The paradigm of distractor-induced blindness has previously been used to track the transition from unconscious to conscious visual processing. In a variation of this paradigm used in this study, participants (n = 13) had to detect an orientation change of tilted bars (target) embedded in a dynamic random pattern; the onset of the target was signaled by the presentation of a color cue. Occasional orientation changes preceding the cue served as distractors and severely impaired the target's detection. ERPs showed that a frontal negativity was cumulatively activated by the distractors, and early sensory components were not affected. In a control condition, the target was defined by a coherent motion of the bars. Orientation changes preceding the motion target did not affect its detection, and the frontal suppression process was not observed. However, we obtained a significant reduction of the sensory components. The data support the notion that distractors that share the target's features trigger a cumulative inhibition process preventing the conscious representation of the inhibited features. Explorative source modeling suggests that this process originates in the pFC. A top–down modulation of sensory processing could not be observed.
Which neuronal processes determine whether a visual stimulus reaches perceptual awareness? It remains a question of debate which neural correlates and which cognitive mechanisms are critically involved (Fisch et al., 2009; Dehaene, Changeux, Naccache, Sackur, & Sergent, 2006; Rees, Kreiman, & Koch, 2002). To answer this question, various experimental procedures were developed inducing a short-term “blindness” in normal observers, such as visual masking (Breitmeyer, 1984) or attentional blink (Shapiro, 1994; for a review, see Kim & Blake, 2005).
Using electrophysiological and functional magnetic resonance neuroimaging, the state of visual awareness has been linked to activation at the level of sensory processing in a number of studies. Neural activation of primary visual cortex (V1) measured by means of ERPs was found to be linked to the detection of near-threshold stimuli (Pins & Ffytche, 2003), and hemodynamic activation in V1 also correlated with switches in the perceptual state induced by binocular rivalry (Tong & Engel, 2001). The latter was also obtained for visual areas in the ventral visual system (Polonsky, Blake, Braun, & Heeger, 2000; Leopold & Logothetis, 1996). Within the dorsal visual system, the processing of motion stimuli in area MT/V5 has been investigated, and single-unit recordings as well as ERPs demonstrated a close link between the observers' perceptual state and neural activity (Rodriguez & Valdes-Sosa, 2006; Salzman & Newsome, 1994).
However, it is questionable whether the activation in the visual cortex is sufficient to predict visual awareness. Neither in visual masking (Dehaene, Sergent, & Changeux, 2003) nor in studies on the attentional blink (Luck, Vogel, & Shapiro, 1996) was target detection associated with an increase of sensory activity. Therefore, it has been suggested that visual awareness rather relies on the long-range interaction of a distributed cortical network involving parietal and frontal brain regions. Indeed, fMRI studies on change blindness provided evidence that a fronto-parietal network appears to be involved (Beck, Rees, Frith, & Lavie, 2001). This higher-order network has been proposed to trigger a top–down attentional amplification of the sensory input (Dehaene et al., 2003, 2006), leading to a conscious representation of the stimulus. Further experimental evidence for a neuronal “global workspace” underlying conscious perception has been provided in studies on the attentional blink (Kessler et al., 2005; Sergent, Baillet, & Dehaene, 2005) and visual masking (Del Cul, Baillet, & Dehaene, 2007).
Although the above-mentioned approaches are helpful to identify brain regions critically involved in visual awareness, they do not specify the transition from unconscious to conscious processing because they primarily focus on the brain's state associated with visual awareness but neglect the preceding state. A number of studies using fMRI (Hesselmann, Kell, & Kleinschmidt, 2008; Boly et al., 2007) and EEG (Busch, Dubois, & VanRullen, 2009; Monto, Palva, Voipio, & Palva, 2008) have indicated, albeit on different time scales, that the conscious access to a visual stimulus critically depends on the activation state immediately preceding the presentation of the target. However, results from these studies rely on a post hoc analysis of prestimulus activity; in other words, the prestimulus state of activation was not controlled for in a rigorous experimental fashion.
