Abstract
Despite its bedrock status in the search literature, the notion of a set size effect breaks down almost entirely when ported into the real world, where objects are often difficult to delineate and typically far outnumber the set sizes used in most search experiments. To explore various forms of set size effects in real-world contexts, we had observers search for the presence of a target (a tank) in pseudorealistic landscapes created with 3-d modeling software. This approach allowed us to quantify the number of objects in the scene and to manipulate the types and proportions of distractor objects. Populating the landscape scenes were rocks, shrubs, and trees in varying proportions (25/25/25, 75/25/25, 25/75/25, 25/25/75). Distractor objects also varied in their color and size similarity to the target. Although RTs generally increased with the number of objects in the scene, there were exceptions to this standard set size effect. In displays dominated by rock distractors, observers searching 75 object scenes (25/25/25) were no faster at indicating target-absence compared to the comparable 125 object scenes (75/25/25). Slopes were also steeper in the target-present (∼330 ms) and target-absent (∼1470 ms) data when the dominant distractor group consisted of target-similar objects (e.g., bushes, trees), and eye movements were smaller in amplitude and more frequent under these conditions. We interpret these data as evidence for similarity-based grouping in realistic scenes. When distractors are dissimilar to the target, these distractors are more easily rejected as a group.
This work was supported by grants from the Army Research Office (DAAD19-03-1-0039) and National Institute of Mental Health (R01 MH63748)