A unitary focus of spatial attention during attentional capture: Evidence from event-related brain potentials

Abstract

Abstract

Abstract:

Abstract  We studied whether attentional capture in vision can be elicited simultaneously at multiple locations in the visual field or whether it is always restricted to a single location. Participants searched for color singleton targets in search arrays that were preceded by spatially uninformative color cue arrays. Single cue arrays contained a unique color singleton among gray background items, and double cue arrays included two different-colored objects. Behavioral spatial cueing effects and N2pc components indicative of attentional capture were triggered by single as well as double cue arrays, but these capture effects were smaller with double cues. Response latencies were used to distinguish double cue trials where attention was captured by a left or right cue and trials where it was attracted by a cue on the vertical meridian. The N2pc was only present on horizontal capture trials, indicating that attentional capture is not triggered in parallel and equally at different locations, but is strongly weighted towards one object and one location at a time. Results show that when multiple salient visual objects compete for attentional selection, the capacity of each object to capture attention is reduced, but the focus of spatial attention tends to remain unitary.

Introduction

Visual attention is responsible for selecting a subset of all currently available visual information for access to perception, working memory, and action control. In many circumstances, task-relevant objects and events occupy one specific part of the visual field, and their selection can be effectively guided by spatial location. Space-based selection has been studied intensively over the past 30 years, and is often compared to the operation of an attentional “spotlight” (Posner, Snyder, & Davidson, 1980). According to Posner et al. (1980), this hypothetical spotlight has a unitary focus, because attention cannot be simultaneously allocated to different non-contiguous locations in visual space (see also Eriksen & Yeh, 1985; McCormick & Klein, 1990, for similar arguments). Many models of visual search (e.g., Treisman, 1988; Wolfe, 2007) also postulate a single focus of spatial attention that is allocated sequentially to different candidate target locations. However, others have argued that the focus of attention can be split and be divided across different positions (e.g., Bichot, Cave, & Pashler, 1999; Castiello & Umiltà, 1992; Kramer & Hahn, 1995). For example, separate foci of attention may be involved in multiple-object tracking tasks (Pylyshyn & Storm, 1988), which require the simultaneous monitoring of objects at nonadjacent locations (e.g., Cavanagh & Alvarez, 2005). 

The debate about the unitary versus separable nature of focal visual attention has continued for decades. It has remained unresolved, although the balance of the current experimental evidence appears to be in favor of a single-focus account (e.g., McCormick, Klein, & Johnston, 1998; see also Jans, Peters, & De Weerd, 2010, for a recent detailed review). Research in this area has typically focused on the question whether it is possible to voluntarily divide visual attention between locations when such an attentional allocation strategy is advantageous for a task at hand. However, the issue of single versus split foci of visual attention does not just arise for fully endogenous attentional control. Salient visual events are known to capture attention in an involuntary exogenous fashion (e.g., Jonides, 1981). Attentional capture by multiple salient visual events and its implications for the debate about single versus divided attentional foci has been studied relatively rarely (e.g., Wright & Richard, 2003; Yantis & Johnson, 1990). What happens when such events occur simultaneously at different locations in the visual field? Do all of them capture attention independently and in parallel, or is attention only allocated to one of these locations at any given moment? 

It is often assumed that attentional capture by salient visual events is triggered in a bottom-up fashion irrespective of currently active task sets. For example, reaction times (RTs) during visual search for shape singletons (e.g., unique diamond targets presented among distractor circles) are delayed by the presence of salient but task-irrelevant color singletons (e.g., Theeuwes, 1991), suggesting that such color singletons are able to attract attention even when they are irrelevant for current selection goals (see also Ester & Awh, 2008; Nordfang, Dyrholm, & Bundesen, 2012, for similar observations). However, the ability of salient visual objects to capture attention is often contingent on whether they match target-defining attributes. This task-set dependent nature of attentional capture was demonstrated in a series of experiments by Folk and colleagues (Folk & Remington, 1998; Folk, Remington, & Johnston, 1992; Folk, Remington, & Wright, 1994) where spatially nonpredictive feature singleton cues preceded visual search arrays. Attentional capture by these cues was assessed by comparing RTs to visual search targets that were presented at cued versus uncued locations. The critical finding was that spatial cueing effects (i.e., faster RTs to targets at cued locations) were only triggered by cues with features that matched a currently active task set, but not by nonmatching cues, indicating that only the former captured attention. For example, color singleton cues produced spatial cueing effects when participants searched for color-defined targets, but not in a task where targets were abrupt onset items. According to the contingent involuntary orienting hypothesis (Folk et al., 1992), salient visual objects such as feature singletons will capture attention only if they possess attributes that match current task goals, but not when their visual attributes are task-irrelevant. 

