Why scrambling sometimes fails: Target specification modulates context effects in visual search

Poster Presentation 36.430: Sunday, May 17, 2026, 2:45 – 6:45 pm, Pavilion
Session: Visual Search: Features, scenes, real-world stimuli

Antje Nuthmann1, Franz Faul1; 1Kiel University, Germany

Scene context is known to facilitate visual search in images of real-world scenes by constraining where a target object is likely to appear. These contextual benefits are commonly investigated by comparing search performance for objects located in typical versus atypical positions, or in intact versus scrambled scenes. Unexpectedly, Foulsham, Alan, and Kingstone (2011) reported no effect of scene scrambling on search time when scenes were subdivided into a 4 × 4 grid. This result seems to challenge the widespread assumption that disrupting global scene structure reliably impairs search. We suspected this null effect arose from using whole grid cells as target cues, a method that may bypass the need for object identification and render contextual information unnecessary. To test this, we conducted an extended replication where targets were cued either by naming the object (word cue) or by the content of the grid cell containing the object (image cue). These cue types were crossed with scene scrambling (normal vs. scrambled) and object position typicality (typical vs. atypical), allowing us to investigate how target specification interacts with contextual disruption. Targets were always present, and we collected behavioral and eye-movement data. The results support our hypothesis: Scene scrambling and atypical object position both produced substantial increases in search time and error rate in the word-cue condition, whereas these effects were minimal in the image-cue condition, where the cue effectively removed the need for object identification. With word cues, atypical object positions were particularly disruptive, suggesting that contextual expectations play a crucial role in locating objects. In this case, the lack of context sometimes prevented target discovery altogether. These findings demonstrate that scene context supports visual search only when semantic information about the target is available, offering an explanation for prior discrepancies in the literature regarding the effects of scrambling.