Byron A. Johnson¹ (), Shravan Murlidaran¹, Ansh Soni¹, Michael Beyeler¹, Miguel P. Eckstein¹; ¹University of California, Santa Barbara

Macular degeneration and retinitis pigmentosa are two leading causes of vision loss in which complementary parts of the retina are degraded or spared over time. Previous research has shown that vision loss impairs visual search, reading, and navigation, however, less is known about how these conditions affect scene perception. In order to understand how central and peripheral vision loss affects scene perception, fourteen observers were presented with 120 scenes in one of three conditions: central vision loss (CVL), peripheral vision loss (PVL), and no simulation. The scene images were categorized into two groups: 50% social interaction scenes and 50% MS COCO images. The CVL condition presented scenes with a gaze contingent central scotoma (Gaussian, SD = 5 degrees). The PVL condition presented scenes with a clear central aperture (5 degrees diameter) and Gaussian blurred periphery to simulate tunnel vision. The simulation tracking latency used a maximum update speed of 7 ms. Trials were terminated after one or three saccades. Scene description ratings were calculated with Sentence BERT Scores on participant description’s semantic similarity to gold standard descriptions generated with unlimited viewing time. Results with stop words removed show that both CVL and PVL conditions led to descriptions that were rated significantly lower than descriptions generated without impairment (ANOVA, F = 8.60, p = 0.006). The effect of viewing condition significantly interacted with scene type, suggesting that descriptions made with CVL and PVL were worse than the no impairment condition for scenes with social interactions (F = 7.38, p = .009). Although Tukey’s HSD revealed no difference in ratings between CVL and PVL, the CVL condition led to increased saccade amplitudes compared to PVL and no impairment (F = 9.05, p = .006). These findings show how CVL and PVL can influence eye movements for rapid scene understanding.