How does attentional capture operate when more than one salient and task-set matching visual object is present in the visual field? Is attention simultaneously and independently attracted by each of these objects, or is capture only triggered for a single location in visual space at any given moment? In other words, can attentional capture be associated with multiple foci of spatial attention? To assess this question, we modified the spatial cueing procedure introduced by Folk et al. (1992). On each trial, a cue array that included one or two colored items was followed by a search array that contained one colored target bar among three gray distractor bars. Participants' task was to respond to the orientation (horizontal or vertical) of the color singleton bar in the search array. Target arrays were preceded with equal probability by one of two types of cue arrays (see Figure 1). On single cue trials, the cue array contained one colored item (randomly red or green) among three gray items. Targets appeared at the cued location on 25% of all trials, and at one of the three uncued locations on the remaining 75% of trials. On double cue trials, one item in the cue array was red, one other was green, and the two remaining items were gray. One colored item always appeared on the horizontal meridian (to the left or right of fixation), and the other on the vertical meridian (above or below fixation). The locations of the colored items in single or double cue arrays and the locations of subsequent target bars were uncorrelated, that is, color cues were spatially uninformative with respect to the target location. 

Figure 1

 

Schematic illustration of trial sequences on single cue trials (left) and double cue trials (right). Red and green items are shown here in black and light gray, gray items are shown in white.

Because target bars were color singletons and their color (red or green) varied randomly across trials, participants were expected to adopt a feature-unspecific singleton search mode (“any color discontinuity, regardless of its value”; Bacon & Egeth, 1994). On single cue trials, both red and green cues should therefore capture attention, and this should be reflected by faster RTs for targets at cued versus uncued locations. The critical question was how attentional capture would operate on double cue trials where a green and a red item appeared together. If attentional capture can be triggered in parallel, independently, and equally strong for different locations in the visual field, each of the two colored items in double cue arrays should attract attention just as effectively as in single cue trials. Thus, behavioral correlates of attentional capture should not differ between single and double cue trials. But if attentional capture can only be triggered by one object at a time, results should be different. If only one of the two color cues was able to capture attention on each double cue trial, targets presented at the location previously occupied by one of these cues would have a 50% chance to be located within the current focus of attention. Therefore, RT spatial cueing effects indicative of attentional capture should be approximately half as large on double cue trials relative to single cue trials, where cued targets are always presented within the current focus of attention. Another possibility is that two color items, which are simultaneously present in a cue array, neutralize each other in terms of their ability to attract attention. In this case, there should be no attentional capture at all and thus no spatial cueing effects on double cue trials. 

In order to gain temporally precise insights into the nature of attentional capture on single and double cue trials, we also recorded event-related brain potentials (ERPs) during task performance. To assess whether and to what extent color cues were able to attract attention to their position, we measured the N2pc component in response to single cue and double cue arrays. The N2pc is an enhanced negativity over posterior scalp electrodes contralateral to the side of an attended stimulus that emerges between 180 ms and 240 ms after stimulus onset, and is assumed to reflect the attentional selection of candidate target items among distractors (Eimer, 1996; Girelli & Luck, 1997; Hickey, McDonald, & Theeuwes, 2006; Kiss, Van Velzen, & Eimer, 2008; Luck & Hillyard, 1994; Mazza, Turatto, Umiltà, & Eimer, 2007; Woodman & Luck, 1999). Previous ERP investigations of attentional capture have measured the N2pc in response to cue stimuli that preceded target search arrays (e.g., Eimer & Kiss, 2008, 2010; Lien, Ruthruff, Goodin, & Remington, 2008), and found that this component was triggered by feature singleton cues only when they matched current task sets. For example, Lien et al. (2008) employed a task where search arrays including a target-color letter were preceded by cue arrays containing a target-color cue and a nontarget-color item on opposite sides together with two white items. Faster RTs for targets presented at the location previously occupied by the target-color cue were accompanied by a robust N2pc contralateral to the visual field of this cue, indicative of task-set contingent attentional capture. 

