Differential Effects of Peripheral and Central Vision Loss on Scene Perception and Eye Movement Patterns

Talk Presentation 14.15: Friday, May 16, 2025, 3:30 – 4:45 pm, Talk Room 1
Session: Eye Movements: Natural tasks, neural mechanisms

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

Peripheral (PVL) and central vision loss (CVL) are irreversible visual impairments that significantly affect visual tasks like search and reading. Studies have shown that patients with vision loss exhibit reduced visual search accuracies and reading speeds. While previous studies have documented reduced performance in these domains, less is known about how PVL and CVL impact the perception of natural scenes and social cues. To investigate, we tested 32 sighted observers using a gaze-contingent simulation for PVL, CVL, or no impairment. Participants viewed 120 natural scenes (half depicting social interactions and half from the MS COCO dataset) and generated descriptions after one or three saccades. PVL was simulated with a 10-degree clear window surrounded by a Gaussian blur, while CVL applied a 10-degree Gaussian blur to the center of fixation. Eye movement data was analyzed by determining the correlation between fixation heat maps for each viewing condition and scene. Description quality for each scene was rated for semantic similarity to gold-standard descriptions. Results revealed a significant three-way interaction between viewing condition, scene type, and saccade count (F=3.978, p=.018). Interestingly, when viewing social interaction scenes with one saccade, descriptions generated with PVL were rated lower than CVL (p=.0001, Cohen’s D = .3045) and no impairment (p < .0001, Cohen’s D = .5206). Fixation heat map correlations between CVL and no impairment were lowest across scene types and saccades (F=52.907, p < .0001), suggesting greater changes in fixation patterns compared to no impairment (p < .0001, Cohen’s D = 1.29) and PVL (p < .0001, Cohen’s D = 1.02). These findings suggest distinct effects underlying scene perception in PVL and CVL: PVL reduces semantic understanding of scenes, while CVL alters gaze behavior. This work underscores the need for tailored interventions based on impairment type to improve daily functioning for individuals with vision loss.