Psychological Research (2012) 76:8–19 DOI 10.1007/s00426-011-0329-4
ORIGINAL ARTICLE
Unique sudden onsets capture attention even when observers are in feature-search mode Thomas M. Spalek • Matthew R. Yanko Paola Poiese • Hayley E. P. Lagroix
•
Received: 10 July 2010 / Accepted: 11 March 2011 / Published online: 30 March 2011 Ó Springer-Verlag 2011
Abstract Two sources of attentional capture have been proposed: stimulus-driven (exogenous) and goal-oriented (endogenous). A resolution between these modes of capture has not been straightforward. Even such a clearly exogenous event as the sudden onset of a stimulus can be said to capture attention endogenously if observers operate in singleton-detection mode rather than feature-search mode. In four experiments we show that a unique sudden onset captures attention even when observers are in feature-search mode. The displays were rapid serial visual presentation (RSVP) streams of differently coloured letters with the target letter defined by a specific colour. Distractors were four #s, one of the target colour, surrounding one of the non-target letters. Capture was substantially reduced when the onset of the distractor array was not unique because it was preceded by other sets of four grey # arrays in the RSVP stream. This provides unambiguous evidence that attention can be captured both exogenously and endogenously within a single task.
Introduction What are the rules that govern the selection of a visual target-object presented among other items displayed in close spatial proximity or temporal contiguity? Two sources of attentional control have been proposed: goal-directed (also known as endogenous or top-down) and stimulusT. M. Spalek (&) M. R. Yanko H. E. P. Lagroix Department of Psychology, Simon Fraser University, 8888 University Drive, Burnaby, BC V5A 1S6, Canada e-mail:
[email protected] P. Poiese Centro Polifunzionale Don Calabria, Verona, Italy
123
driven (also known as exogenous or bottom-up). When attentional control is goal-directed, selection is said to be governed by factors that implement the observer’s strategic objectives or attentional set. For example, when looking for a red book amongst books of many other colours, selection is driven by goal-directed factors that act as a filter set to pass only red objects for further processing. Alternatively, when under stimulus-driven control, selection is said to be governed by involuntary factors tied to specific stimulus characteristics. For example, the sudden onset of a stimulus has been said to attract attention reflexively and independently of the observer’s attentional set (e.g., Yantis & Jonides, 1984). Evidence concerning the influence of goal-directed and stimulus-driven factors on visual selection has been obtained in studies of a phenomenon called attentional capture. Broadly defined, attentional capture occurs when a task-irrelevant stimulus involuntarily receives attentional priority (Folk & Remington, 1998; Theeuwes, 1992). Two contrasting perspectives have emerged regarding the influence of stimulus-driven and goal-directed factors in attentional capture. One is that visual selection is exclusively stimulus-driven, at least in the first 150 ms of vision (Theeuwes, 1994). On this account, salient stimuli are said to capture attention regardless of the observer’s attentional set. In an opposite viewpoint, known as the contingent involuntary orienting hypothesis, visual selection is said to be governed entirely by the observer’s attentional set, namely, to be goal-directed (Bacon & Egeth, 1994; Folk, Remington, & Johnston, 1992). In the studies of Bacon and Egeth (1994) and of Folk et al. (1992), the observers searched for a target that was defined either as a singleton (i.e., the most salient item in the display) or as possessing a specific feature (e.g., the colour red). In either case, visual selection was said to be
Psychological Research (2012) 76:8–19
goal-directed: in the former case, the observers were said to be in singleton-detection mode; in the latter they were said to be in feature-search mode. A task-irrelevant stimulus that either matched or did not match the observer’s attentional set, was presented some 150 ms before the target display. The critical finding was that the task-irrelevant stimulus captured attention only when it matched the observer’s attentional set. The salience-driven (Theeuwes, 1992) and the contingent-involuntary-orienting models (Folk et al., 1992) lie at opposite poles on a continuum of models of attentional control (e.g., Cave & Wolfe, 1990; Lamy, Leber, & Egeth, 2004; Treisman & Sato, 1990; Wolfe, 2006). The issue then becomes one of assessing the relative contributions of stimulus-driven and goal-directed factors to visual selection. The present study examined this issue with a variant of a paradigm first used by Folk, Leber, and Egeth (2002, Experiment 2). In that study the display consisted of a stream of letters presented in the centre of the screen in rapid serial visual presentation (RSVP). The target was a red letter for one group, and a green letter for another group. The remaining letters in the RSVP stream were a mixture of other colours. The critical manipulation was the presentation of four number signs (#) surrounding one of the non-target items in the RSVP stream. Except for the condition in which there were no surrounding #s (No-Distractor condition), the #s could appear either two items before (Lag 2), one item before (Lag 1), at the same time as (Lag 0), or one item after the target (Lag -1). Furthermore, the #s could be either all grey or three grey with one green, or three grey with one red, with equal probability. The finding of principal interest was that capture was in evidence only when the chromatic # was the same colour as the target (Colour-Match condition). When the chromatic # was in the non-target colour (ColourMismatch condition), performance did not differ from the condition in which the #s were all grey (Grey condition). This finding was taken as evidence that stimulus-driven factors—as indexed by the Colour-Mismatch condition— were not capable of capturing attention. The conclusion that attention cannot be captured in a stimulus-driven manner, however, is open to question on the grounds that an important source of stimulus-driven salience—sudden stimulus onset—may have been overlooked in the study of Folk et al. (2002). In that study, as in all extant studies of contingent capture, the distractor that shared the target’s defining feature (i.e., its colour) was presented at a screen location that had been blank up to that point. Thus, the array of # distractors had a sudden onset and, as noted above, it is known that sudden onsets can trigger stimulus-driven attentional capture (e.g., Yantis & Jonides, 1984). It is, therefore, possible that capture might
9
have been mediated by the sudden onset of the distractor array, thereby, introducing a stimulus-driven source of attentional capture even though the observer was in feature-search mode. Some support for this option comes from Folk et al.’s finding that, relative to a control condition in which no distractors were presented, attentional capture was in evidence in all conditions in which the distractor array had a sudden onset, even if it did not contain a # of the same colour as the target. At any rate, even if the sudden onset was not itself the determining factor, it may have enhanced capture by the # that shared the target’s defining feature. The present experiments were designed to examine two related issues. First, whether a colour singleton can capture attention in the absence of a unique sudden onset and, second, whether goal-directed attentional capture is still in evidence after the effects of colour singletons and unique sudden onsets have been factored out.