In the present study, the stimulus onset asynchrony (SOA) separating cue and target arrays was sufficiently long (200 ms) to measure N2pc components in response to single cue and double cue arrays prior to ERP activity triggered by the onset of subsequent target arrays. Because the N2pc is a lateralized component that is triggered at electrodes contralateral to the visual field of an attended visual object, it is absent when such an object appears on the vertical midline. For single cue trials, the N2pc could therefore only be measured in response to cue arrays that contained a color singleton item on the left or right side, but not for arrays with a cue on the vertical meridian. For double cue arrays, the N2pc was measured relative to the side of the horizontal color cue. In these arrays, the other color cue that was simultaneously present at the top or bottom position will not trigger any additional lateralized ERP activity, which ensures that the N2pc is a pure measure of attentional capture by the color cue in the left or right visual field. The critical question was whether N2pc components in response to horizontal color cues would differ between single cue and double cue trials. If attentional capture is triggered in parallel and strictly independently for different locations, N2pc components to horizontal cues on double cue trials should be very similar to N2pc components observed for horizontal color cues on single cue trials. In other words, the additional presence of a color cue on the vertical meridian should have little if any effect on the N2pc to horizontal color cues. If two simultaneously present color cues neutralized each other in their ability to attract attention, no N2pc should be found at all on double cue trials. If only one of the two cues was able to capture attention on any given trial, the N2pc triggered by horizontal color cues in double cue arrays should be approximately half as large as the N2pc triggered by these cues in single cue arrays. 

While the finding of reduced behavioral spatial cueing effects and attenuated N2pc amplitudes on double cue as compared to single cue trials would be consistent with the predictions of a single-focus account of attentional capture, it would not necessarily be conclusive. It could be argued that attention is always captured in parallel by both color cues on double cue trials, but that the amount of attentional capture by each cue is affected by the second color item in the cue array. For example, the presence of two colored objects is likely to reduce the bottom-up salience of both objects, and thus possibly their ability to elicit attentional capture. To further distinguish between these alternatives, additional analyses compared N2pc components in double cue trials that were separated as a function of RT latencies to targets at cued locations. 

Methods

Participants

Thirteen observers were paid to participate in this experiment. One was excluded because of a large amount of horizontal eye movements. Four of the remaining 12 participants (aged 22–38 years; mean age, 30 years) were female. One participant was left-handed, and all had normal or corrected-to-normal vision. 

Stimuli and procedure

Stimuli were presented on a 22-inch Samsung wide SyncMaster 2233 LCD monitor (100 Hz refresh rate; 16 ms black-to-white-to-black response time, as verified with a photodiode; Samsung, Tianjin, China) against a black background. On each trial, a cue display preceded a search display. Both displays were presented for 50 ms and were separated by a 150 ms blank interval. Cue displays contained four sets of four closely aligned dots (0.8° × 0.8°), which were centered 4° above, below, to the left, and to the right of central fixation (see Figure 1). In each trial, one of two cue array types appeared with equal probability and in random order. In Single Cue arrays, one of the four sets of dots was colored (red or green), and the others were gray. The location and color of this singleton cue was randomly determined on each trial. Double Cue arrays contained one red, one green, and two gray sets of dots. One color cue was always located on the vertical meridian (above or below fixation), and the other on the horizontal meridian (left or right of fixation). 

Search arrays consisted of four horizontal and vertical bars (0.5° × 1.2°) that were presented at the same locations as the preceding cues. The orientation of each bar was randomly and independently selected for each trial. Each search display contained one color singleton bar that was presented randomly and equiprobably at one of the four locations. In half of all trials, the color singleton bar was red, and in the other half it was green. The three other bars were always gray. Stimulus colors were identical for cue and search arrays (CIE color coordinates for red: 0.609/0.338; green: 0.262/0.558; gray: 0.324/0.348), and all stimuli were equiluminant (12 cd/m²). 

Color singleton bars in the search arrays served as targets, and participants were instructed to report their orientation by pressing one of two vertically arranged response keys with their left or right index finger (top key: vertical target; bottom key: horizontal target). Hand-to-key mappings were swapped for each participant after three of the six experimental blocks. Because the locations of color cues and color singleton target bars were independently selected on each trial, color cues were uninformative with respect to the location of the subsequent target bar. 

