Selecting and ignoring salient objects within and across dimensions in visual search

There are two strategies for selecting relevant information from a visual scene: according to either its salience or its relevance for behavioral goals. Although there is broad evidence for the existence of both mechanisms, there has been a debate concerning the impact of top-down control on salience-based selection. We investigated whether salient but irrelevant information is filtered by the visual system and what role the organization of the visual system plays in selection. Using event-related brain potentials (ERPs), we compared processing of the same salient objects when relevant versus irrelevant to the task at hand. Observers had to detect a target singleton defined in a specific dimension (e.g., orientation), while ignoring a singleton non-target defined in either another dimension (e.g., color; Experiment 1, requiring dimension-based search) or the same dimension (Experiment 2, requiring feature-based search). For dimension-based search, the results revealed an enhanced posterior N2, indicative of singleton selection, and an enhanced P3 for target singletons, but no difference between non-target singletons defined in another dimension and blank trials. Also, N2pc results indicated that the allocation of attention was modulated by the task. In contrast, for feature-based search, some ERP enhancement was also observed for non-target singletons (defined in the same dimension as the target) in the N2, N2pc, and P3 latency ranges. This pattern argues in favor of a strong top-down influence on singleton selection that operates on dimensions rather than features.