Abstract
Characterizing the conditions under which irrelevant stimuli are efficiently filtered out has long been a central driving force in attention research. Although several factors, including load and local competition, have been identified to be important, a parsimonious and straightforward account has yet to be proposed. In a series of experiments using color-word flanker and Stroop paradigms, we show that the relative saliency of distractors determines the level of distractor processing. For each subject, we first measured the unique yellow point where yellow appears neither reddish nor greenish. We then devised three saliency conditions, with the color difference between red and green being large, medium, and small for the high, medium, and low saliency conditions, respectively. With the saliency of the target manipulated while the distractor held constant, the relative saliency of the distractor was directly manipulated. Subsequently, subjects were to categorize the color of each central patch (red/reddish yellow or green/greenish yellow). When the target and distractor were spatially separated (flanker task), increasing the relative saliency of the distractor also increased the interference effect from the distractor, regardless of whether the conflict came from sensory information, semantic information, or both. When the target and distractor were spatially overlapped (Stroop task), the same pattern persisted. An individual difference analysis of sensitivity to color difference further revealed that the relative saliency of the distractor as scaled by sensitivity correlated with the interference effect. However, precueing the distractor 80ms before the target increased the interference effect only in the low saliency condition but decreased the interference effect in the medium and high saliency conditions. Thus, these results demonstrate that the relative saliency of the distractor, with saliency defined by the strength of stimuli-response association, determines distractor processing. This saliency account can also parsimoniously explain a number of (sometimes puzzling) findings in selective attention.
Supported by Graduate Research Partnership Program Fellowship and Eva O. Miller Fellowship (ZL).