The relative roles of endogenous (voluntary and goaldirected) attention and exogenous (involuntary and stimulus-driven) attention in visual search are still under dispute. There is considerable evidence that perceptually salient but task-irrelevant visual events can capture attention independently of current task sets. For example, reaction times (RTs) to shape singleton targets presented among nontarget shapes (e.g., a diamond target among circle distractors) are delayed when a salient but taskirrelevant color singleton is present in the visual search array, relative to trials without a color distractor (e.g., Theeuwes, 1991). This suggests that color singletons capture attention due to their bottom-up salience, and irrespective of current task set. However, other studies (e.g., Folk & Remington, 1998;Folk, Remington, & Johnston, 1992;Folk, Remington, & Wright, 1994) have suggested that the ability of salient visual events to capture attention is determined by current top-down task sets. In these studies, spatially nonpredictive singleton cues preceded visual search displays, and the presence or absence of attentional capture was inferred from the pattern of spatial cuing effects. Faster RTs in response to visual search targets at cued rather than uncued locations were found only when cue features matched the current task set (e.g., for color singleton cues in blocks where targets were also color singletons), but not when cue features were task irrelevant. These findings support the contingent involuntary orienting hypothesis (Folk et al., 1992), which postulates that attentional capture will be triggered by salient visual stimuli only when these possess attributes that are currently task relevant.To resolve the dispute between the salience-based bottom-up account of attentional capture advocated by Theeuwes (1991Theeuwes ( , 1992 and the hypothesis that capture is mediated by top-down task set (Folk et al., 1992), Bacon and Egeth (1994) have suggested that attentional capture can be determined either by bottom-up salience or by top-down control, depending on which generalized search strategy is used in a given task context. In search tasks where target items are feature singletons, participants may choose to adopt a singleton detection mode, where they search for any feature discontinuity irrespective of its value. When this search mode is adopted, salient but task-irrelevant visual singletons will capture attention, as observed by Theeuwes (1991). However, this singleton search mode is not available in search tasks where targets are not singletons. Here, participants have to adopt a more finely tuned feature search mode, where they search for specific target-defining features or dimensions. When feature search is used, singletons that do not possess targetdefining features will not capture attention, in line with the observations of Folk et al. (1992). Bacon and Egeth demonstrated that RT costs associated with a salient but task-irrelevant color singleton distractor emerge only when shape targets are themselv...