Experiment 1 An important preliminary consideration is to verify that the pattern of results reported by Folk et al. (2002) can be replicated using our paradigm. Notably, in the context of the issue of whether a unique sudden onset can capture attention when the observer is in feature mode, it is important to verify that performance in a condition in which the distractors do not share the target’s defining feature is impaired relative to a condition in which no distractors are presented. In essence, this involves a replication of Folk et al. Experiment 2. In some earlier studies of attentional capture the distractor conditions were blocked (e.g., Lamy et al., 2004) and in other studies they were mixed randomly across trials (e.g., Folk et al., 2002). It is not known whether the strength of attentional capture is affected by whether the conditions are blocked or mixed. It is possible, e.g., that the procedure of blocking the conditions might induce the observers to adopt different strategies in different blocks. Therefore, a second objective of the present experiment was to investigate this issue by replicating Experiment 2 of Folk et al. using a blocked rather than a mixed design. Method Observers Twenty-four undergraduate students at Simon Fraser University participated in the experiment either for class credit or for payment. All had self-reported normal or correctedto-normal visual acuity and colour vision.
123
10
Apparatus and stimuli Stimuli were displayed on an NEC 17-inch AccuSink 70 computer monitor driven by a 1024 9 768 graphic board at the rate of 70 Hz. The experiment was implemented in E-Prime software (version 1.1). The observers viewed the screen in a dimly lit room at a distance of approximately 60 cm. The stimuli consisted of an RSVP stream of uppercase letters presented in the centre of the black screen. Each letter measured 1.3° in height and had a maximum width of 1.2°, with a stroke width of 0.3°. On trials in which a distractor array was present, one of the letters in the stream was surrounded by four #s whose inner edges were located at 5.2° above, below, right, and left of the centre of the letter. Design and procedures The sequence of events on any given trial is illustrated in the left-hand panel of Fig. 1. Each trial began with a small white fixation cross which remained in the centre of the screen until the observer pressed the spacebar. After a 500-ms blank interval, the RSVP stream of 15 letters was presented. Each letter was displayed for 56 ms, followed by a 42-ms blank interval, yielding an SOA of 98 ms. One of the letters in the stream was defined as the target by its colour: red (RGB 145, 0, 0) for 12 observers, green (RGB 0, 86, 0) for the remaining 12. The colour of the non-target letters was chosen randomly from four possible colours: grey (RGB 69, 69, 69), blue (RGB 0, 0, 255), purple (RGB 123, 0, 123), and brown (RGB 88, 63, 0). These RGB values were selected so as to produce a luminance value of 13 cd/m2 for every colour, as measured by a Minolta CS100 Chroma Meter. On each trial, the letters were selected randomly without replacement from the English alphabet, with the exception of I, O, Q, and Z. Across trials, the target appeared equally often in positions 8 through 12 in the RSVP stream. Both when the target was red and when it was green, there were four distractor conditions. In the No-Distractor condition, the RSVP stream contained only letters. In the Grey-Distractor condition, the four #s surrounding one of the letters were all grey. In a third condition, three #s were grey and one was red; this condition was denoted as the ‘‘Colour-Match’’ condition on trials in which the target was red, and as the ‘‘Colour-Mismatch’’ condition on trials in which the target was green. In the fourth condition, three #s were grey and one was green; this condition was denoted as the ‘‘Colour-Match’’ condition on trials in which the target was green, and as the ‘‘Colour-Mismatch’’ condition on trials in which the target was red. Thus, on 25% of the trials, the colour of one of the four #s matched the colour of the target. When present, the #s appeared and disappeared with the letter they surrounded.
123
Psychological Research (2012) 76:8–19
The four distractor conditions were crossed with four temporal lags between the distractor array and the target. The distractor array could appear simultaneously with the target (Lag 0), in the frame following the target (Lag -1), or one frame (Lag 1) or two frames (Lag 2) before the target. In practice, the fact that in the No-Distractor condition the #s were the same colour as the background is tantamount to saying that Lag was dummy-coded in that condition. The 16 conditions resulting from the factorial combination of 4 distractor conditions and 4 lags were presented randomly with equal probability throughout the experimental session. The experiment comprised 20 practice trials followed by 25 trials for each of the 16 conditions for a total of 400 trials. The observer’s task was to identify the target letter and to type it on the keyboard, guessing if unsure.
