THE TRANSFER OF STRATEGIES FROM VIRTUAL TO REAL ENVIRONMENTS : AN EXPLANATION FOR PERFORMANCE DIFFERENCES ? Rudolph

Given recent attention to the use of virtual environments (VEs) for familiarization of unknown spaces, there has been no discussion to date concerning behavioral or qualitative analyses of how navigating in a VE differs from navigating in the real world and what effects these differences have on performance. We discuss such an analysis of several recent experiments we have completed. We break the task into route planning, rehearsal, and execution with associated maintenance and error recognition and recovery behaviors. In many instances, we can show that behaviors and strategies evident in the VE training phase are also evident in the real world transfer task. Consequently, while the data does not support the use of VEs as training aids for navigation, it does suggest that the VE can predict performance levels based on strategies employed. INTR ODUCTION There has been a significant amount of attention given recently to the use of virtual environments (VEs) for the purposes of navigation training and environmental familiar-ization. To a limited degree, we can now replicate real environments so that people can navigate through their virtual counterparts. It would seem obvious that there is something to be gained in terms of spatial knowledge acquired in the virtual and applied to the real environment. However, thus far results have been mixed on this issue. While we have seen very few studies reporting negative results (i.e. the VE always comes out ahead), we have experienced a number of cases in our experiments where exposure to a VE does not help in the real environment. In fact, we have seen cases where it actually lowers performance. These are admittedly counter-intuitive results. However, if we look beyond the quantitative data and delve into the specific behaviors that are at the heart of the issue, we begin to see why this might be the case. This paper presents results from a number of recent studies to show how an attempt to use real world strategies in a VE, along with specific shortcomings in the VEs themselves, gives rise to predictable behaviors which can begin to explain performance shortfalls. Our experiments are specifically designed to allow for flexibility in strategy. While this approach can tend to increase variation in the data, it is a close match to real navigation and therefore is the behavior we really want to study. Given that no VE can ever be a complete replica of a real …