To overcome these limitations, we have recently developed a paradigm in which the probability to detect a simple visual feature can be modulated by a simple experimental parameter. Importantly, this paradigm allows us to track the parameter's behavioral effects and, as will be explained below, the changes in brain activity preceding visual awareness. Originally, the behavioral effect evoked by our paradigm has been dubbed as “attention-induced motion blindness” (AMB; Sahraie, Milders, & Niedeggen, 2001). In the standard AMB paradigm, two spatially separate rapid serial visual presentation (RSVP) streams were presented. In a local sequence, the color of a fixation point changed at 10 Hz. This central area was surrounded by a random dot kinematogram whose dots followed a random walk (Scase, Braddick, & Raymond, 1996). The random global motion was interrupted by short episodes of coherent motion for 100 msec. The subject's task was to attend to the color “red” in the local stream and to detect a simultaneous coherent motion episode in the global stream (target motion). Thus, the color change in the local stream served as a cue to shift attention to the global stream. The behavioral results showed that the detection of the salient motion target can be significantly affected by the presence of “distractors” (Hesselmann, Allan, Sahraie, & Milders, 2009; Hesselmann, Niedeggen, Sahraie, & Milders, 2006), that is, coherent motion epochs before the cue onset that were irrelevant to the task had to be ignored.
Follow-up experiments from our laboratory further specified the effect of distractors in the AMB paradigm. We found that the probability to miss the target was related to the number of presented distractors: The more distractors were presented within the precue epoch, the higher was the probability to miss the target (Hesselmann et al., 2006). On the basis of these findings, we proposed that the distractors trigger an inhibitory mechanism that leads to a cumulative suppression of target feature processing. Using ERPs, we have been able to identify a neural correlate of this inhibitory mechanism. With increasing number of distractors, a frontal negativity (FN), peaking at about 250 msec after distractor onset, was found to be more expressed (Niedeggen, Hesselmann, Sahraie, Milders, & Blakemore, 2004). In a recent study on the attentional blink, a similar ERP component has been reported (Zhang, Zhou, & Martens, 2009). The process has been related to the activation of a category-specific negative attentional set. In the study by Zhang and colleagues, however, only a single distractor was presented so that a cumulative effect could not be examined.
In this study, we examined the processing of visual distractors using a variation of the original AMB paradigm. Random dots were substituted by a random visual array of diagonally tilted bars. In addition to the onset of a coherent motion of the bars, an orientation flip of all bars provided a second signal in the global stream (see Figure 1). Both features could serve as target and/or distractor. Note that the term “distractor-induced blindness” instead of “motion blindness” will be used in the following paragraphs, because the target's feature is no longer restricted to coherent motion but extends to changes of orientation. The use of a second visual feature allowed us to extend our previous research and further test the characteristics of the distractor-induced inhibitory process. Specifically, our experiment focused on three questions:
Is distractor-induced blindness feature-specific? In our previous models of the AMB, we did not specify whether the activation of the inhibitory process requires a match of the target's and distractor's features (Hesselmann et al., 2006; Sahraie et al., 2001). In priming experiments requiring a rapid change between the task settings, feature-specific enhancement and inhibition were obtained (Kiefer & Martens, 2010). We assumed that the inhibition process is also feature-specific and cannot be activated by deviant events, which do not share the targets' characteristics.
Is the cumulative frontal activation process linked to behavioral performance, that is, the reduction in target detection? Our experimental design allowed the comparison of responses to the different precue events while keeping the participants' task constant. We assume that a suppression process indexed by the FN should be exclusively triggered by distractors for which the detection of the upcoming target is reduced.
Can we observe differences in the early sensory processing of distractors? If the reduced access to the target stimulus is associated with a top–down regulation of sensory processing, one might assume that visually evoked responses to the distractors will be reduced. This reduction should be restricted to the processing of distractors that reduce the conscious access of the target and impair detection performance.