Results and discussion The mean proportion of correct target identifications across lags is illustrated in the right-hand panel of Fig. 1. The data were entered in a mixed-factor analysis of variance (ANOVA) with one between-subjects factor: Target colour (red or green), and two within-subject factors: Distractor (No Distractor, Grey Distractor, Colour-Matched Distractor, and Colour-Mismatched Distractor), and Lag (-1, 0, 1, and 2). The analysis revealed significant effects of Lag, F (3,66) = 21.77, p \ .001, MSE = .013, and Distractor, F (3,66) = 36.48, p \ .001, MSE = .008. The interaction effect between Lag and Distractor was also significant, F (9,198) = 8.70, p \ .001, MSE = .007. The effect of Target colour was significant, F (1,22) = 4.78, p = .040, MSE = .069, suggesting that green-coloured targets were somewhat less perceptible than red-coloured targets. However, no interactions involving Target colour were significant. The pattern of results in Fig. 1 is remarkably similar to that found in Folk et al. (2002) Experiment 2. As was the case in that study, simple main-effects analyses yielded significant effects of distractor condition for both lags 1 and 2, F (3,66) = 14.22, p \ .001, MSE = .008, and F (3,66) = 34.40, p \ .001, MSE = .010, respectively. The direction of all pairwise comparisons was predicted on the basis of the results reported by Folk et al. (2002). Therefore, one-tailed probabilities were used for all pairwise comparisons. Planned pairwise comparisons of the results for Lag 1 showed that the Colour-Match condition differed significantly from all the other conditions (all p’s \ .005), which did not differ from one another (all p’s [ .176). Planned pairwise comparisons of the results for Lag 2 showed a more complex pattern. As was the case for Lag 1, the Colour-Match condition differed significantly
Psychological Research (2012) 76:8–19
11
Fig. 1 Left-hand panel schematic representation of events on any given trial in Experiment 1. The target letter was coloured red for one group of observers and green for another group. Depicted in the figure is the case in which the target was red, rendered in solid bold font. There were four distractor conditions in three of which a non-target letter was surrounded by four # distractors. The # distractors were either all grey (Grey distractor condition), three grey and one green (Colour-Mismatch condition), or three grey and one red (Colour-
Match condition). The distractor arrays were presented anywhere between two frames before the target (Lag 2) and one frame after the target (Lag -1). The colour of each non-target letter (rendered in various fonts) could be either grey, blue, purple, or brown, chosen randomly for each letter. Right-hand panel mean proportion of correct target identifications in each distractor condition in Experiment 1 as a function of the temporal lag between the onset of the distractor array and the onset of the target. One lag = 98 ms. (ms milliseconds)
from all other conditions (all p’s \ .005). In addition, the Colour-Mismatch and the Grey conditions, while not differing from one another (p = .376), differed significantly from both the Colour-Match and the No-Distractor conditions (all p’s \ .005). The outcomes of these comparisons match nicely with the corresponding analyses reported by Folk et al. (2002, Experiment 2) and strongly suggests that the strength of attentional capture is not affected by whether the conditions are blocked, as in the present experiment, or mixed, as in the experiment of Folk et al. On the face of it, the present results uphold Folk et al.’s (2002) conclusion that a salient singleton does not produce attentional capture unless it shares the target’s defining property. However, the results for Lag 2 (Fig. 1) and the corresponding evidence in Fig. 3 of Folk et al. point to the possibility that stimulus-driven factors might indeed have contributed to the overall results. Specifically, the present results for Lag 2 (Fig. 1) show that performance in the Colour-Mismatch and the Grey conditions was significantly lower than in the No-Distractor condition, even though the observers were in featuresearch mode and the distractor array did not possess the target’s defining feature. On the other hand, a feature that the former two conditions shared with one another but not with the No-Distractor condition was that the distractor arrays appeared suddenly at screen locations that had been blank up to that point. Thus, the distractor stimulus was characterized by a unique sudden onset that might have
captured attention exogenously. If this were true, such stimulus-driven capture would also have been a factor in the Colour-Match condition in which the distractor array also had a unique sudden onset. This raises the possibility that performance in the Colour-Match condition might reflect the action of stimulus-driven (sudden onset) as well as goal-directed (colour-contingent) factors. We checked on this possibility in Experiment 2.
Experiment 2 Experiment 2 was designed to determine whether the unique sudden onset of the distractor array in Experiment 1 caused attention to be captured in a stimulus-driven manner even though the observers were in feature-search mode. This was done by comparing performance in two conditions. The Unique-onset condition was the same as the Colour-Match condition in Experiment 1. In the Repeatedonset condition, four grey # distractors were presented around every letter in the RSVP stream up to the frame in the RSVP stream that contained the coloured # which always occurred at the Lag-2 position. A third (control) condition without # distractors was added to permit an assessment of attentional capture. We expected that the distractor array in the Uniqueonset condition would be seen as having a unique sudden onset, much as in the Colour-Match condition in
123
12
Experiment 1. In contrast, the corresponding distractor array in the Repeated-onset condition would not be perceived as having a unique sudden onset because it was preceded by the onsets of the distractor arrays presented throughout the RSVP stream up to that point. A difference in the magnitude of attentional capture between the Unique-onset and the Repeated-onset conditions would then reveal the influence of stimulus-driven factors triggered by the unique sudden onset. In view of the results of Experiment 1, and to maximize the number of trials in the present within-subjects design, only Lag 2 was included in the experiment. In addition, a Colour-Mismatch condition was not included both because it yielded the same pattern of results as the Grey condition in Experiment 1, and because it was not directly relevant to the sudden-onset issue under investigation. Method Observers Twenty-four undergraduate students drawn from the same population as Experiment 1 participated in Experiment 2. None had participated in Experiment 1. Fig. 2 Left-hand panel schematic representation of events on any given trial in Experiment 2. Characters rendered in solid bold font were coloured red. Non-target letters, rendered in various fonts, could be either grey, green, blue, purple, or brown, chosen randomly for each letter. In the Unique-onset condition, the non-target letter presented two frames before the target (Lag 2) was surrounded by four #s, three grey and one red. The Repeated-onset condition was the same as the Unique-onset condition except that all letters presented before Lag 2 were surrounded by four grey #s. Right-hand panel mean proportion of correct identifications in each distractor condition in Experiment 2
123
Psychological Research (2012) 76:8–19
Apparatus and stimuli Apparatus and stimuli were the same as in Experiment 1. Design and procedures The design and procedures were the same as in Experiment 1, with the following exceptions. First, given that Target colour did not interact significantly with any other factor either in Experiment 1 or in the study of Folk et al. (2002), only red targets were used in Experiment 2. Second, the experiment comprised the three conditions illustrated in the left-hand panel of Fig. 2. The No-Distractor condition was the same as the corresponding condition in Experiment 1. The Unique-onset condition was the same as the ColourMatch condition in Experiment 1. The Repeated-onset condition was the same as the Unique-onset condition except that four grey # distractors surrounded every item in the RSVP sequence up to the frame in which one of the #s in the distractor array was coloured red. When present, the distractor array that contained the coloured # was displayed only in the Lag-2 position, namely, unlike Experiment 1, no coloured #s were presented at Lags -1, 0, and 1. The #s appeared and disappeared with the letter they surrounded.
