Top-down control and bottom-up priming

The present study explored the mechanisms involved in the contingent capture phenomenon using a variant of the classic precueing paradigm of Folk et al. (1992). Rather than keeping the target fixed over a whole block of trials (as traditionally done with contingent capture experiments), we encouraged participants to adopt a top-down set before each trial. If top-down attentional set determines which property captures attention, as is claimed by the contingent capture hypothesis, then one would expect that only properties that match the top-down set would capture attention. We showed that even though participants knew what the target would be on the upcoming trial, both relevant and irrelevant properties captured attention (Experiment 1). An inter-trial analysis (Experiment 1 and 2) showed that previous contingent capture findings may to a large extent be explained by inter-trial priming. In addition, when participants were further forced into adopting the required top-down set (Experiment 3 and 4), irrelevant cues were suppressed, suggesting that top-down control might operate through disengagement of attention from the location of a property that does not match top-down goals. The present findings suggest that top-down control and inter-trial priming have their own distinct contributions to the contingent capture phenomenon. Top-down control and bottom-up priming 3 What is top-down about contingent capture? Since the early 1990s, there has been a considerable debate regarding the extent to which the allocation of spatial attention is under top-down or bottom-up control. On the one hand it has been argued that the allocation of attention is initiated voluntarily in line with the behavioural goals (Folk, Remington, & Jonhnston, 1992). Alternatively, it has been claimed that events that are salient enough are selected regardless of the current top-down set (Theeuwes, 1992, 1994). In the latter case, when objects receive priority of processing independent of the volitional goals of the observer, one refers to this as attentional capture. When such an object not only captures attention but also triggers an exogenous saccade to the location of the event, this is referred to as oculomotor capture (Theeuwes, Kramer, Hahn, Irwin, & Zelinsky, 1999). Evidence for bottom-up attentional capture comes from the irrelevant singleton paradigm developed by Theeuwes (Theeuwes, 1991, 1992, 1994). In this task, participants search for one particular feature singleton (the target) while in some trials another irrelevant salient singleton (the distractor) is also present. Critically, the presence of the irrelevant singleton increases the time to find the relevant singleton. Reaction time (RT) cost led Theeuwes (1991, 1992) to conclude that the irrelevant salient singleton captured attention automatically. On the basis of these findings, Theeuwes (see also Hickey, McDonald, & Theeuwes, 2006; Schreij, Owens, & Theeuwes, 2008; Theeuwes, 2004) argued that attentional capture is basically bottom-up and not affected by volitional top-down attentional set. In a recent study, Theeuwes and Van der Burg (2008) explicitly addressed the question whether volitional top-down set can affect attentional capture in a variant of the additional singleton paradigm in which the exact target was unknown (i.e., target was either a diamond among circles or a circle among diamonds). On each trial an instructional cue (the word “diamond” or “circle”) indicated with 100% validity the identity of upcoming target. For example, when the word “diamond” was presented participants knew that the target singleton on the upcoming trial would be a diamond. The results showed that the instructional cue was effective in reducing RT, indicating that participants actively used the cue. However, at the same time, this top-down set could only reduce but not eliminate the distracting effect of the irrelevant colour singleton in the additional singleton paradigm. The idea that selection is completely under volitional top-down control comes from a different paradigm known as the spatial precueing paradigm developed by Folk and colleagues (Folk et al., 1992; Folk, Remington, & Wright, 1994). In this paradigm, the Top-down control and bottom-up priming 4 search display is preceded by a cue display. In their classic 1992 study (Folk et al., 1992) the target display consisted of either a color or an onset singleton and observers were required to identify the unique element. In the color display the target was red while the other three elements were white. In the onset display, only one element was presented, and so the target was characterized as being the only element presented with an abrupt onset. Immediately preceding the target display at an 150 ms SOA, a cue display was presented: this cue display either consisted of a color cue (in which one location was surrounded by red dots and the other three locations were surrounded by white dots) or an onset cue (in which one location was surrounded by an abrupt onset of white dots and the remaining locations remained empty). All conditions were factorially combined and the target type remained constant within block of trials. The critical finding of Folk