During the Boeing Joint Strike Fighter Concept Development Phase, we investigated the feasibility of using three-dimensional (3D) solid models, implemented within a Virtual Environment (VE), as a low-cost partial replacement for conventional hardware mockup trainers for aircraft maintainers. Currently, there are few studies directly comparing performance using VE-based training to more conventional methods. This paper summarizes the results of several empirical studies conducted to evaluate the effectiveness of aircraft maintenance training within VEs. In these studies, trainees were taught a simple remove-and-replace maintenance procedure. The task, while not complex, required a number of ordered steps involving visual and physical obstructions. Training effectiveness was assessed with a written test of task procedures and with an objective assessment of task performance on a hardware mockup. Measures of performance included task completion time and procedural errors (e.g., incorrect action, wrong tool). The initial study compared hardware mockup training to two alternative display formats: solid model-based VEs and 3D line drawings implemented as computer-based displays (CBD). Within each of these display formats, we compared passive "hands-off" training with user-interactive training. Results of the study indicated that as realism in the virtual training environment increased, performance approached that achieved with the more costly, time-intensive hardware mockup training. Another study assessed immersive Virtual Reality (VR) for task training. Participants in this condition trained for the maintenance task in an immersive VR, wearing a head-mounted display and interacting with task components using a 3D mouse. Results indicated that training time for the immersive condition was longer than the other CBD training methods, with a diminished task performance. Finally, using participants from the initial training study, we addressed the effectiveness of using VEs for maintenance rehearsal three months after completing initial training. Rehearsal involved a review of the task using one of two CBD methods. The first rehearsal condition was a review with annotated technical drawings; the second was the solid model-based interactive VE. A third group, the control, had no rehearsal. Performance for all participants was evaluated as before - with a written test of task procedures and performance on the hardware mockup. Results of this study showed a trend for better performance after interactive VE rehearsal over that of the other two conditions. Collectively, these studies indicate that solid model-based VEs provide a potentially significant alternative to hardware mockup based training, resulting in savings in training time and cost. Further research is needed to identify the types of training scenarios for which VEs are most effective..