Saloni Sharma¹ (), Marcel Linka^2,3, Maximillian Davide Broda², Benjamin de Haas², Margaret Livingstone¹; ¹Harvard Medical School, ²Justus-Liebig-Universität Giessen, ³Stanford University

Human free-viewing behavior towards natural scenes can be predicted by semantic object information (Xu et al., 2014) with substantial variability across individuals (de Haas et al., 2019) and age (Linka et al., 2025). An open question is whether such gaze biases reflect conserved primate characteristics or if they are human-specific. Here, we analyzed free-viewing behavior of 103 humans (age = 25.17 ± 5.5; 72 females) and 8 monkeys (age = 9.13 ± 5.1; 1 female) across 200 natural scenes. Humans exhibited a pronounced center-bias with fixation density concentrated near the center of the scenes. Monkeys also showed a center-bias, but it was weaker and less spatially concentrated. Humans made more fixations per image than monkeys (humans: 9.25 ± 0.91; monkeys: 7.73 ± 1.88, p = 0.0011), though the mean fixation duration differed weakly across species (humans: 240.52ms ± 24.25; monkeys: 242.33ms ± 55.96; p = 0.044). Next, we quantified group differences in the dwell time and first fixations proportion across six semantic categories (Faces, Text, Emotion, Hands, Touched and Edible objects). Despite a history of face deprivation during earlier development in two of the monkeys, both species showed the strongest attentional bias towards Faces. However, this face bias was more pronounced in humans (dwell time: humans 26.48%, monkeys 21.21%; first fixations: humans 39.31%, monkeys 28.96%; both p < .001). Compared to humans, monkeys looked less at Text (dwell: 4.22% vs. 14.15%; first fixation: 4.38% vs. 7.23%, both p < .001) and spent slightly less dwell time on Touched (13.08% vs. 9.48%; p < .001) and Edible objects (10.32% vs. 9.06%; p < .05). Overall, these findings indicate that free-viewing behavior in humans and monkeys is characterized by both shared and species-specific differences in oculomotor behaviors and in the strength and structure of semantic object-directed biases.