Enilda Velazquez¹, Nelson Roque²; ¹University of Central Florida, ²The Pennsylvania State University

Visual clutter impairs search performance, interfering with top-down and bottom-up attentional guidance. Previous research suggests localized clutter impairs top-down object segmentation, leading to non-objects being included in the search set, resulting in search errors (Neider and Zelinsky, 2006). Further, global clutter disrupts bottom-up information accumulation across a scene, leading to reduced search efficiency (Wolfe et al., 2002). Quantifying clutter through perceptual grouping (Subband Entropy - SE) and feature covariance (Feature Congestion - FC), allows us to more comprehensively understand the impact of top-down vs. bottom-up guidance mechanisms (Rosenholtz et al., 2007). The present study aimed to disambiguate the effects of clutter on guidance by investigating how SE and FC clutter impact search, as a function of top-down or bottom-up mechanisms, using an ecologically valid multi-target search task. Participants searched for boat targets in marina scenes taken from a satellite image set (DOTA, Xia et al., 2018). Targets were placed according to condition: (1) in marina (in clutter), (2) in water (out of clutter). If search was guided bottom-up by clutter, we predicted that global clutter would decrease search efficiency, while local clutter would not. Whereas, if search was guided top-down by clutter, we predicted that local clutter would decrease search efficiency, while global clutter would not. Linear multilevel models suggest that global SE clutter decreases search efficiency (b = 1976.3 ms, p < .001), but global FC clutter does not. This effect reverses for local clutter, where local FC clutter decreases search efficiency (b = 477.6 ms, p < .001), but local SE clutter does not. Findings suggest that clutter may impact guidance at the level of priority map formation (Wolfe, 2021). Specifically, that SE clutter may serve to quickly accumulate target location probabilities globally, then guidance uses local FC around probable locations to select targets for further processing.