et al.’s studies was that only when the search display was preceded by a to-be-ignored featural singleton (the "cue") that matched the singleton for which observers were searching, did the cue capture attention. Thus, when searching for a red target singleton, attention automatically shifted to the location of the irrelevant red cue that preceded the search display, while the irrelevant onset had no effect on performance. The result suggests that the top-down attentional set determines the selection priority: when observers are set for a particular feature singleton, only elements that match this top-down attentional set will capture attention. Feature singletons that do not match top-down attentional sets will simply be ignored: “With a control setting established, events exhibiting the critical properties will involuntarily summon attention, whether or not the event is actually relevant to task performance. Stimuli not exhibiting these properties will not involuntarily summon attention” (p. 1041, Folk et al., 1992). The underlying theoretical notion of the contingent capture hypothesis proposed by Folk and colleagues is that capture is fully contingent on the top-down set adopted by the observer and salient objects with irrelevant properties are simply filtered out (Folk & Remington, 1998). On the other hand, finding attentional capture by an event that does not exhibit target-related properties while the appropriate attentional set has been established would constitute a violation of the strong version of the contingent capture hypothesis. The contingent capture hypothesis appears to be the generally accepted way to account for top-down control in visual search (see recent reviews by Burnham, 2007; Rauschenberger, 2003 #3053, but see Schreij et al., 2008; Theeuwes, 2004). The present study was designed to explore the mechanisms involved in contingent capture and pursued two goals. First of all, since in the classic precueing paradigm the target remained the same for a whole block of trials (e.g., search for color singleton), it leaves the possibility that such fixed attentional set for the target feature was not truly Top-down control and bottom-up priming 5 top-down, but instead driven by automatic inter-trial priming (Maljkovic & Nakayama, 1994). This was examined in Experiment 1 and 2, in which we mixed the trials with onset and color targets within a block, preventing participants from adopting a top-down set for a specific feature. If inter-trial priming can explain contingent capture then we expect the same contingent capture pattern of results here: only the cues that match the target on the previous trial should capture attention. The second goal of the present study was to test whether a top-down set is capable of guiding attention in a contingent fashion on a trial-by-trial basis. Similar to Theeuwes and Van der Burg (2008) we presented an instructional cue at the beginning of each trial indicating with 100% validity the target that would be presented on the upcoming trial (Experiment 1). In Experiment 3 the cue indicated which target, onset or color, participants had to search for and respond to. They had to withhold their response if a different target was presented. In Experiment 4 we even let participants choose themselves which target dimension they wanted to search for and respond to. We employed a variant of the classic precueing paradigm of Folk et al., (1992) which allowed us to determine whether the precue captured spatial attention or not. If visual selection is truly dependent on volitional top-down set, we expect basically the same results as reported by Folk et al. (1992): only features that match the current attentional set should capture attention. However, if it turns out to be impossible to prepare for a specific target on the upcoming trial in a top-down way, then one expects that not only the singleton that matches the top-down set would capture attention, but also that the feature singleton that does not match the top-down set would capture attention. Experiment 1 Experiment 1 examined the possibility of creating a top-down set for a specific target feature on a trial-by-trial basis. The task used was basically the classic precueing paradigm of Folk and colleagues (1992), in which participants had to search either for a color or onset singleton. The search display was immediately preceded by a nonpredictive color or onset cue. Every trial began with an instructional cue (“RED” or ‘WHITE”), which with 100% validity indicated the feature of the upcoming target (see Figure 1). Using the word cue was crucial to our goal of inducing a top-down set, since it ensured that a top-down set was not contaminated by any bottom-up information about the upcoming feature (Theeuwes, Reimann, & Mortier, 2006; Theeuwes & Van der Burg, 2008). This condition was compared to a condition in which a neutral cue (“NEUTRAL”) was provided. If top-down set created on every trial can influence selection, then only the cues that match this set would produce a cueing effect. However, if such top-down set cannot modulate selection, then all types of cues, even the ones that do not match Top-down control and bottom-up priming 6 top-down set would produce cueing effects.