Sixteen subjects participated in the study. Three subjects were excluded because of difficulty to follow the fixation instructions and excessive EEG artifacts. The final sample consisted 13 participants (nine women) aged between 21 and 35 years. All participants had normal or corrected-to-normal visual acuity, had no history of neurological disorders, and were students of the Freie Universität Berlin and received course credit after giving informed consent.
Stimuli, Task, and Design
The visual stimuli were presented on an SVGA monitor using a VSG stimulus generator (VSG 2/5; Cambridge Research Systems Ltd., Kent, United Kingdom) and customized software written in C++ (Version 3.0, Borland, Austin, TX). The stimuli were presented at a viewing distance of 57 cm. The local stream consisted a 10-Hz color change of the central fixation point (0.5° in diameter). The used colors were of different luminance and easily separable. The fixation was centered in a gray circular patch (3.5° in diameter), which was surrounded by 150 randomly distributed white bars on a gray background (25° × 25°). Each bar was diagonally oriented and defined by three dots (0.18° in diameter). As shown in Figure 1, all bars were oriented in the same direction. The global stream consisted random walk noise, in which individual direction of motion was randomly assigned every 10 msec. In a fraction of trials, we embedded occasional changes of the bars' orientation. The onset of a change in orientation (diagonally left to right to diagonally right to left or vice versa) was a global coherent event, and all bars changed orientation simultaneously. The onset of such a “flip” event was synchronized with the changes in the color of the fixation. We will refer to those flip events as “distractors.” In each RSVP trial, only one distractor was predefined as a “probe” and provided a trigger event for the subsequent EEG segmentation. To prevent the superimposition of ERP responses to subsequent distractors, the “probe” was separated from the following orientation flip by 500 msec. The final probe occurred at least 500 msec before cue onset so that it was clearly separable from the target (Hesselmann et al., 2006).
Two blocks, each comprising 252 trials, involving a different target were presented. Each trial contained a cue that occurred randomly between 2400 and 4000 msec after trial onset, which signaled that a forthcoming target had to be detected. In one third of the trials, no targets were presented to minimize false alarm rates. In the remaining trials, targets appeared with short (0 msec) and long (400 msec) SOAs with respect to the cue onset. In 75% of the trials (n = 189), between 6 and 10 distractors were presented. For each of the two tasks, an equal number of probes (n = 63) from early (second distractor, approximately 2.4 sec before cue onset), medium (fourth or fifth distractor, approximately 1.7 sec before cue onset), and late (seventh or eighth distractor, approximately 1.0 sec before cue onset) positions were considered for ERP analysis. In 25% of the trials (n = 63), no distractors were presented. Here, EEG probes were recorded in the precue epoch to analyze the steady-state potentials evoked by color changes. The temporal position of EEG probes was matched to the position (early, medium, or late) of distractors, as described above.
In the “orientation task,” the cue was defined by the onset of a red fixation point, and the target consisted an orientation flip presented either simultaneously with the cue (short SOA) or after 400 msec (long SOA). Distractors before cue onset were defined as “congruent distractor” because they shared the target's feature. In the “motion task,” the target consisted a brief episode of coherent motion, that is, all bars moved in one of the four cardinal directions (upward, downward, leftward, or rightward) for 100 msec. The cue and the cue-target SOA were defined as in the orientation task. Distractors in the motion task were defined as “incongruent distractors” because a different visual feature defined the target. On the basis of behavioral findings from our previous experiments on the effect of feature specificity (Michael, Hesselmann, Kiefer, & Niedeggen, 2011), a “mixed” distractor block was not included in the design: We have shown that the expression of distractor-induced blindness was not affected if incongruent features were added to RSVP sequences determined by congruent distractors (Experiment 3).
Following each trial, participants were required to indicate whether they had detected the predefined target by pressing a response key. The sequence of the two blocks (orientation task and motion task) was balanced across participants, and the sequence of experimental conditions (target-present vs. target-absent conditions; long vs. short cue-target SOAs; early, medium, or late probe position) was randomized in each participant.
