Abstract
Both path integration and visual landmarks could be used to guide homing during spatial navigation. While the interaction of these cues during animal navigation is extensively investigated, how humans use multiple sources of information is poorly understood. We tested two hypotheses about how humans combine path integration and visual landmarks during a homing task in an virtual environment: (1) The Bayesian integration hypothesis
predicts that behavior will be determined by a weighted combination of cues, with the weights based on their individual reliability (Cheng et al., 2007). (2) The multiplesystem hypothesis predicts that different cues will dominate under different environmental conditions (Shettleworth & Sutton, 2005). Participants performed a
triangle completion task in the presence of three visual landmarks, which were increasingly shifted prior to the home-bound leg (by 15 to 135). With local landmarks
(5.5 m from home) participants followed the landmarks completely up to a 90 shift, and then switched to rely on path integration at 135: a bimodal response was often observed around 90°. With distal landmarks (500m from home), by contrast, most participants primarily relied on path integration and ignored the landmarks. Although some
individuals appeared to integrate the information in the distal condition, they also switched to path integration beyond 90. These findings are consistent with previous
animal navigation studies, indicating that path integration not optimally combined with visual landmarks during human navigation. Instead, it may serve as a back-up or
reference system when external visual landmarks become noticeably unreliable or are absent.
