Abstract
Previous work has shown that humans can quickly and accurately recognize objects within briefly presented natural scenes. Multiple theoretical accounts have argued for an important role of scene context adding to this process. Evidence for these accounts comes from studies that found improved recognition of objects that were presented in semantically congruent scenes (e.g., a sandcastle on a beach) relative to semantically incongruent scenes (e.g., a sandcastle on a football field). In the current study, we aimed to discriminate between three scene properties that might account for this congruency effect: 1) low-level visual features, which typically overlap more for congruent than incongruent scene-object pairs; 2) local scene properties that require focused attention to be processed, such as other objects in the scene; and 3) global scene properties, or "scene gist", that are known to be processed without the need for focused attention. Experiment 1 replicated the scene congruency effects of a previous report (Davenport & Potter, 2004). Using a new carefully controlled stimulus set, Experiment 2 showed that the scene congruency effect could not be explained by low-level feature overlap between scenes and target objects. Experiment 3 investigated whether focused attention modulates scene congruency effects. Using a location cueing manipulation, attention was focused either at the target object or at the scene background. Recognition of the target object was better when attention was focused at its location. Importantly, the effect of scene congruency on target object recognition was independent of spatial attention. These results are most consistent with the hypothesis that congruent scene context benefits object recognition through the processing of global scene properties that do not require focused attention.
Meeting abstract presented at VSS 2013