Abstract
In real-world environments, stimuli move and change over time, yet most change detection studies use still-frame images. How is change detection modulated by the amount and direction of changes in spatiotemporal continuity when watching real-world videos, relative to the identical changes in still-frame images? In Experiment 1, observers were instructed to detect changes within continuous video clips. During critical saccades, videos shifted forward or backward in time by 0ms, 500ms, or 1000ms. Experiment 2 replicated the identical stimulus changes and removed motion percepts by exchanging the video stimuli with the still-frame images immediately before and after each change executed in Experiment 1. Detection accuracy with videos was surprisingly low (54%-68%), and significantly lower than with still-frame images (70%-82%, p<.001), suggesting a greater insensitivity to spatiotemporal continuity changes in videos. Larger temporal changes were detected more accurately (both experiments p<.001). Also, both experiments showed trends towards less accurate detection of forward relative to backward changes (significant when collapsed across experiments, p<.02), suggesting that predictive motion may mask some changes. Detection was also modulated by oculomotor properties; detection in videos was worse when observers were in stable fixation immediately after the change rather than in smooth pursuit (47%-76%, p<.001), despite all changes occurring during saccade rather than pursuit. Moreover, changes associated with stable fixation in videos were again detected less accurately in Experiment 2 (70%-85%, p<.001), where still-frame images were presented. This suggests a stimulus-based contribution to the stable fixation disadvantage in videos, as pursuit fixations could not occur with still images. However, the stable fixation disadvantage was larger in videos (p<.001), suggesting that the oculomotor act of pursuit itself may also improve detection. Collectively, the results suggest that forward and backward spatiotemporal continuity changes in videos are strikingly difficult to detect even when compared to the identical changes in still images.
Meeting abstract presented at VSS 2014