Psychological Research (2012) 76:8–19
Results and discussion The mean proportion of correct target identifications is illustrated in the right-hand panel of Fig. 2. The data were entered in a within-subjects ANOVA with two factors: Order in which the three blocks were presented (all six permutations), and Distractor Condition (three conditions: No Distractor, Unique-onset, and Repeated-onset). Because neither the effect of Order, F (1,18) = 1.11, p = .390, MSE = .065 nor the interaction between Order and Distractor Condition, F (10,36) = 1.22, p = .312, MSE = .031 were significant, the data were combined across the six orders, and a new ANOVA was performed on the combined data. The analysis revealed significant effects of Distractor condition, F (2,46) = 19.03, p \ .001, MSE = .032. Planned pairwise comparisons revealed that target identification was less accurate in the Unique-onset condition than in either the No Distractor, t (23) = 5.33, p \ .001, or the Repeated-onset conditions, t (23) = 3.21, p = .004. Also, accuracy was lower in the Repeated-onset than in the No Distractor condition, t (23) = 3.67, p = .001. The outcome of critical importance for the objective of the present experiment is that accuracy was higher in the Repeated-onset than in the Unique-onset condition, indicating that attentional capture was more pronounced in the latter. This outcome is in line with the hypothesis that attention was captured by the unique onset of the distractor array in the Unique-onset condition, thus evidencing a stimulusdriven source of capture even though the observers were in feature-search mode. In contrast, the onset of the corresponding distractor array in the Repeated-onset condition was not unique because it was preceded by the onsets of other leading distractor arrays in the RSVP stream. At a first approximation, it can be concluded that capture was mediated mainly by goal-directed factors in the Repeated-onset condition but that it was mediated jointly by goal-directed and stimulus-driven factors in the Unique-onset condition. For this conclusion to be unequivocal, two additional issues must be considered. First, that the attentional capture evidenced in the Repeated-onset condition relative to the No-distractor condition may not have been mediated entirely by goal-directed factors. The red # in the Repeated-onset condition was a colour singleton which, as was pointedly noted by Lamy et al. (2004), might trigger an attentional shift that is stimulus-driven rather than goaldirected. We consider this option in Experiment 3. Second, to interpret the results obtained in the Repeated-onset condition in terms of attentional capture, it is necessary to consider an alternative interpretation in terms of forward masking. The option must be considered that, in addition to its intended function, the leading stream of grey #s in the Repeated-onset condition might have had the unintended
13
effect of acting as a forward mask for the colour-matched #. If so, masking would have reduced the visibility of the colour-matched # and, hence, its ability to capture attention. The weaker capture obtained in the Repeated-onset condition could then be ascribed to reduced visibility of the coloured # distractor due to masking rather than to diminished stimulus-driven capture due to the reduction in the uniqueness of the sudden onset. Although seemingly plausible, a masking account is ruled out by findings in the masking literature showing that forward masking becomes progressively weaker as the SOA between the leading mask and the trailing target is increased. The strength of forward masking is maximal at an SOA just beyond zero, is considerably weaker at an SOA of 50 ms, and never occurs at SOAs beyond about 100 ms, which is the SOA employed in the present work (Spencer & Shuntich, 1970). The masking option is further disconfirmed by findings in the temporal-integration literature. It is known that the mechanism underlying forward masking is the temporal integration of the leading mask with the trailing target ¨ g˘men, 2006). It is also (Breitmeyer, 1984; Breitmeyer & O known that temporal integration of two successive displays viewed in photopic conditions fails at SOAs beyond about 50 ms and never occurs at SOAs beyond about 100 ms (Di Lollo, 1980). These temporal contingencies have been formalized by Michaels and Turvey (1979). Given that the SOA between successive items in the present RSVP streams was 98 ms, the findings in both the masking and the temporal-integration literatures put the results of the present experiments beyond the temporal range of forward masking.
Experiment 3 Experiment 3 examined the option that the colour singleton in the Repeated-onset condition in Experiment 2 might have triggered at least some degree of stimulus-driven capture because it was a colour singleton. To this end, we modified the paradigm so as to create an RSVP stream in which the red # was no longer a colour singleton. This was done by displaying each of the four #s in a different colour, with the colours changing randomly in successive frames. There were four conditions, all ‘‘repeated-onset’’ as illustrated in the left-hand panel of Fig. 3. In the Heterogeneous Control condition each # was either grey, green, blue, purple, or brown, but never red which was the colour of the target letter. The Heterogeneous Red Distractor condition was the same as the Heterogeneous Control condition except that one of the #s in the Lag-2 frame was coloured red. In the Homogeneous Control condition, all #s were grey. The fourth condition, Homogeneous Red Distractor, was the same as the Homogeneous Control condition except that one # in the
123
14
Psychological Research (2012) 76:8–19
Fig. 3 Left-hand panel schematic representation of events on any given trial in Experiment 3. The target letter was coloured red (rendered in solid bold). Each of the letters in the RSVP stream was coloured either grey, green, brown, blue, or purple (rendered here in various fonts). In the two Heterogeneous # conditions, each of the #s up to Lag 2 had a different colour (either grey, green, blue, purple, or brown, rendered here in various line types). In the two Homogeneous # conditions all #s preceding Lag 2 were coloured grey. The two Heterogeneous conditions differed with respect to the #s in the Lag 2
frame: in the Control condition each # had a different colour as in the preceding frames; in the Red distractor condition one # was coloured red as the target and three #s were of various colours as in the preceding frames. The two Homogeneous conditions also differed with respect to the #s in the Lag 2 frame: in the Control condition each # was coloured grey as in the preceding frames; in the Red distractor condition one # was coloured red as the target and three #s were grey. Right-hand panel mean proportion of correct target identifications in each distractor condition in Experiment 3
Lag-2 frame was coloured red. The main reason for using the two control conditions rather than, say, a common allgrey # condition, was to enable valid estimates of goaldirected and stimulus-driven effects, as detailed below. For example, to obtain a valid estimate of goal-directed effects in the Heterogeneous Red Distractor condition, a control condition must be employed that is identical to the former in all respects (notably in respect to stimulus-driven factors) except for the presence of the red #. Comparison between the two Heterogeneous conditions provides an estimate of goal-directed capture on the assumption that both likely sources of stimulus-driven capture associated with the presentation of the red # (unique sudden onset, colour singleton) have been removed. Any differences in accuracy between the two conditions can then be ascribed to a goal-directed factor, namely, to the fact that the red # matched the target’s defining feature.