Participants completed six blocks of 160 trials per block (each divided in two sub-blocks containing 80 trials, and separated by a self-paced break), resulting in a total of 960 experimental trials. Single cue arrays were presented on 64 trials per block. On 16 of these trials, targets appeared at the cued location, and on the remaining 48 trials, targets were presented at one of the three uncued locations. On 96 trials per block, double cue arrays were presented, and targets appeared at one of the two cued and one of the two uncued locations with equal probability. Before the start of the experiment and after block 3 when response hands were swapped, participants completed a practice block containing 40 trials. 

Electroencephalogram recording and data analyses

Electroencephalogram (EEG) was DC-recorded from 23 scalp electrodes mounted in an elastic cap at standard positions of the extended 10/20 system at sites Fpz, Fz, F3, F4, F7, F8, FC5, FC6, Cz, C3, C4, T7, T8, CP5, CP6, Pz, P3, P4, P7, P8, PO7, PO8, and Oz. The continuous EEG was sampled at a rate of 500 Hz, with a digital low-pass filter of 40 Hz. All electrodes were online referenced to the left earlobe and re-referenced offline to the average of both earlobes. No further filters were applied after EEG acquisition. Trials were segmented from 100 ms before to 500 ms after cue onset. Trials with artifacts (HEOG exceeding ± 30 μV; Fpz ±60 μV; all other electrodes ±80 μV), or with incorrect or missing responses to target bars were excluded from all further analyses. 

EEG was averaged separately for all combinations of cue type (Single Cue, Double Cue) and visual field of the horizontal cue (left or right). Trials with single cues on the vertical meridian (above or below fixation) were not included in the EEG analysis. N2pc components in response to cue arrays were quantified on the basis of mean amplitudes obtained in a 190–290 ms time window after cue onset at lateral posterior electrodes PO7 and PO8. All t-tests applied were two-tailed. 

Results

Behavioral performance

Trials with anticipatory RTs (faster than 200 ms) and very slow responses (RTs longer than 1000 ms) were excluded from analysis (0.6% of all trials). Mean correct RTs were subjected to a repeated-measures ANOVA with the factors cue type (Single Cue, Double Cue) and target position (cued, uncued). A main effect of cue type, F(1, 11) = 45.5, p < 0.001, was accompanied by a main effect of target position, F(1, 11) = 77.0, p < 0.001, reflecting the presence of reliable spatial cueing effects indicative of attentional capture. Most importantly, there was a significant cue type x target position interaction, F(1, 11) = 18.0, p = 0.001, demonstrating that spatial cueing effects differed between trials with single and double cues. These effects were reliably present for single cues, t(11) = 7.9, p < 0.001, as well as for double cues, t(11) = 7.4, p < 0.001, but they were almost twice as large on single cue trials as compared to double cue trials (40 ms vs. 22 ms; see Figure 2). Response errors were generally infrequent (3.3% of all trials). An ANOVA conducted for error rates revealed main effects of cue type, F(1, 11) = 9.3, p < 0.02, and target position, F(1, 11) = 21.5, p < 0.001. Error rates were higher on double cue trials (3.7%) as compared to single cue trials (2.6%), for targets at uncued (4.5%) as compared to cued positions (1.9%). There was no significant cue type x target position interaction for error rates, F(1, 11) = 3.3, p = 0.095. 

Figure 2

 

Left panel: Mean correct RTs to targets at cued and uncued locations, shown separately for single cue and double cue trials. Right panel: RT spatial cueing effects measured in single cue and double cue trials.

Additional analyses were conducted to investigate whether RT spatial cueing effects indicative of attentional capture differed between trials where cues and targets had the same or two different colors (color match vs. color nonmatch trials). For single cue trials, there was no such difference [spatial cueing effects of 38 ms and 43 ms on color match and nonmatch trials, respectively, t(11) < 1]. For double cue trials, spatial cueing effects were computed by comparing RTs to uncued targets to RTs on cued trials where a target was preceded by a color matching or nonmatching cue at the same location. Again, no reliable difference was observed [spatial cueing effects of 18 ms and 25 ms on color match and nonmatch trials, respectively, t(11) = 2.2, p = 0.07]. The absence of differential attentional capture effects on color match versus nonmatch trials confirms that participants adopted a feature-unspecific singleton search mode. 

