Abstract
We have reported that humans depend on topological spatial knowledge and landmarks rather than metric spatial knowledge when navigating in a virtual maze (Foo et al, JEP: LMC, 2005; Harrison et al, VSS, 2001, 2002; Zhong et al, VSS 2005, 2006, Psychonomics, 2005). Topological ordinal place structure consists of the sequence (and sign) of places passed on routes between locations in 2D. Metric place structure consists of the distances and angles between places in 2D. Zhong et al (Psychonomics, 2006) found that making previously learned places visible dramatically reduced variability in novel shortcuts to hidden targets, consistent with ordinal knowledge. Here we manipulate the configuration of visible places to test whether people rely on ordinal structure or the metric distance of the target from neighboring objects. Participants actively walk in a virtual hedge maze containing five pairs of places marked by distinctive objects (A–B). During learning, participants freely explore and are then trained to walk from Home to A and then B. During testing, they walk from Home to A, the hedges are removed (making all objects visible except the target), and they take a shortcut to B. On expansion trials, the metric distance of each object from the target is increased while ordinal structure remains constant. On rotation trials, each object is rotated about the target by a random angle, changing ordinal structure but leaving distance from the target constant. If participants rely on ordinal place structure, variable error (but not constant error) should increase in expansion trials, and should increase dramatically in rotation trials. If they rely on metric distance from visible objects, constant error (but not variable error) should increase in expansion trials, and both should remain constant in rotation trials. The results allow us to compare the contributions of ordinal and metric knowledge to active navigation.
Funded by NSF BCS-0214383