Behavioral data were analyzed separately for the task (motion vs. orientation), the presence of distractors, and the cue-target SOA (short vs. long) using a repeated-measures ANOVA. Degrees of freedom were corrected according to the Greenhouse–Geisser criterion (Greenhouse & Geisser, 1959).
An elastic cap with predefined electrode positions (EasyCap, Herrsching-Breitbrunn, Germany) was mounted on the participant's head. The 35 active Ag–AgCl electrodes were referenced to linked mastoids, with impedance kept below 10 kΩ. Additional electrodes attached at the outer canthi and the suborbital and supraorbital ridges of the right eye controlled for horizontal (hEOG) and vertical (vEOG) eye movements. Biosignals were recorded continuously with a 40-channnel NuAmps amplifier (Software Acquire, Neuroscan Labs, Neurosoft, Inc., El Paso, TX). Data were band-pass filtered on-line (0.1–200 Hz) and sampled at 500 Hz. Off-line EEG data were analyzed using the “Brain Vision Analyzer” (Version 1.05, Brain Products GmbH, Gilching, Germany). EEG was segmented according to the probe onset in each trial (−100 to 600 msec epoch length), filtered (0.3–30 Hz), and baseline corrected (−100 to 0 msec before probe onset). Single EEG sweeps containing muscular or ocular artifacts were excluded from analysis. The remaining sweeps were separately averaged according to the participant's task (orientation and motion target), the probe's temporal position (early, medium, and late), and the electrode position.
On the basis of the grand-averaged ERPs triggered by the probes, we determined three distinct time ranges (P1: 80–120 msec, N2: 140–180 msec, and FN: 250–450 msec; see also Figure 2). In each time region, mean amplitudes were computed for each participant separately for the experimental conditions and electrodes. Statistical effects were analyzed independently for each task running a two-way repeated-measures ANOVA (Probe Position × Electrodes). Degrees of freedom were corrected according to the Greenhouse–Geisser criterion (Greenhouse & Geisser, 1959).
Analysis of the experimental effects in the three time ranges was restricted to the electrodes in which the components were maximally expressed. For this reason, reference-independent ERP amplitudes were computed. According to the scalp distribution, analysis focused on two electrode clusters: The posterior cluster comprised the electrodes O1, Iz, O2, P3, Pz, and P4, and the anterior cluster comprised the electrodes FP1, FP2, AFz, F3, Fz, and F4.
EEG data analysis was restricted to distractor-evoked ERPs. The analysis of different distractor-induced effects on target-evoked ERPs would have been difficult because different visual features were used to define the targets. An in-depth analysis of the target-related ERPs can be found in our previous studies (Niedeggen, Hesselmann, Sahraie, & Milders, 2006; Niedeggen, Sahraie, Hesselmann, Milders, & Blakemore, 2002). Please also note that the experimental design did not allow us to run a separate analysis for hits and misses with respect to the target. This is primarily because of the fact that incongruent distractors did not significantly affect detection performance (see Results section below).
EEG Source Reconstruction
To estimate the cortical sources of selected ERP components, we used source localization modeling based on the freely available software package Brainstorm 3.1 (neuroimage.usc.edu/brainstorm/) by Sylvain Baillet and colleagues (Tadel, Baillet, Mosher, Pantazis, & Leahy, 2011). We applied a whitened and depth-weighted linear L2 minimum-norm estimates algorithm (Hamalainen & Ilmoniemi, 1994) to the grand-averaged data from 33 active electrodes using a default three-shell sphere head model. The applied algorithm assumes that, at each vertex of the cortex surface, there is only one dipole and that its orientation is normal to the cortex surface at this point (“constrained solution”). Note that minimum-norm estimates have a bias toward superficial currents and that deeper sources will be identified less accurately. Taken together, this analysis should be considered as explorative.
