Air Force pilots and control subjects were tested on a visual “mental rotation” task. Nine of the 16 pilots, as well as all of the 16 control subjects, required more time to rotate greater angular distances. The performance of the other 7 pilots was unique: their response time did not increase with greater angular rotations. The results suggest that visual mental rotation can be accomplished by at least two different processes. One process involves incremental object rotations in a multi-step mapping –like an actual physical rotation of an object– going through intermediate stages. This process requires more time to rotate greater angular distances. The other process involves direct translation in a single-step mapping. In this process, the starting position transforms into the final position in one mapping without any intermediate steps, and thus does not require more time to rotate greater angular rotation. The lack of intermediate stages, which may allow small perturbations in location to be corrected, affects the accuracy of this process; this is particularly apparent when more complex stimuli are rotated. The pilots who did not show incremental rotation effects had different and distinct error patterns, their errors increased when rotating the more complex stimuli.