N2pc components to single and double cue arrays

Figure 3 shows ERPs in response to single cue and double cue arrays at electrodes PO7/8 contralateral and ipsilateral to the side of a horizontal color cue and N2pc difference waveforms obtained by subtracting ipsilateral from contralateral ERPs. Note that on double cue trials, a horizontal color cue was always accompanied by a second color cue at the top or bottom position, which does not trigger lateralized ERP activity. N2pc components appear to be present for both cue types, but were considerably larger in response to single cues. This was confirmed by a repeated-measures ANOVA conducted with the factors cue type (Single Cue, Double Cue) and laterality (electrode contralateral vs. ipsilateral to the horizontal color cue) for mean amplitudes measured in the N2pc time window (190–290 ms) after cue array onset. A main effect of laterality, F(1, 11) = 21.8, p = 0.001, was accompanied by a significant interaction between cue type and laterality, F(1, 11) = 33.0, p < 0.001, confirming that N2pc mean amplitudes were reliably attenuated in response to double cues (−0.5 μV) as compared to single cues (−1.3 μV). However, t-tests conducted separately for both cue types confirmed that reliable N2pc components indicative of attentional capture were present not only for single cue arrays, t(11) = 5.3, p < 0.001, but also for cue arrays that contained two colored items, t(11) = 3.1, p < 0.01. 

Figure 3

 

Top panels: Grand-average ERPs elicited in response to color cue arrays at posterior electrodes PO7/8 contralateral and ipsilateral to a color cue in the left or right visual field, shown separately for single cue arrays (left) and double cue arrays (right). Bottom panel: N2pc difference waveforms obtained by subtracting ipsilateral from contralateral ERP waveforms at PO7/8, shown separately for single cue arrays (solid line) and double cue arrays (dashed line).

The fact that N2pc components were smaller for double cue arrays relative to single cue arrays is consistent with the hypothesis that attention was attracted by only one of the two color cues in a double cue array on each trial. In this scenario, a horizontal color cue would capture attention and thus trigger an N2pc on approximately 50% of all double cue trials. On the other half of all trials, attention is captured by the cue on the vertical meridian, which does not elicit an N2pc. Averaging across these two types of trials would yield a reduced N2pc as compared to arrays with a single color cue on the left or right side, which elicit an N2pc on all trials. To obtain more direct evidence for this interpretation, ERPs measured in response to double cue arrays were averaged separately as a function of target RT latencies. If only one of the two color cues was able to capture attention on each trial, this should be reflected in the RT distribution for trials where targets appear at one of the two cued locations. Attentional capture by a cue at the top or bottom position should result in fast RTs on trials where a subsequent target appears at this position, and slower RTs on trials where the target is presented at the position of the other (horizontal) cue. Conversely, attentional capture by a horizontal cue should produce fast RTs for cued horizontal targets, but slower RTs on trials with cued vertical targets. To separate these hypothetical “horizontal capture” and “vertical capture” trials, we sorted double cue trials with targets at cued locations on the basis of individual RT median splits conducted separately for vertical (top/bottom) and horizontal (left/right) targets. For cued vertical targets, mean fast and slow RTs across all participants were 448 ms and 578 ms. For cued horizontal targets, the respective mean RTs were 456 ms and 597 ms. We then combined trials with fast RTs to cued left/right targets and slow RTs to cued top/bottom targets (horizontal capture trials) and trials with fast RTs to cued top/bottom targets and slow RTs to cued left/right targets (vertical capture trials). Importantly, mean RTs did not differ between these two types of trials [517 ms vs. 523 ms; t(11) = 1.6, p = 0.133]. 

If attention was captured by only one of the two color cues, and the RT median split procedure was effective in separating trials with horizontal capture and trials with vertical capture, this should be reflected by a distinctive pattern of N2pc results for these two types of trials: A reliable N2pc should be observed on horizontal capture trials, where attention is attracted by a color cue on the left or right side, but not on vertical capture trials where attention is captured by a cue on the vertical meridian. Figure 4 (top panels) shows ERPs obtained in response to double cue arrays at PO7/8 contralateral and ipsilateral to the side of a horizontal color cue for these two types of trials. As predicted, a clear N2pc was present on horizontal capture trials, but not on vertical capture trials. This was confirmed by an ANOVA with the factors laterality and trial type (horizontal vs. vertical capture), which revealed a reliable interaction between both factors, F(1, 11) = 5.0, p < 0.05. Follow-up t-tests demonstrated the presence of a reliable N2pc on horizontal capture trials, t(11) = 2.9, p = 0.013. In contrast, there was no significant difference between contralateral and ipsilateral ERPs in the N2pc time window on vertical capture trials, t(11) = 1.4, p = 0.182. Any hypothesis which assumes that attention is always captured in parallel by both colored items in double cue arrays cannot easily account for this pattern of N2pc results. 

