The purpose of this study was to investigate the effects of unilateral vestibular neurotomy on humans ability to perform navigation tasks. These tasks provided self-motion feedback by way of either locomotor activity only (nonvisual navigation or "locomotor task") or visual motion cues only (visually simulated navigation or "visual task"). After exploration of an environment in which 4 locations were marked by different objects, subjects attempted to navigate to those locations either by reproducing the same paths as those followed during exploration, by reversing routes, or by making spatial inferences (shortcuts). Vestibular defective patients were tested one day before surgical treatment and during the recovery time course following unilateral vestibular nerve lesion (1 week, 1 month, and 3 month later). Their performance was assessed by measuring turn error and distance error in both navigation tasks and was compared to that of control subjects tested 4 times at similar time intervals. Turn error in the reproduction of previously explored routes in the locomotor task was lower in patients before surgery than in controls, suggesting the existence of compensatory processes. In the acute stage (1 week) after unilateral vestibular lesion, turn error was greater in patients than in controls for the highest level of mental representation (spatial inferences or reversing routes); impairment at making accurate rotations had disappeared by 1 month after vestibular lesion in both navigation tasks. These results point to the role of vestibular cues, in interaction with other sensory modalities, in the elaboration of an accurate internal representation of the environment. In addition, they suggest that unilateral suppression of vestibular information would induce transitory spatial memory disorganization at a high level of information processing.