What if they knocked down the library and nobody noticed? The failure to detect large changes to familiar scenes

The purpose of the current study was to investigate people's ability to detect changes to familiar scenes. College students were asked either to identify what was wrong with a picture of a familiar location on their college campus (e.g., the library had been removed from the scene), or to estimate the difficulty of change detection for a hypothetical cohort performing the same task. Performance in the change-detection condition was extremely poor, even when changes were large. Participants who were familiar with the scenes and those who were unfamiliar with the scenes both overestimated the actual levels of change-detection performance. A follow-up analysis indicated that the participants who were unfamiliar with the scenes produced estimations of difficulty that were highly correlated with the mathematical area of the change, whereas participants who were familiar with the scenes produced estimations of difficulty that were highly correlated with the actual difficulty of change detection. The results indicate that people's visual long-term memory for familiar scenes lacks the precision to be able to effectively identify even large-scale changes, although subjectively people believe this should be relatively easy.