Abstract
Humans have two somewhat independent navigation systems based on rough path integration and precise visual landmarks. When taking a short cut between well-learned locations, participants follow surreptitiously shifted landmarks completely despite their lack of reliability (Foo et al., submitted). However, large shifts (18 deg) are detected and in such cases participants fall back on coarse spatial knowledge built up from path integration. Such large shifts may be detectable within the resolution of the path integration system (SD ∼20 deg, Kearns, et al. 2002). Here we test the shift magnitude required for explicit detection during continuous path integration in a triangle completion task. Participants walked in an immersive virtual environment (12 ∼ 12 m) while wearing a head-mounted-display (60û × 40û, 50 ms latency). On each trial, they walked two specified legs of a triangle and then returned to the starting position (four different triangles). To assess the accuracy of path integration, no landmarks were visible in the first and last blocks of trials. In intermediate blocks, a landmark (red post) was positioned near the starting position, and was surreptitiously displaced during the trial by 0, 12, 20 or 28 deg (viewed from the third vertex). All participants faithfully followed the shifted landmark in all conditions, and none noticed the shift. In a second experiment, participants were told explicitly that the post was often unreliable. In this case they ignored the landmark and relied on path integration across all conditions. These results suggest that dependence on visual landmarks, even ones that move by as much as 28 deg, supersedes the use of path integration. Thus, although both systems provide information useful for navigation, our evidence suggests that they interact in a “winner-take-all” manner in which landmarks typically dominate navigation.
Funded by NSF IRI-9720327, NSF BCS 0214383 & NSF IGERT 9870676