Three dimensional visualizations are becoming one of the popular web visualization techniques; however, studies show that for some users, comprehending 3D visualizations is not easy. This research investigated what makes 3D visualizations difficult to understand with the purpose of using this knowledge to develop training to improve visualization literacy. In particular, we focused on the difficulties that individuals have with orientation problems in which they are asked to re-orient themselves to a visual scene from a different viewpoint, something that happens quite often in 3D information visualizations. For our research, we used complex 3D cubes as our experiment problems and sliced the cubes in different ways such that the slices matched the visual orientation a person would naturally take when viewing the slice or required the person to take a different orientation to best determine the features of the slice being made. We then compared performance on recognizing slices that matched a viewer's orientation to that of slices that required a re-orientation. Our results indicated that the problems requiring a reorientation were significantly less accurate.