Abstract
Foveal vision loss can impact basic perceptual tasks such as face recognition, visual search, and scene classification. Day-to-day life also requires rapidly understanding scenes with social interactions, but the extent to which vision loss affects social interactions is not well understood. Here, we investigate the impact of foveal vision loss on scene understanding of real-world images using a simulated gaze-contingent central scotoma (Gaussian; SD = 4 deg). Eighteen observers viewed and described 120 scenes (50% with social interactions, 50% images from MS COCO). Half of the observers viewed the scenes with the simulated scotoma. Trials were terminated after one or three scene fixations (randomly counter-balanced). Scene descriptions were rated (8 evaluators) on their semantic similarity relative to gold standard descriptions (with unlimited viewing time). Results showed that the scotoma led to scene descriptions with significantly lower similarity to the gold standard (F = 5.99, p = .027). Participants with a simulated scotoma were less correlated and more variable in first fixation location after stimulus onset compared to participants without a scotoma (t = 5.0569, p < 0.0001), suggesting that the scotoma’s presence affected fixation strategies. Manual annotations of relevant objects (humans, animals, and faces) revealed that participants showed similar preference for humans and faces regardless of viewing condition (t = 0.436, p = 0.7). Our findings provide a method to assess the impact of vision loss on scene understanding of real-world scenes. These findings have important implications for understanding the challenges faced by people with central vision loss.