Xavier H. Pham¹, Jessica N. Goetz¹, Angelica S. Busciglio¹, Lillie C. del Real¹, Mark B. Neider¹; ¹University of Central Florida

During free-viewing of natural scenes, oculomotor deployment is highly biased toward the center. However, when accounting for central fixation bias, image features such as luminance, luminance contrast, and edge density are highly predictive of fixation location (e.g., Nuthmann & Einhäuser, 2015). Studies have suggested that the ability to detect and differentiate low-level features like luminance degrades with age (e.g., Monge & Madden, 2016), but few studies of age-related change control for central fixation bias. While controlling for central fixation bias, we investigated the influence of image features on oculomotor behavior between age groups. In the current study (N = 64; 32 younger and 32 older adults) participants completed a free-viewing task. Scenes were segmented into 8 x 6 regions, and four feature values (central fixation bias, edge density, luminance, luminance contrast) were calculated for each region. For analysis, we utilized a generalized linear mixed model containing ten fixed effects (one intercept; four main effects of image features (central fixation bias, luminance, luminance contrast, edge density); five interactions (age group x intercept, age group x each image feature)) and two random effects (participant and image). We only analyzed the first three fixations from stimulus onset. The outcome variable was whether a region was fixated. Across age groups, there were no significant differences in the predictive power of luminance (b = 0.01), luminance contrast (b < -0.001), and edge density (b = 0.01) (all ps ≥ .20); visual features were similarly diagnostic of fixation location between older and younger adults. Interestingly, central fixation bias was significantly less predictive for older adults (b = -0.25, p < .001). Our findings suggest that older adults process scenes more globally, and despite evidence of age-related degradation of visual feature discrimination, low-level features continue to predict oculomotor behavior.