Abstract
Mammalian navigation depends both on visual landmarks and on self-generated (e.g. vestibular and proprioceptive) cues that signal the organism's movement from one moment to the next. Previous studies have found that under conflict, landmarks can reset estimates of self-motion, or be integrated with them. We asked whether humans combine these cues optimally, given their relative reliability, and whether children, who use both kinds of information from a young age, combine them as adults do. In a dark arena with illuminated peripheral landmarks, participants attempted to return an object to its original place, given (i) non-visual self-motion information only, (ii) visual landmarks only, or (iii) both. In the “self-motion only” condition, participants responded in the dark. In the “landmarks only” condition, participants were disoriented by turning before responding. In the “both” condition landmarks were available and participants were not disoriented. In a further “conflict” condition (iv), landmarks were covertly rotated by 15°, creating a conflict between self-motion and landmark information that enabled us to assess how the two were weighted relative to each other. When self-motion and landmark information conflicted, adults and children aged 4 – 8 years weighted them optimally given their relative reliability. However, only adults were Bayes-optimal in reducing variability in their location estimates when both information sources were available and consistent, whereas children's variability was not lower than with either information source alone. These results indicate that in human navigation, development of individual representational systems greatly precedes development of the capacity to combine multiple sources of spatial information optimally within a common reference frame. Further, optimal weighting of conflicting spatial cues precedes ability to combine them to reduce variance.