Comparison between the two Homogeneous conditions provides an estimate of the combined effects of a goaldirected factor (because of the colour match between the red # and the target letter) and a stimulus-driven factor (because the red # is a colour singleton). An estimate of the stimulus-driven component can be obtained by subtracting the amount of capture obtained in the Heterogeneous condition (Heterogeneous Control minus Heterogeneous Red Distractor) from the amount of capture obtained in the Homogeneous condition (Homogeneous Control minus Homogeneous Red Distractor).
123
Methods Observers Forty-eight undergraduate students drawn from the same population as the previous experiments participated in
Psychological Research (2012) 76:8–19
Experiment 3. None had participated in the earlier experiments. Apparatus, stimuli, and procedures Apparatus and stimuli were the same as in the previous experiments. The four conditions used in the present experiment have been described above and are illustrated in the left-hand panel of Fig. 3. The conditions were presented in separate blocks of trials, counterbalanced across observers. All other procedural details were the same as in the Repeated-onset condition of Experiment 2.
Results and discussion The mean proportion of correct target identifications is illustrated in the right-hand panel of Fig. 3. The data were entered in a mixed-design ANOVA with two within-subject factors: Red Distractor at two levels (Present or Absent), Homogeneity at two levels (Heterogeneous or Homogeneous), and one between-subjects factor: Order (all 24 permutations). Because neither the effect of Order nor any of the interactions involving Order were significant (all F’s \ 1.1), the data were combined across the 24 orders, and a new within-subject ANOVA was performed on the combined data. The analysis revealed significant effects of Red Distractor, F (1,47) = 34.08, p \ .001, MSE = .016, Homogeneity, F (1,47) = 23.48, p \ .001, MSE = .006, and a significant interaction effect, F (1,47) = 16.88, p \ .001, MSE = .007. Combined with the graphical evidence in Fig. 3, the significant interaction effect indicates that the degree of attentional capture was greater in the Homogeneous than in the Heterogeneous conditions. On the reasoning outlined in ‘‘Introduction’’ of the present experiment, the larger amount of capture in the Homogeneous condition reveals the effect of a stimulus-driven component, namely the fact that the red # was a colour singleton. This finding is consistent with the results reported by Lamy et al. (2004). Given that the results clearly implicate stimulus-driven effects in contingent capture, a question arises as to whether the present results contain any evidence at all of goaldirected capture. An affirmative answer is provided by the finding that accuracy in the Heterogeneous Red Distractor condition was significantly lower than in the Heterogeneous Control condition, t(47) = 3.28, p = .002. As noted above, the two conditions were equated on stimulus-driven factors, including the possible effect of offset transients which have been shown to influence attentional capture (Du & Abrams, 2006). That difference, therefore, must be ascribed to goal-directed factors.
15
What is still unclear, however, is the nature of the goaldirected mechanism that mediated the impairment. Thus far, we have considered only one goal-directed option, namely, that attention was captured by the red # because it had the same colour as the target. But another option must also be considered. In the present experiment, as in the previous experiment, the red # was presented invariably at Lag 2. As such, it may have acted as a temporal cue for the upcoming target that appeared reliably 196 ms later. On this option, attention might have been deployed to the red # distractor not because it shared the colour of the target but because it acted as a temporal cue for it. To be sure, attentional capture would be governed by goal-directed factors in either case; but the underlying mechanisms would differ. Distinguishing between the two options is theoretically important because predictions from the contingent involuntary orienting hypothesis would be disconfirmed if the capture obtained in the present experiment were ascribable to temporal cueing rather than to the # distractor sharing the target’s colour. These two options were examined in Experiment 4 by invalidating the effectiveness of the red # as a temporal cue for the target. This was done by presenting the red # unpredictably either at Lag 2, as in the present experiment, or at Lag 0 (simultaneously with the target) or at Lag -1 (directly after the target). If the goal-directed attentional capture in the present experiment was mediated by the red # acting as a cue for the upcoming target, no capture should be in evidence in Experiment 4 because the cue would no longer be valid. On the other hand, if capture occurred because the # shared the target’s defining characteristic, then capture should be as much in evidence in Experiment 4 as it was in the present experiment.
