Eye movements reveal the role of attention in change blindness for shadows

In a change blindness paradigm, participants are presented with two images (alternating or side-by-side) and are asked to detect a change created with digital image manipulation. A recent study has suggested that changes in cast shadows are more difficult to detect than changes in objects (Ehinger, Allen, & Wolfe, 2016). Research has also suggested that when judging faces, observers show reduced attention to cast shadows (Hermens & Zdravković, 2015). A plausible reason for change blindness for shadows is therefore that observers fail to attend the relevant areas in an image. We here test this hypothesis by measuring eye movements while observers freely view the images from Ehinger et al. (2016). To compare change blindness in the same set-up we also asked participants to perform a change detection task where we presented images side-by-side and balanced the number of object and shadow changes. During free viewing, observers paid little attention to the relevant shadows and objects, but significantly more often fixated the relevant shadows than objects. During the change blindness task, changes in shadows were more easily detected than changes in objects. Attention during free viewing of individual images, however, did not predict change detection performance. The results suggest that during free viewing observers do not automatically discard cast shadows, and that change blindness may not be directly related to visual attention during free viewing.