Abstract
What geometric properties of the environment are preserved in spatial knowledge for navigation? In the first of a series of studies of this question, we investigate whether human navigation on learned routes relies on metric distances and angles or ordinal relations among junctions and paths. The experiments were conducted during active walking in a 40 × 40 ft virtual environment of a garden or hedge maze, generated on an SGI workstation. Participants wore a head-mounted display (60 deg H × 40 deg V) and head position was measured by a hybrid sonic/inertial tracking system (50 ms latency). Initially, participants learned the layout of ten places in the garden (statue, well, bird bath, fountain, etc.) by free exploration. They were then tested on their ability to walk to specified places from a “home” location while their routes and times were recorded. On random catch trials, the virtual garden was distorted in two ways. To test metric distance, the garden was stretched along its main axis by 160 or 137 percent, such that walking a learned distance would result in turning at the wrong junction. Similarly, to test metric angle, a “radial arm” section of the garden was rotated by 45 deg, such that turning a learned angle would result in taking the wrong path. Results suggest that route navigation relies more on ordinal relations among junctions and paths than on metric distances and angles.
NSF LIS IRI-9720327.