The authors examined the roll-over shape alignment hypothesis, which states that prosthetic feet are aligned by matching their roll-over shapes with an "ideal" shape. The "ideal" shape was considered to be the roll-over shape of the able-bodied foot-ankle system. An alignment algorithm and computational alignment system were developed to set trans-tibial alignments based on this hypothesis. Three prosthetic feet with considerably different roll-over shapes were either aligned using the alignment system or not aligned (i.e. used previous foot's alignment), and then were aligned by a team of prosthetists. No significant differences were found between roll-over shapes aligned by the computational alignment system and those based on standard clinical techniques (p = 0.944). Significant differences were found between the "no alignment" shapes and the prosthetist alignment shapes (p = 0.006), and between the "no alignment" shapes and the computational alignment system shapes (p = 0.024). The results of the experiment support the hypothesis that the goal of alignment is to match the prosthetic foot's roll-over shape, as closely as possible, with an "ideal" shape. The hypothesis is also supported by its ability to explain the results of previous studies. Using an "ideal" roll-over shape or surface as a goal for prosthetic alignment could lead to a priori alignment, eliminating the need for alignment hardware in some cases. Being able to build the alignment into a prosthesis without special hardware could be beneficial in low-income countries and in the fabrication of lightweight prostheses for the elderly.