed 10 replicas of a trabecular structure from a sheep talus cropped along the 3 principal axes of the bone and in line with the principal direction of loading (denoted on-axis model). Next, we have rotated the same cropped trabecular structure in increments of 10° up to 90° to the bone principal direction of loading (denoted off-axis models) and printed 10 replicas of each off-axis model. Finally, all on-axis and off-axis 3D printed replicas were loaded in compression until failure and trabecular structure stiffness and strength were calculated. Contrary to our prediction, and conflicting with Roux's principle of bone functional adaptation, we found that a trabecular structure loaded off-axis tended to have higher stiffness and strength values when compared to the same trabecular structure loaded on-axis. These unexpected results may not disprove Roux's principle of bone functional adaptation, but they do imply that trabecular bone adaptation may serve additional purposes than simply optimizing bone structure to one principal loading scenario and this suggests that we still don't fully understand bone modeling in its entirety.