Abstract
Walking without vision to previously viewed targets is a task commonly used to measure the perception of absolute distance. Previous work indicates that subjects are accurate at this task in full cue real world environments to targets up to 20 meters (Loomis et al.,1992). We reexamined claims of accuracy while looking for evidence of individual differences and changes in performance over time by analyzing data pooled from previous studies in our laboratory. This data came from over 100 subjects and involved 1,200 direct blind walking trials to targets at distances up to 12 meters. We found that, on average, subjects walked 96% of the distance to the target. One sided t-tests (p < .05) indicate that approximately one third of the subjects walked significantly less than the global average and another third of the subjects walked significantly more than the global average. These two groups walked 83% and 108% of the way to the target respectively on average. These individual differences indicate that blind walking experimenters should not necessarily compensate for a small number of subjects by running more trials per subject. An analysis of the percent walked by individual subjects over trials supports the suggestion by Philbeck et al. (2004) that subjects become more accurate at blind walking as they complete more trials even when no feedback is provided. Similar analyses would be useful to perform on distance judgments within HMD-based virtual environments, in which systematic biases of underestimation of distance are typically found.
This work was supported by NSF grant IIS-0121084.