Abstract
When passively exposed to optic flow, implied travel distance is underestimated. Conversely, participants stop early when deciding how much optic flow specifies a given distance, as if travel distance were overestimated. Lappe et al. (2006) argued this was due to a leaky integrator that loses magnitude while trying to accumulate traveled distance from a starting point, but loses magnitude from its representation of the distance remaining to the target. An alternative theory suggests that the decision to stop is stochastically biased. We tested for leakiness in non-visual path integration, by describing a task either as reproducing a distance or returning to a starting point. Blindfolded subjects walked 5–40 m along a guide wire until told to stop. Step counting was prevented with an alphanumeric memory task. In the Return condition, subjects tried to walk back to their starting point. In the Reproduce condition, subjects switched to a second guide wire and tried to reproduce the distance walked. Consistent with leaky integration, the slopes of distances produced in the Return condition were lower than in the Reproduce condition. For the longest distance subjects stopped short in both conditions, but reliably shorter in the Return condition (27.4 m) than in the Reproduce condition (31.3 m), t(12) = 2.733, p = .0182. Consistent with the results of Sun et al. (2004), performance at 10–20 m was relatively unbiased. The success of the instructional manipulation is consistent with the idea of a leaky integrator for non-visual path integration. However, the data also indicate memory contamination from previous trials: Compared to an initial 7m practice trial (6.9 m), subjects traveled reliably farther for the 7m target later on (9.0 m), t(5) = 2.72, p = .0434, whereas the average travel distance for the 28m target, initially 26.6 m, decreased to 22.5 m.