The Interaction of Clutter and Scene Size on Visual Search in Natural Scenes

Poster Presentation 36.359: Sunday, May 19, 2024, 2:45 – 6:45 pm, Banyan Breezeway
Session: Visual Search: Cueing, context, scene complexity, semantics

Wentao Si1 (), Lindsay Houck1; 1Bates College

It is well understood that content in natural scenes impacts attention. Clutter, the level of disorder within a scene, also impacts visual search, where search time increases with scene clutter. Clutter is operationalized as feature congestion, edge density, and subband entropy, which all correlate with target response time (RT). Clutter also correlates with object quantity, though this depends on scene size - larger scenes have higher object capacities. It is less known how scene size interacts with visual clutter to impact search time. Here, we used stimuli from Park, Konkle, and Oliva (2015) with visual clutter and scene size ratings to test this interaction. Scenes were categorized by dimensions of clutter (low, medium, high) and size (small, medium, large) from levels 1, 3, and 5 from the stimuli. 49 online participants completed 216 visual search trials, half target-present and half target-absent, with an 18-item T and L array overlaid on the scene. A linear mixed-model analysis revealed a main effect of scene clutter on RT (F(2, 9366.1) = 116.55, p < 0.001), with more clutter increasing RT, but no main effect of scene size (F(2, 9366.1) = 2.33, p = 0.097). However, the interaction between clutter and size was significant (F(4, 9366.1) = 33.32, p < 0.001): for low and medium clutter, RT increased with scene size, but for high clutter, RT decreased with scene size. A post-hoc analysis revealed no impact of the background contrast surrounding the target on RT, ruling out low-level feature contrast in more cluttered scenes as a confounding factor. This reveals a previously unknown interaction between scene size and clutter beyond the operationalized measures of clutter as visual disorder within a scene. Future work will untangle this interaction with eye-tracking and other measures of visual clutter, such as feature congestion.

Acknowledgements: Thanks to the Bates Neuroscience Program for funding data collection for this study