Abstract
We have found that humans depend heavily on ordinal spatial knowledge and landmarks, rather than metric spatial knowledge, when navigating in a virtual hedge maze (Harrison et al, VSS, 2001, 2002; Zhong et al, VSS 2005, Psychonomics, 2005). A related form of topological structure is provided by boundaries that carve up space into adjacent neighborhoods. Kuipers et al (2003) found that people tend to choose paths that participate in more boundary relations when navigating in grid-like desktop virtual environments. Here we investigate whether knowledge of boundary relations improves the accuracy of shortcuts when walking in a virtual maze. Participants actively walk in the VENLab, an immersive virtual environment (10m x10m) with a head-mounted display (60 deg H x 40 deg V) and a sonic/inertial tracking system (50–70ms latency). In the learning phase, participants freely explore a hedge maze with primary, secondary, and tertiary paths (the boundaries) and are trained to walk from a Home location to each of 10 places successfully. In the testing phase, they walk to place A, the maze is removed, and they are instructed take a shortcut to place B. On control trials, all features of the maze are removed. On probe trials, either (a) primary, (b) primary and secondary, or (c) all three paths remain visible. If participants utilize knowledge of boundary relations, their shortcut accuracy and precision should improve as more paths remain. The results allow us to determine the role of topological boundary relations in active navigation.
Acknowledgements: Funded by NSF BCS-0214383