Abstract
Previous research has shown that visual long-term memory (VLTM) has large capacity (Shepard, 1963; Standing et al., 1970, 1973) and can store rich details (Brady et al., 2008; Konkle et al., 2010) for both object images and scenes. Here we investigated the fidelity of VLTM by occluding images with “bubbled” masks that had various degrees of coverage. The bubbles were comprised of six circles with varying sizes and positions, and only the region inside these “bubbles” was visible. Participants viewed 228 scenes for 20 minutes, followed by a 15 minute Ospan task to measure working memory capacity, and then completed a series of forced-choice memory tests for the previously viewed images. Across all experiments, there was no correlation between performance on the VLTM task and Ospan score. Experiment 1 used bubbled scenes (coverage: 20%, 40%, 60%) for both study and test and subjects were asked to remember the specific scene/bubble combination. At test, subjects performed a four-alternative-force-choice task with the foils being old-Picture-new-Bubbles, new-Picture-old-Bubbles and new-Picture-new-Bubbles. Although subjects correctly choose the target on 41% of the trials (above chance, t(9) = 4.8, p < .01), they false alarmed to the foils on 26%, 17% and 17% of the trials, respectively. These data suggest that subjects are more sensitive to the scenes than the specific bubbling patterns. There was no effect of the degree of bubbling coverage (F < 1). We next asked whether bubbling would impair memory when the task was to remember only the scenes. Experiment 2 used non-bubbled scenes for study and bubbled scenes (coverage: 0%, 30%, 60%) for the test. Although there was a significant drop in accuracy for 60% coverage condition, performance was still above chance. Together these results suggest that memory for scenes are resilient but not impervious to occlusion.