Abstract
In contextual cueing experiments, participants are sensitive to the repetition of contextual elements during visual search, resulting in faster search through familiar arrays than through novel arrays (Chun & Jiang, 1998). Studies using random arrays of simple objects have found that contextual cueing is driven primarily by memory for local elements near the target location (Olson & Chun, 2002). In contrast, studies using real-world scene stimuli have found that contextual cueing is driven primarily by memory for global scene features (Brockmole et al., 2006). Here we investigated whether the context provided by global scene structure and the context provided by local search elements can be used independently to cue target location or whether both are integrated within a holistic scene representation. Participants performed a feature negative search within a 3-D rendered search display. At the center of the display was a local search array consisting of vertically oriented “dumbbell” objects. The target was missing one dumbbell segment, and participants reported which segment was missing. This local search array was embedded within a 3-D rendered real-world scene. The search array and scene were spatially segregated, allowing us to manipulate the local search context and the global scene context independently. We found that repeated local context produced contextual cueing even when global scene context did not predict target location. However, repeated global scene context did not produce contextual cueing when local array context did not predict target location. Although we did not find contextual cueing on the basis of global context alone, when both global and local cues were informative, contextual cueing was disrupted if either context was changed in a transfer block. This suggests that although contextual cueing was driven primarily by local information, global and local context were integrated into a holistic scene representation.