As can be seen in Table 1, the presence of distractors decreased the probability of target detection, and the impairment recovered with increasing cue-target SOA. In the orientation task, the ANOVA yielded a main effect of Congruent Distractors (F(1, 11) = 60.85, p < .001, η2 = 0.85) as well as a main effect of Cue-target SOA (F(1, 11) = 100.12, p < .001, η2 = 0.90). The significant interaction of both effects (F(1, 11) = 32.30, p < .001, η2 = 0.75) indicated that the SOA effect is significantly strongly expressed if distractors were presented (Δhit rate = 31.7%, t(11) = 10.06, p < .001) as compared with the absence of distractors (Δhit rate = 8.7%, t(11) = 3.43, p < .01). The latter factor also affected the false alarm rates; in case of distractor presence, false alarms occurred more frequently as compared with distractor absence (presence: M = 13.9, SD = 11.2; absence: M = 2.5, SD = 4.0; t(11) = 3.45, p < .01).
|Orientation||56.9 (13.5)||88.6 (5.6)||86.5 (8.9)||95.2 (5.3)|
|Motion||94.7 (4.6)||97.5 (3.9)||95.6 (7.2)||98.4 (5.5)|
|Orientation||56.9 (13.5)||88.6 (5.6)||86.5 (8.9)||95.2 (5.3)|
|Motion||94.7 (4.6)||97.5 (3.9)||95.6 (7.2)||98.4 (5.5)|
In the motion task, incongruent distractors did not affect target detection (F(1, 11) = 2.74, p = ns), and a positive effect of hit rate with increasing cue-target SOA was not obtained (F(1, 11) = 2.64, p = ns), independent of the presence or absence of distractors (interaction: F(1, 11) = 0.00, p = ns). False alarms occurred rarely, but only in the case of the presentation of distractors (presence: M = 1.8, SD = 1.1; absence: M = 0.8, SD = 0.8; t(11) = 2.73, p < .05).
The grand-averaged ERPs evoked by distractors presented in the precue epoch is shown in Figure 2. At posterior electrodes, a positive deflection of the EEG—further labeled as P1—was observed, peaking at about 100 msec. The topography of the P1 was focused at midline occipital and parietal leads. The ERP positivity is released by a transient negativity—further labeled as N2—with a peak maximum at 160 msec. As compared with the P1, the topography of the N2 is shifted toward the parietal leads but remains at the midline position. The topography of both components was adequately captured by the posterior electrode cluster. Because our previous ERP study allowed the identification of an FN, we aimed for this component in the current experiment, too. The grand-averaged potentials indicated a negativity (FN) at frontal leads extending from 250 to 450 msec following the probe onset. As depicted in Figure 4B, changes in the activation of FN were successfully captured by the anterior electrode cluster.
The EEG recorded in the current paradigm is not only affected by the presentation of the probes but also by the repetitive color changes in the local stream. In a previous study, we have already demonstrated that the 10-Hz activity triggered by changes in color is negligible as compared with the response to the individual probes (Niedeggen et al., 2004). Here, the same effect was observed. As shown in Figure 3A, cyclic local color changes did not evoke a steady-state activity, which masked the ERP response to probes. The visual components, primarily the N2, were easily separable from the color-driven steady-state activity.
Analysis of the P1 Effects
The P1 amplitude was not affected by the presentation of distractors neither in the orientation task (where distractors were congruent) nor in the motion task (where distractors were incongruent). For the orientation task, mean amplitudes remain stable for distractors presented at early, medium, and late positions in the RSVP stream (F(2, 24) = 0.33, p = ns). Within the posterior electrode cluster, we obtained no evidence for a local effect of Probe Position (Probe Position × Electrodes: F(10, 120) = 0.49, p = ns). In the motion task, P1 was marginally more expressed with cumulative presentation of incongruent distractors, but the effect did not reach significance (F(2, 24) = 1.92, p = ns).