Experiment 4 Methods Observers Forty-eight undergraduate students drawn from the same population as the previous experiments participated in Experiment 4. None had participated in the earlier experiments. Apparatus, stimuli, and procedures Apparatus and stimuli were the same as in the previous experiments, with the following exceptions. There were two conditions: Heterogeneous Red Distractor, and Heterogeneous Control, as illustrated in the left-hand panel of Fig. 4. These were identical to the corresponding
123
16
Psychological Research (2012) 76:8–19
The mean proportion of correct target identifications is illustrated in the right-hand panel of Fig. 4. The data were entered in a mixed-design ANOVA with one betweensubjects factor: Order in which the two blocks were presented (First or Second), and two within-subject factors: Condition (Red Distractor or Control), and Lag (-1, 0, or 2). Because neither the effect of Order, F (1,46) = 1.48, p = .230, MSE = .127, nor any of the interactions involving Order were significant (all F’s \ 1), the data
were combined across the two orders, and a new ANOVA was performed on the combined data. The analysis revealed significant effects of Condition, F (1,47) = 5.38, p = .025, MSE = .005, Lag, F (2,94) = 5.26, p = .007, MSE = .007, and a significant interaction effect, F (2,94) = 9.13, p \ .001, MSE = .006. Planned pairwise comparisons revealed that the two conditions did not differ significantly from one another at either Lag 0, t(47) = 0.54, p = .591, or Lag -1, t(47) = 0.71, p = .484. The outcome of critical importance for the objective of the present experiment, however, is that target identification at Lag 2 was less accurate in the Heterogeneous Red Distractor condition than in the Control condition, t(47) = 3.66, p = .001, indicating that attentional capture was in evidence even when the red # could not be used as a valid temporal cue for the upcoming target. This outcome strongly suggests that the attentional capture obtained in the Heterogeneous Red Distractor condition in Experiment 3 occurred because the red # was of the same colour as the target, not because it acted as a temporal cue. Indeed, the magnitude of attentional capture was the same (6%) in the two experiments, further discounting the option that the red # might have acted as a
Fig. 4 Left-hand panel schematic representation of events on any given trial in Experiment 4. The target letter was coloured red (rendered in solid bold). Each of the letters in the RSVP stream was coloured either grey, green, brown, blue, or purple (rendered here in various fonts). Each of the #s in each frame had a different colour (either grey, green, blue, purple, or brown, rendered in various line types). In the Heterogeneous Red Distractor condition one of the four
#s was coloured red (rendered in solid bold) and was presented either at Lag 2 (two frames before the target) or at Lag 0 (in the same frame as the target) or at Lag -1 (directly after the target). In the Figure the red # is shown at Lag 2. Right-hand panel mean proportion of correct target identifications in the two conditions in Experiment 4 as a function of the temporal lag between the onset of the distractor array and the onset of the target. One lag = 98 ms. (ms milliseconds)
conditions in Experiment 3 except that the red # in the Heterogeneous Red Distractor condition was presented unpredictably either at Lag 2, or at Lag 0, or at Lag -1. The two conditions were presented in separate blocks of trials, counterbalanced across observers. Each of the three lags was presented 50 times in each block, randomly across trials. Thus, the factorial design consisted of two withinsubject factors: Condition (at two levels: Red Distractor and Control) and Lag (at three levels: 1, 3, and 7), and one between-subjects factor (at two levels: order in which the two blocks were presented).
Results and discussion
123
Psychological Research (2012) 76:8–19
cue. To buttress this conclusion, the data for Lag 2 of the present experiment were compared with the corresponding data from Experiment 3 (Heterogeneous Red Distractor and Heterogeneous Control conditions) in a mixed ANOVA with one between-subjects factor (Experiment: 3 or 4) and one within-subject factor (Condition: Red Distractor or Control). The analysis revealed a significant effect of Condition, F (1,94) = 23.85, p \ .001, MSE = .007, but neither the effect of Experiment nor the interaction effect were significant (both F’s \ 1).
General discussion The contingent involuntary orienting hypothesis of attentional capture holds that a salient singleton (e.g., a coloured #) captures attention provided that it matches the target’s defining characteristic (e.g., its colour; Folk et al., 2002). On this view, capture is said to be based entirely on goal-directed factors. But the evidence is not entirely unambiguous: in these studies the distractor that shared the target’s defining feature (a goal-directed factor) invariably had a unique sudden onset that might have captured attention in a stimulus-driven manner (Yantis & Jonides, 1984). The present work pursued this possibility by eliminating the uniqueness of the sudden onset of the salient distractor in a condition in which the salient distractor array was preceded by a stream of non-salient distractor arrays. The results obtained with this procedure showed that the unique sudden onset of a distractor (a stimulus-driven factor) captures attention even when the observer is in feature-search mode. The design used in some earlier demonstrations of goal-directed capture did not allow this stimulus-driven component to be factored out from the goal-directed component thus confounding the experimental outcome. Stimulus-driven attentional capture The proposition that attentional capture can be mediated by stimulus-driven factors has been supported in several studies (Dalton & Lavie, 2006; Maki & Mebane, 2006; Spalek, Falcon, & Di Lollo, 2006; Wee & Chua, 2004). That proposition has been questioned, however, on the grounds that the procedures in those studies permitted a singleton-detection mode to be adopted, thus allowing for an interpretation based on goal-directed factors alone (Folk, Leber, & Egeth, 2008). Folk et al.’s objection does not apply to the present work because the observers were never in singleton-detection mode. Given that the letters in the RSVP stream were of different colours, the target could not be identified unless the observers were set for its specific colour, i.e., unless
17
they were operating in feature-search mode. Given that a singleton-detection mode was ruled out, the results of Experiments 2 and 3 provide unambiguous evidence that attention can indeed be captured by stimulus-driven factors. The pattern of results in Figs. 2 and 3 strongly suggests that stimulus-driven factors can work in conjunction with goal-directed factors in mediating attentional capture within the ambit of a single experiment. If this is indeed the case, a question arises regarding the combinatorial rule: do the two sources of capture combine additively, overadditively, or underadditively? An answer to this question can be computed from three independent estimates of attentional capture: one obtained under conditions in which stimulus-driven and goal-directed factors are active concurrently, a second in which only stimulus-driven factors are at work, and a third in which only goal-directed factors are at work. The combinatorial rule can then be assessed by adding together the stimulusdriven and the goal-directed estimates and subtracting the sum from the estimate containing both sources [both - (stimulus-driven ? goal-directed)]. If the difference is equal to zero the two factors are additive, if it is positive they are overadditive, and if it is negative they are underadditive. A first approximation to these estimates can be gleaned from the results of Experiments 1 and 2. An estimate of stimulus-driven capture is available in Experiment 1 by subtracting the score in the Grey-Distractor condition from that in the No-Distractor condition (84 - 75% = 9%). An estimate of goal-directed capture is available in Experiment 2 by subtracting the score in the Repeated-onset condition from that in the No-Distractor condition (83 - 62% = 13%). An estimate of the conjoint effects of stimulusdriven and goal-directed factors is available in Experiment 1 by subtracting the score in the Colour-Match condition from that in the No-Distractor condition. A corresponding estimate is also available in Experiment 2 by subtracting the score in the Unique-onset condition from that in the No-Distractor condition. To obtain an estimate of the conjoint effects of stimulus-driven and goal-directed factors, we averaged the scores in the corresponding conditions in Experiments 1 and 2. The average of the two No-Distractor conditions was 83.5%, and the average of the Colour-Match (Experiment 1) and Unique-onset (Experiment 2) conditions was 59%. The subtraction yielded an estimated conjoint effect of 24.5% which is 2.5% higher than the sum of the two independent effects (9 ? 13% = 22%), perhaps hinting at some overadditivity. These estimates, however, must be regarded as tentative. An answer to the question of additivity must await the outcome of experiments designed explicitly for that purpose.
