Abstract
Limited attention capacities necessitate statistical summary representations such as statistical averaging. We examined how statistical averaging might be affected by feature type and variability in the set. Do we average length and orientation in a similar way? How similar should the items be in order to be averaged? Participants were presented shown sets of lines (n = 2, 3, 4, 9) for 133 ms. In Expts. 1-2, participants were asked to identify the length that represented the average or a member of the set. The lines were uniformly vertical in Expt. 1 and were randomly oriented (between -75 and +75 deg) in Expt. 2. In Expts. 3-4, participants were asked to identify the orientation that represented the average or a member of the set. The lines were all 2.7 deg in Expt. 3, and had randomly selected lengths (between 1.0 and 2.33 deg) in Expt. 4. Across all four experiments, we found that accuracy for identifying a set’s average feature, either length or orientation, was higher than the accuracy for identifying a member feature. Adding variability to the set in a feature dimension irrelevant to the task did not affect the pattern of accuracies. These data suggest that statistical averaging is robust even in sets with heterogeneous members.
Meeting abstract presented at VSS 2012