Four experiments investigated whether a highly salient color singleton can be ignored during serial search. Observers searched for a target letter among nontarget letters and were instructed to ignore an irrelevant, highly salient color singleton that was either compatible or incompatible with the response to the target letter. The results indicate that it was possible to prevent attentional capture by the irrelevant singleton only when both the target and the distractor color were known. When either the color of the target or the color of the to-be-ignored singleton were varied over trials, the irrelevant singleton captured attention. The ability to selectively falter singleton distractors during serial search depends on the presence of an attentional set for a specific feature value of both target and distractor. In the absence of a consistently predictable feature value of both target and distractor, top-down control is not possible. Among the most fundamental issues of visual attention research is the extent to which visual selection is controlled by the properties of the image or by the intentions, goals, and beliefs of the observer (for recent reviews, see Egeth &