123
18
Temporal characteristics of stimulus-driven and goaldirected factors It has been suggested that, depending on the processing stage, attentional capture can be governed by either stimulus-driven or goal-directed factors. As noted above, Theeuwes and colleagues have proposed that visual selection and, therefore, attentional capture is governed exclusively by stimulus-driven factors during the first 150 ms of processing and by goal-directed factors thereafter (1994; Theeuwes, Atchley, & Kramer, 2000). On the face of it, this conjecture is not supported by the results of Experiment 1. When the distractor arrays were presented 98 ms before the target, neither the Grey nor the Colour-Mismatch conditions differed from the No-Distractor condition (Fig. 1, Lag 1). Namely, there was no evidence of stimulus-driven capture during the first 100 ms of processing. In contrast, target identification in the Colour-Match condition was significantly impaired at Lag 1, pointing to goal-directed effects occurring early in the processing sequence. In contrast, stimulus-driven factors came into evidence only when the distractor arrays preceded the target by 196 ms (Grey and Colour-Mismatch conditions, Fig. 1, Lag 2). This pattern of results is, if anything, the reverse of what might be expected on the basis of the conjecture that stimulus-driven factors precede goal-directed factors in the course of visual selection. The finding that in Experiment 1 the impairment in the Grey and Colour-Mismatch conditions occurred later than in the Colour-Match condition suggests that stimulus-driven effects may have longer onset latencies than goaldirected effects. This seems counterintuitive because stimulus-driven effects are commonly regarded as arising earlier in the chain of processing events. The earlier response to Colour-Match distractors, however, can be understood if it is assumed that the system responds more readily to stimuli that match an existing attentional set. Neurophysiological and brain-imaging evidence shows that such an attentional set can be implemented at very low levels in the visual system, notably Area V1 (Bahrami, Lavie, & Rees, 2007; Kastner & Ungerleider, 2000; Motter, 1993). Thus, even though both sets of distractor arrays had sudden onset, only the Colour-Match condition contained a # distractor that matched the system’s setting, thereby triggering a response earlier than a distractor in the other conditions. Implications for the attentional blink The outcome of the present work has direct implications for the phenomenon known as the attentional blink (AB) in which perception of the second of two targets is impaired if it is presented within about 500 ms after the first
123
Psychological Research (2012) 76:8–19
(Raymond, Shapiro, & Arnell, 1992). Theoretical interpretations have centred on the idea that the AB occurs when the second target’s access to a high-level processing stage is delayed while that stage is busy with the first target (Chun & Potter, 1995; Jolicœur & Dell’Acqua, 1998; Di Lollo, Kawahara, Ghorashi, & Enns, 2005). It is known that the AB can occur even if the display sequence contains only a single target, provided that it is preceded by a distractor that shares some of the target’s characteristics (e.g., Chun, 1997; Folk et al., 2002; Spalek et al., 2006). The present pattern of results is consistent with the view that an AB can be triggered by a salient temporally leading stimulus even if it does not share the target’s defining characteristic. Specifically, the identification of the target was impaired when attention was captured by the unique sudden onset of a task-irrelevant stimulus. In light of this result, the question that needs to be asked is how can such a task-irrelevant stimulus gain access to a high-level processing stage, given that the system is set to pass only stimuli that share the target’s defining characteristic. One plausible answer has been provided by Spalek et al. (2006) who proposed a hybrid input-filtering model in which goal-directed and stimulus-driven factors are subserved by different pathways, both converging on a highlevel processing stage. Access to one pathway is restricted to stimuli that match the goal-directed filter-setting. The other pathway is accessible to biologically relevant stimuli that, while not matching the goal-directed filter setting, are nevertheless salient by virtue of some other attribute such as sudden onset or gradient discontinuity. Because the higher processing stage is kept busy by a stimulus arriving from either pathway, an AB deficit can be triggered by both goal-directed and stimulus-driven factors. This account bears clear similarities to Wolfe’s (2006) Guided Search 4.0 model in which a similar dual-pathway architecture has been proposed.
Conclusions In the present work we decoupled the relative contributions of stimulus-driven and goal-directed factors to attentional capture. The outcomes of all four experiments showed consistently that either goal-directed or stimulus-driven or a combination of both factors can produce attentional capture. It can, therefore, be concluded that neither the contingent involuntary orienting hypothesis (Folk et al., 1992) nor the hypothesis that capture is invariably stimulus-driven (Theeuwes, 1994) can fully account for the present results. Acknowledgments This work was supported by a Canadian Foundation for Innovation New Opportunities Grant, by a grant from
Psychological Research (2012) 76:8–19 the British Columbia Knowledge Development Fund, and by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada to TMS. We thank Shannon Gaudry for expert assistance in data collection.