Figure 4

 

Top panels: Grand-average ERPs elicited in response to double cue arrays at posterior electrodes PO7/8 contralateral and ipsilateral to a color cue in the left or right visual field. The left panel shows ERPs for trials classified as horizontal capture trials; the right panel shows ERPs for vertical capture trials. See text for details. Bottom panel: N2pc difference waveforms obtained by subtracting ipsilateral from contralateral ERP waveforms at PO7/8 for single cue trials (left) and double cue trials (right). Difference waves are shown separately for trials where the horizontal color cue matched the color of the target on the immediately preceding trial (matching cues, solid lines), and trials where the color of the horizontal cue differed from the color of the previous target (nonmatching cues, dashed lines).

If only one color cue attracted attention on any given double cue trial, its selection could depend on the color of the target bar on the immediately preceding trial. In this case, RTs should be faster on double cue trials where the target appeared at the location of the color cue that matched the target color on the previous trial than on trials where the target was presented at the location of the other non-matching cue. In fact, RTs were virtually identical for these two types of trials [519 ms vs. 520 ms; t(11) < 1]. Likewise, RTs to targets at cued locations on single cue trials were not affected by whether the cue matched or did not match the color of the preceding target [490 ms vs. 489 ms; t(11) < 1]. There were also no color-specific intertrial priming effects for N2pc components to color cues in the left or right visual field. This is illustrated in Figure 4 (bottom panels), which shows contralateral-ipsilateral difference waveforms for single cue and double cue arrays, separately for arrays where the color cue on the horizontal meridian had the same color as the target on the preceding trial (matching cues) and arrays where the color of this cue differed from the preceding target (nonmatching cues). There were no reliable N2pc amplitude differences between matching and nonmatching cues for either single cue or double cue arrays [both t(11) = 1.2; both ps = 0.25]. 

Discussion

The aim of the current study was to investigate the operation of attentional capture during color singleton search in response to cue arrays that contained either one or two color singleton objects. Can attention be captured in parallel and strictly independently by two simultaneously present salient visual objects, or is only one of these objects able to attract attention to its location at any time? Is the capacity of one of both items to capture attention reduced relative to a situation where only a single salient visual event is present? Or is attentional capture eliminated altogether when two competing color singletons are present in the visual field? The pattern of behavioral and electrophysiological attentional capture effects observed in the present experiment conclusively rules out some of these scenarios. The first notable finding was that reliable behavioral spatial cueing effects and N2pc components were triggered not only on single cue trials, but also on trials where cue arrays included two colored items. This demonstrates that attentional capture is not abolished when two equally salient color singletons appear simultaneously. In other words, these objects do not neutralize each other in terms of their ability to attract attention. The question is therefore whether single-focus or multiple-foci models provide a better account of the mechanisms of attentional capture on double cue trials. 

Although behavioral and electrophysiological correlates of attentional capture were reliably present on double cue trials, these effects were smaller than on single cue trials: Behavioral spatial cueing effects were almost exactly twice as large on single as compared to double cue trials (40 ms vs. 22 ms), and N2pc components triggered by horizontal color cues in double cue arrays were significantly attenuated relative to single cue arrays (Figure 3). These results are inconsistent with a scenario where attentional capture is triggered in parallel and strictly independently by two simultaneously present color cues. If this was the case, behavioral spatial cueing effects of similar size should have been observed on single cue and double cue trials. Furthermore, N2pc amplitudes in response to horizontal color cues should have been unaffected by the presence (on double cue trials) or absence (on single cue trials) of another colored item on the vertical meridian. The observation that attentional capture effects, as reflected by RTs and N2pc components, were reliably reduced on double cue trials thus rules out a multiple-foci model which assumes that attentional capture can be triggered simultaneously and entirely independently at different locations in the visual field. 

