Spatial survey estimation is incremental and relies on directed memory structures

This study examined how navigators of large-scale environmental spaces come up with survey estimates of distant targets. Participants learned a route through a virtual city by walking it multiple times in one direction on an omnidirectional treadmill. After learning, they were teleported to intersections along the route and pointed to multiple other locations. Locations were always queried in chunks of related trials relative to a participant’s current position, either to all locations route forwards or all locations route backwards. For their first pointing, participants took twice as long as for the later pointings and latency correlated with the number of intersections to the target, which was not the case for later pointings. These findings are inconsistent with reading out coordinates from a cognitive map but fit well with constructive theories which suggest that participants integrated locations between their current location and the target along the learned path. Later pointings to adjacent intersections within a chunk of trials continued this process using the previous estimation. Additionally, in first pointings participants’ estimates were quicker and more accurate when targets were located route forwards than route backwards. This route direction effect shows that the long-term memory employed in generating survey estimates must be directed – either in form of a directed graph or a combination of a directed route layer and an undirected survey layer.