Abstract
Salient-but-irrelevant objects have the potential to distract attention and interfere with the task of searching for relevant objects. A very potent mechanism for reducing this interference, the (top-down) down-weighting of the distractor’s (bottom-up) saliency signal, is suggested by the dimension-weighting account. Various strands of behavioral and neuroimaging evidence indicate that people can very effectively down-weight a whole distractor dimension, but not a specific distractor feature when searching for a target that stands out from a background of homogenous non-targets (pop-out): Most notably, (a) physically identical distractors cause strong or weak interference depending on the currently searched-for target (same vs. different dimension), and (b) a distractor standing out in the same dimension as the target induces extremely reliable electrophysiological markers of attentional capture (distractor N2pc) and delays attention allocations towards the target (target N2pc), whereas a distractor standing out in a different dimension is typically directly suppressed (Pd) without attracting attention. These patterns emerge even when the task design is otherwise held constant and both types of distractor trials are randomly intermixed, and even when targets and same-dimension distractors are highly dissimilar. Electrophysiological as well as behavioral data can be well accounted for quantitatively by a dimension-weighting-based computational model of visual search that assumes a partial (though not perfect) down-weighting of the distractor whenever target and distractor stand out in different dimensions.