Analysis of the N2 Effects
As shown in Figure 4A, the N2 amplitude was reduced in the orientation task following the first presentation of congruent distractors. However, the ANOVA did not confirm a significant decrease in amplitude (probe position: F(2, 24) = 2.38, p = ns; Probe Position × Electrodes: F(10, 120) = 1.83, p = ns). In the motion task, the cumulative effect of incongruent distractors on the N2 amplitude was more pronounced: The ANOVA indicated a significant decrease of amplitude (probe position: F(2, 24) = 4.135, p < .05, η2 = 0.26). The test of linear trend revealed that N2 amplitudes were reduced with increasing number of deviants following a linear trend (F(1, 12) = 8.94, p < .05, η2 = 0.43). To determine the scalp distribution of the effect, the difference waves between the ERP responses to late and early probe presentations were computed. The scalp distribution of the experimental effect confirms that the posterior N2 component was reduced in amplitude (see Figure 4A).
Analysis of the FN Effects
In the orientation task, the first presentation of a congruent distractor evoked a positive-going wave at frontal electrodes. With increasing number of distractor presentations, the amplitude in this time range became increasingly more negative (see Figure 4B). This effect was significant (probe position: F(2, 24) = 10.54, p < .001, η2 = 0.47). The amplitude function followed a significant linear trend (F(1, 12) = 33.40, p < .001, η2 = 0.74) indicating that the distractor effect was not restricted to a negative shift from early to medium probe positions. The topography of the difference waves (ERP[late distractor] − ERP[early distractor]) confirmed that the effect was pronounced at midline frontal electrodes. The posterior positivity depicted in Figure 4B was not systematically affected by the number of distractors. In the motion task, an early incongruent distractor also triggered a positive ERP shift, but no negative shift at later probe positions was expressed. Correspondingly, the ANOVA did not yield a significant effect of the probe position (probe position: F(2, 24) = 1.02, p = ns; Electrode × Probe Position: F(10, 120) = 1.38, p = ns).
Source Reconstruction of the FN
We applied source localization modeling to identify the cortical sources of the FN using a minimum-norm estimates algorithm. As a first “proof of principle” step, we aimed at identifying the cortical sources of the N2 component, evoked by the onset of the early distractor in our paradigm. Previous source analysis showed that the motion-evoked N2 originates in or around the middle temporal area (Probst, Plendl, Paulus, Wist, & Scherg, 1993). Figure 5A shows that the estimated distribution of sources is in good agreement with these earlier findings. In the second step, we applied the same algorithm to estimate the cortical sources of the FN component. We based this analysis on the differential ERP activity between the late and early distractors. Figure 5B shows the resulting distribution of sources, located primarily in prefrontal cortical areas, including anterior parts of the left and right middle frontal gyrus.
ERP Correlates of Target Processing
The effect of distractors on the sensory processing of the target has already been analyzed in a previous article (Niedeggen et al., 2002). Here, we focus on the differences between the “distractor-absent” and “distractor-present” conditions, which allows us to control for adaptation effects. In line with our previous behavioral (Sahraie et al., 2001) and electrophysiological (Niedeggen et al., 2002) findings, we found no evidence that the repeated presentation of distractors affects the sensory processing of the target. As shown in Figure 3B, the lack of distractors in the precue epoch did not enhance the N2 response to the orientation target. Analysis of the amplitudes obtained for the short-cue target SOA did not reveal a significant difference at the posterior electrode cluster (distractor-present vs. distractor-absent conditions: F(1, 12) = 0.121, p > .5, η2 = 0.01).
In this study, our aim was to probe the brain mechanisms underlying the distractor-induced blindness effect. Our behavioral results showed that subjects' detection performance was impaired at short cue-target SOAs if task-irrelevant distractors shared the target's feature, that is, congruent distractors were presented. This significant behavioral impairment was associated with the generation of a distinctive FN in the distractor-evoked ERPs. If distractors did not share the target's feature (i.e., incongruent distractors), the detection performance was not impaired neither at short nor at long cue-target SOAs. Incongruent distractors did not trigger an FN, but sensory processing of distractors appeared to be gradually reduced. In the following paragraphs, each of these results will be discussed with respect to our research questions.
