Since the implementation of the Tower of London (ToL) test by Shallice in 1982, numerous variants differing in the tower's physical appearance have been developed. Here we compare behavioral performance (n = 31) on the original Tower of London task consisting of three rods of unequal lengths with a three-ball version of the Ward and Allport Tower Task (1997) using three equally sized rods. In the problem set used, the start and goal states for both tower configurations were identical across all trials. The experiment was divided into two parts: in the first block, the problems presented were equalized with respect to the number of paths for achieving an optimal solution, the minimum number of moves, goal hierarchy, subgoaling patterns, and suboptimal alternatives between the two tower versions. As expected, participants showed the same performance scores for both types of towers when structural problem parameters were identical. In the second block, participants had to solve five-move problems which-due to the different rod sizes of the towers-had only one optimal solution in the original version, but two optimal solutions in the variant with three rods of equal length. Participants revealed lower performance scores and showed longer planning times in the original version than in the second tower version. These findings demonstrate that the two tower versions are only interchangeable when specific planning parameters are equalized. Otherwise, even if problems look identical, significant differences in performance may be found due to the differing problem spaces in the two tower versions.