Abstract
Shape similarity judgments involve more than immediate perception. Prior experience and attentional biases may also be important. We investigate these issues using 2D contour shape triplets generated by morphing two novel "parent" shapes into a third "daughter" shape that is, by this measure, equally similar to both parent shapes. Participants were presented with shape triplets and judged which parent shape was more similar to the daughter shape. Each of 100 triplets was presented twice per session (e.g., the triplet presented on trial 1 was repeated on trial 101), on each of 4 sessions (days 1, 2, 8, & 31). Consistency of responding (the probability of choosing the same parent shape on two different presentations of the same triplet) was measured. Despite objectively equivalent similarity, average consistency was above chance and approximately equal within and between sessions (0.70 vs. 0.71). Since between-session intervals were much longer (1-30 days) than within-session intervals (<15 minutes), equivalent performance suggests participants weren’t basing future responses on explicit memory of earlier choices. Next, for each participant, we selected consistent-response triplets from each session, and measured response consistency for these triplets in all other sessions. Average consistency was higher (0.78) and independent of the session from which the shapes were selected and the session on which consistency was measured. This result agrees with a model in which individual participants respond consistently to some shapes and not others. Finally, doing the same analysis using consistent-response triplets from different participants (e.g., measuring consistency in subject A using consistent-response triplets from subject B) had little affect on consistency, indicating that different participants responded consistently to different subsets of the triplets. Together, these results suggest that individualized, stable attentional biases drive judgments of similarity under conditions of uncertainty.
Meeting abstract presented at VSS 2012