Set-size effects during visual search in realistic scenes.

摘要

Despite its bedrock status in the search literature, the notion of a set-size effect breaks down when ported into the real world, where objects are often difficult to delineate and typically outnumber the set-sizes used in most search experiments. To explore set-size effects in real-world contexts, we had observers search for the presence of a target (a modern military tank) in pseudorealistic landscapes created with 3d modeling software. In Experiments 1 and 2 scenes were populated with varying proportions of shrub, boulder, and tree distractors, with each distractor type varying in its degree of feature (color, shape, size) similarity to the target. In Experiment 3 observers searched for the target among a distractor set of trees, within which color-similarity to the target and distractor group heterogeneity were varied. Finally, in Experiments 4 and 5 observers searched for the target among tree distractors that were similar to the target in color, but varied in terms of density (set-sizes of 25 and 200). Findings from Experiments 1 and 2 showed that distractors that were similar to the target in the size and shape (shrubs and boulders) feature dimensions produced set-size effects when they were also similar in color to the target; larger distractors (trees) produced set-size effects regardless of target-color similarity. This latter result was supported by data from Experiment 3; search times remained unchanged as tree color was systematically manipulated. In Experiments 4 and 5 reverse set-size effects were observed; as trees were added to the display search performance improved. An analysis of eye movements showed that observers tended to fixate individual trees at smaller set-sizes, but tended to fixate the open regions between trees at larger set-sizes. Overall, these data suggest that set-size effects in realistic scenes are likely determined by multiple factors. In some cases traditional factors, such as target-distractor similarity, dictate whether or not a given distractor is included in the relevant search set, however in other cases set-size effects are influenced by non-traditional factors, such as general clutter (e.g., trees) and contextual constraints (e.g., open fields).