The observation that behavioral and electrophysiological correlates of attentional capture were attenuated on double cue trials but remained reliably present is consistent with a different dual-foci account which assumes that attention was attracted in parallel by both colored items in double cue arrays, but that the capacity of either color singleton to capture attention was reduced relative to single cue arrays. This may be due to the fact that colored items were less salient in the context of double cue arrays: Lower bottom-up salience implies reduced attentional capture, and thus smaller RT spatial cueing effects and N2pc components on double cue trials. Previous spatial cueing studies have demonstrated that task-set contingent attentional capture is primarily driven by top-down search intentions and not bottom-up salience. For example, Lien et al. (2008, experiment 2) employed color cue arrays that were similar to the double cues used in the present study (two different-colored items accompanied by two white items), except that only one colored item matched the current target color while the other was task-irrelevant. RTs were about 50 ms faster for targets at the location of the task-set matching color cue as compared to targets at uncued locations, which is very similar to the spatial cueing effects observed for single cue trials in the present study. This suggests that the mere presence of an additional color item in a cue array does not affect the magnitude of attentional capture. (See also Eimer, Kiss, Press, & Sauter, 2009 and Lamy, Leber, & Egeth, 2004 for further evidence against a major role of bottom-up salience in task-set contingent attentional capture). While these previous studies showed that the presence or absence of additional task-irrelevant color objects has little effect on the ability of task-set matching color items to attract attention, the two colored items in the double cue arrays of the present experiment both matched the current task set (“any color singleton”). Therefore, the possibility remains that both color cues did capture attention equally and in parallel, but that the overall amount of attentional capture was reduced relative to single cue trials. 

An alternative explanation for the presence of reduced but reliable behavioral and electrophysiological attentional capture effects on double cue trials is provided by a single-focus scenario where only one of the two color singleton cues is able to attract attention to its location on any given double cue trial. If this was the case, targets presented at one of the two cued positions will be located within the current focus at attention on approximately half of all trials and outside this focus on the other half, and this will result in a reduction of spatial cueing effects on RTs for double cue as compared to single cue trials. Along similar lines, color cues on the left or right side (which trigger an N2pc when attended) and color cues on the vertical meridian (which do not produce an N2pc) will each attract attention on roughly the same number of trials, thus producing an attenuated N2pc relative to cue arrays with a single colored item on the horizontal meridian. To distinguish between this single-focus model and the alternative dual-foci account, we contrasted N2pc components for different types of double cue trials that were separated on the basis of RT median splits. If the single-focus account is correct and attention was captured either by the horizontal or by the vertical color cue on any given trial, these two different loci of attentional capture should be dissociable on the basis of RTs to targets at one of the two cued locations: Attentional capture by a cue on the horizontal meridian will result in fast RTs to subsequent cued horizontal targets and slower RTs to targets that appear at the location of the other (vertical) color cue. Vertical capture trials will be characterized by the reverse pattern of target RTs. Because N2pc components are only triggered when attention is directed to an object in the left or right visual field, but not when an object on the vertical meridian is attended, the N2pc should be present on horizontal capture trials but not on vertical capture trials. The fact that this pattern of N2pc results was indeed observed (Figure 4) supports a single-focus account of attentional capture. 

The reduction of behavioral and electrophysiological attentional capture effects on double cue trials relative to single cue trials might also be consistent with another version of a single-focus account, which assumes that attention was allocated to a larger contiguous region of visual space that included both the horizontal and the vertical color cue position. In this scenario, reduced capture on double cue trials would reflect an inverse relationship between the size of the attentional focus and the efficiency of attention-induced modulations of target processing. The distribution of attention in the visual field and the possibility of dividing attention between noncontiguous locations should ideally be tested with dense stimulus arrays where some stimuli are located between the postulated separate attentional foci (see Jans et al., 2010, for a detailed methodological argument). As cue and target arrays contained only two horizontal and two vertical items in the current study, attentional capture at positions between the two color cue locations in double cue arrays could not be assessed. However, if attention was distributed equally across a single larger area of visual space that encompassed both color cue locations on double cue trials, an N2pc of equal magnitude should be elicited by horizontal cues on all of these trials. In other words, there should be no evidence for separable types of trials where attention was attracted either by the horizontal or by the vertical cue. The fact that the RT median split analyses revealed reliable N2pc components only for horizontal capture trials, but not for vertical capture trials (Figure 4), is inconsistent with the existence of a unitary large focus of attention on double cue trials. 

