Abstract
Human observers are remarkably adept at extracting information from natural scenes. People can detect objects in scenes at very fast time courses (Thorpe et al., 1996), with limited visual attention (Li et al., 2002), and in the presence of occlusion (Meng & Potter, 2008). Here, we further investigated viewers' ability to identify the category of a scene under conditions of occlusion by “bubbling” target images; targets were sometimes displayed through five or more randomly positioned apertures, with radii chosen so that 20% – 80% of the images' original area was visible. Images had previously been rated as “good” (highly representative) or “bad” (least representative) exemplars of six natural categories: beaches, city streets, forests, highways, mountains, and offices. Participants were shown rapid sequences of three photographs and asked to indicate whether the category of the second (target) image was the same or different from that of the first and third images (which always belonged to the same category). We observed significant effects of both representativeness and occlusion on categorization performance, with higher accuracies for good category exemplars and lower levels of occlusion. However, even with only 20% of the original image visible through the bubbles, participants' performance remained above chance for both the good and bad exemplars. These results support previous reports on the robustness of scene perception to occlusion, but show that this manipulation does impair category recognition.
National Institutes of Health.