Graph properties drive navigational selection between equidistant routes

Cognitive maps, traditionally considered metrically accurate mental representations of space, have been central to navigation research. However, recent studies suggest human navigation often deviates from the predictions of cognitive maps. Instead, cognitive graphs - spatial representations based on landmarks (nodes) connected by routes (edges) with relative distances, angles and limited metric information - may more accurately describe mental spatial representation. Unlike cognitive maps, cognitive graphs emphasize structural relationships over precise details. We designed a two-alternative forced-choice navigational task where participants explored and navigated virtual environments with three ways to a target: left, middle, and right. Critically, the left and right routes were always identical in length but varied in structural features like the number of turns, length of the first path of the route, or the size of unpaved areas. After exploring, the middle route was blocked and participants chose the left or right route to navigate to the target. Across two experiments, participants completed the task using an immersive walking virtual reality interface or a desktop computer to view top-down images. Participants in both experiments preferred routes with fewer turns and larger inner and outer areas despite being metrically identical, but showed no preference for routes with a shorter initial path. These findings suggest that participants did not rely on metrically precise cognitive maps when deciding which route to take to a navigational goal. We interpret this as evidence for the use of topological or labeled graph representations and discuss heuristics that are compatible with or may drive reliance on cognitive graphs over cognitive maps. These findings build on prior evidence for cognitive graphs in physically impossible environments (e.g., wormholes) by showing a bias in the absence of route length differences.