References Bacon, W. F., & Egeth, H. (1994). Overriding stimulus-driven attentional capture. Perception & Psychophysics, 55, 485–496. Bahrami, B., Lavie, N., & Rees, G. (2007). Attentional load modulates responses of human primary visual cortex to invisible stimuli. Current Biology, 17, 509–513. Breitmeyer, B. G. (1984). Visual masking: An integrative approach. New York: Oxford University Press. ¨ g˘men, H. (2006). Visual masking: Time slices Breitmeyer, B. G., & O through conscious and unconscious vision. New York: Oxford University Press. Cave, K. R., & Wolfe, J. M. (1990). Modeling the role of parallel processing in visual search. Cognitive Psychology, 22, 225–271. Chun, M. M. (1997). Temporal binding errors are redistributed in the attentional blink. Perception & Psychophysics, 59, 1191–1199. Chun, M. M., & Potter, M. C. (1995). A two-stage model for multiple target detection in rapid serial visual presentation. Journal of Experimental Psychology: Human Perception and Performance, 21, 109–127. Dalton, P., & Lavie, N. (2006). Temporal attentional capture: Effects of irrelevant singletons on rapid serial visual search. Psychonomic Bulletin & Review, 13, 881–885. Di Lollo, V. (1980). Temporal integration in visual memory. Journal of Experimental Psychology: General, 109, 75–97. Di Lollo, V., Kawahara, J., Ghorashi, S. M. S., & Enns, J. T. (2005). The attentional blink: resource depletion or temporary loss of control? Psychological Research/Psychologische Forschung, 69, 191–200. Du, F., & Abrams, R. A. (2006, November). Synergistic attentional effects of abrupt onsets and color singletons. Poster session presented at the 47th Annual Meeting of the Psychonomic Society, Houston, TX. Folk, C. L., & Remington, R. W. (1998). Selectivity in distraction by irrelevant featural singletons: Evidence for two forms of attentional capture. Journal of Experimental Psychology: Human Perception and Performance, 24, 847–858. Folk, C. L., Remington, R. W., & Johnston, J. C. (1992). Involuntary covert orienting is contingent on attentional control settings. Journal of Experimental Psychology: Human Perception and Performance, 18, 1030–1044. Folk, C. L., Leber, A., & Egeth, H. (2002). Made you blink! Contingent attentional capture produces a spatial blink. Perception & Psychophysics, 64, 741–753. Folk, C. L., Leber, A. B., & Egeth, H. E. (2008). Top-down control settings and the attentional blink: evidence for non-spatial contingent capture. Visual Cognition, 16, 616–642.
19 Jolicœur, P., & Dell’Acqua, R. (1998). The demonstration of shortterm consolidation. Cognitive Psychology, 36, 138–202. Kastner, S., & Ungerleider, L. G. (2000). Mechanisms of visual attention in the human cortex. Annual Review of Neuroscience, 23, 315–341. Lamy, D., Leber, A., & Egeth, H. E. (2004). Effects of task relevance and stimulus-driven salience in feature-search mode. Journal of Experimental Psychology: Human Perception and Performance, 30, 1019–1031. Maki, W. S., & Mebane, M. W. (2006). Attentional capture triggers an attentional blink. Psychonomic Bulletin & Review, 13, 125–131. Michaels, C. F., & Turvey, M. T. (1979). Central sources of visual masking: Indexing structures supporting seeing at a single, brief glance. Psychological Research, 41, 1–61. Motter, B. C. (1993). Focal attention produces spatially selective processing in visual cortical areas V1, V2, and V4 in the presence of competing stimuli. Journal of Neurophysiology, 70, 909–919. Raymond, J. E., Shapiro, K. L., & Arnell, K. M. (1992). Temporary suppression of visual processing in an RSVP task: An attentional blink? Journal of Experimental Psychology: Human Perception and Performance, 18, 849–860. Spalek, T. M., Falcon, L. J., & Di Lollo, V. (2006). Attentional blink and attentional capture: Endogenous versus exogenous control over paying attention to two important events in close succession. Perception & Psychophysics, 68, 674–684. Spencer, T. J., & Shuntich, R. (1970). Evidence for an interruption theory of backward masking. Journal of Experimental Psychology, 85, 198–203. Theeuwes, J. (1992). Perceptual selectivity for color and form. Perception & Psychophysics, 51, 599–606. Theeuwes, J. (1994). Stimulus-driven capture and attentional set: Selective search for color and visual abrupt onset. Journal of Experimental Psychology: Human Perception and Performance, 20, 799–806. Theeuwes, J., Atchley, P., & Kramer, A. F. (2000). On the time course of top-down and bottom-up control of visual attention. In S. Monsell & J. Driver (Eds.) Control of Cognitive Processes: Attention and Performance XVIII (pp. 105–124). Cambridge: MIT Press. Treisman, A., & Sato, S. (1990). Conjunction search revisited. Journal of Experimental Psychology: Human Perception and Performance, 16, 459–478. Wee, S., & Chua, F. K. (2004). Capturing attention when attention ‘‘blinks’’. Journal of Experimental Psychology: Human Perception and Performance, 30, 598–612. Wolfe, J. M. (2006). Guided Search 4.0: Current progress with a model of visual search. In W. Gray (Ed.), Integrated Models of Cognitive Systems (pp. 99–119). New York: Oxford. Yantis, S., & Jonides, J. (1984). Abrupt visual onsets and selective attention: Evidence from selective search. Journal of Experimental Psychology: Human Perception and Performance, 10, 601–621.
123