Overall, the current findings rule out a strong version of a dual-foci account of attentional capture, which assumes that attention will be allocated in parallel and entirely independently to the locations of two simultaneously present salient color objects. They also appear inconsistent with a modified version of this account, which assumes that attention is captured in parallel and equally by both color cues, albeit to a lesser degree than under conditions where only one color singleton is present. In particular, the presence of a reliable N2pc on horizontal capture trials, and its absence on vertical capture trials, strongly suggests that the distribution of attention in response to double cue trials was highly unequal, with attention strongly weighted in favor of one of the two locations that were occupied by a colored item. (See also Bundesen, 1990, and Logan, 1996, for similar considerations with respect to the computation of attentional weights.) What are the factors that determine which of these objects will preferentially attract attention on any given trial? One possibility is that the color of the target bar on the preceding trial biases attentional capture on the next trial. However, when capture effects were analyzed as a function of the preceding target color, no evidence for such a link was found. Responses to target bars were not faster when they appeared at the location of a cue that matched the color of the target on the preceding trial than when they were presented at the location of a nonmatching color cue. Likewise, N2pc amplitudes were unaffected by the match or mismatch between horizontal color cues and the color of the preceding target. In other words, target color on trial n-1 did not affect attentional capture by color cues on trial n, and this was the case not only for double cue trials but also for single cue trials (Figure 4, bottom panel). The absence of color-specific intertrial priming effects is not entirely unexpected, as participants searched for color singleton targets whose feature value (red or green) was unpredictable, and were thus likely to adopt a feature-unspecific singleton search mode. The lack of intertrial color priming provides evidence that target colors were not individuated because both colors were equally task-relevant. As the preceding target color had no effect on which of the two colored items in double cue arrays captured attention, it is most likely that this was determined randomly, possibly on the basis of small baseline fluctuations in spatial bias at the moment when a cue array was presented. 

The observation that attention was not captured equally by both color items in double cue arrays, but was instead strongly biased towards one of these two items on any given trial is obviously more in line with a single-focus account of attentional capture than with the alternative dual-foci scenario. But this conclusion needs to be qualified. If attentional capture was exclusively triggered by only one of the two color items in double cue arrays on any given trial, capture effects on horizontal capture trials (as identified by our RT median split procedure) might be expected to be equivalent to the effects observed on single cue trials. In other words, N2pc components on horizontal capture trials should be similar in size to N2pc components triggered by horizontal color cues in single cue arrays, because both would reflect a single attentional focus on the left or right side. In fact, the comparison of Figures 3 and 4 (top left panels) reveals that the N2pc on horizontal capture double cue trials was considerably and significantly smaller than the N2pc on single cue trials, t(11) = 5.0, p < 0.001. Although this difference might be partly due to the fact that the RT median-split procedure used to separate horizontal and vertical capture on double cue trials is unlikely to have worked with perfect accuracy (i.e., some vertical capture trials may have been incorrectly classified as horizontal capture trials, and vice versa), it is likely to also reflect a real difference between attentional capture on single and double cue trials. When two task-set matching visual objects are simultaneously present in the visual field, competitive interactions between these objects may attenuate attentional capture by one of these objects, which would result in reduced N2pc amplitudes relative to single cue trials. This possibility needs to be investigated more systematically in future experiments. 

The question whether the focus of spatial attention is unitary or can be strategically divided between non-contiguous locations has remained controversial (e.g., Jans et al., 2010). In the present study, we investigated whether attentional capture by salient colored objects during color singleton search can be triggered independently at multiple locations in the visual field or whether it is restricted to a single spatial location. Results indicate that when two color objects are present simultaneously, the allocation of attention is strongly biased towards one of these objects, in line with a single-focus account of attentional capture. They also show that relative to a situation where only one color singleton is present, attentional capture effects are reduced when two items compete for attentional selection. 

Acknowledgments

This research was supported by the Swiss National Science Foundation (PBFRP1-140020) and the São Paulo Research Foundation (2011/20974-4). 

Commercial relationships: none. 

Corresponding author: Martin Eimer. 

Email: [e-mail]. 

Address: Department of Psychology, Birkbeck College, University of London, London, UK.