Daniel Toledano¹, Tamer Abd Alrazaq¹, Dominique Lamy¹; ¹Tel Aviv University

Laboratory studies consistently show that attention and eye movements are strongly biased toward the location where a previous target was recently found. However, it remains unclear whether this robust location priming effect generalizes to more naturalistic search environments, which differ from highly controlled laboratory environments in important ways: they contain numerous distractors, complex objects, non-uniform backgrounds and spatial layouts, and strong context-related statistical regularities. To address this question, we conducted a preregistered re-analysis of a publicly available dataset (Zelinsky et al., 2019). Participants searched for either microwaves (N= 27) or clocks (N = 27) in real-world scene photographs taken from the COCO2024 dataset while their eye movements were recorded. On each trial, participants indicated whether the target was present or absent. To quantify location priming we measured both the absolute distance and the angular distance between the first saccades’ destinations and the previous target location. We then compared these distances to two baseline measures. We found a bias not only towards the previous target location but also towards the target location from two trials back. In addition, we found that participants' eye movements dwelled more on locations that were closer to the previous target location. Overall, these findings provide the first evidence that overt attention is strongly biased towards previous target locations in visually rich environments that closely mimic real-world search.