Question 1: Is Distractor-induced Blindness Feature-specific?
Our behavioral data provide clear evidence that incongruent distractors (here, motion coherence) are not sufficient to trigger an inhibitory process leading to a reduction in target processing. Distractor-induced blindness can obviously not be explained in terms of a bottom–up process, assuming that an automatic attentional process is activated by deviants in the RSVP stream, which interferes with the processing of the upcoming target (Maki & Mebane, 2006; Olivers & Watson, 2006; Folk, Remington, & Johnston, 1992). The feature specificity of the effect is rather in line with the idea that a top–down mechanism is involved. Following Belopolsky, Schreij, and Theeuwes (2010), one might assume that the presentation of congruent distractors forces the top–down system to disengage attentional resources repeatedly. This may lead to a specific inhibition of the visual feature.
A similar mechanism has been proposed to affect target detection in the attentional blink. In the attentional blink paradigm, stimuli are presented in one RSVP stream, and a predefined primary target (T1) has to be detected (Raymond, Shapiro, & Arnell, 1992). The detection of an upcoming second target (T2) critically depends on its temporal distance to T1 (Shapiro, 1994). Although most models on the attentional blink focus on the interaction of T1 and T2 processing (Martens & Wyble, 2010), distractor-like effects were recently identified to play an important role in the attentional blink (Martens & Valchev, 2009; Dux & Marois, 2008). If a visual event sharing the perceptual or semantic features of T2 was presented in the pre-T1 epoch, the hit rate for T2 was significantly decreased. It has been suggested that a category-specific negative attentional set is established by the distractors, which delays the attentional allocation to T2 (Zhang et al., 2009). On the basis of our findings, we assume that a comparable mechanism might be triggered by the presentation of congruent distractors in our study.
Distractor effects in RSVP tasks, as reported by Zhang and colleagues and in our study, appear to be in good agreement with the notion of “contingent attentional capture” (Folk, Remington, & Wright, 1994; Folk et al., 1992). According to this view, as opposed to purely stimulus-driven attentional capture (Theeuwes, 1994), capture is contingent on top–down control settings. Therefore, as in our paradigm, the selection of irrelevant distractors and the subsequent triggering of suppression mechanisms should depend on whether the distractors share the defining property of the target. It is important to note, however, that contingent attentional capture does not imply a cumulative effect as obtained in our study. Moreover, a spatial capture process (Folk, Leber, & Egeth, 2002) appears to rely on a shift in spatial attention, whereas a comparable process cannot be observed in distractor-induced blindness (Hesselmann et al., 2009).
Question 2: Is the Cumulative Frontal Activation Process Linked to the Reduction in Target Detection?
In our previous ERP study, a similar FN was elicited by motion distractors (Niedeggen et al., 2004). As compared with the FN obtained in our current study, it was characterized by an earlier onset and was terminated by a subsequent positivity. The topography, however, was highly similar, as well as the linear increase of amplitude with increasing number of distractors. Therefore, we assume that a similar mechanism was activated by congruent distractors in both experiments.
In our actual study, the frontal ERP effect is closely linked to the behavioral effects. In contrast to our previous experiments on motion blindness (Niedeggen et al., 2004), the onset of the cue required the shift of attention from the local to the global stream in both experimental conditions. However, the incongruent orientation distractors did elicit neither an FN nor a reduction of detection performance. Because both effects were observed for congruent orientation distractors, our initial assumption was confirmed: The activation of the FN is closely associated with a reduced probability to get conscious access to the upcoming target. In other words, the distractor-induced blindness paradigm allows us to influence the perceptual state of the observer, although avoiding that the prerequisites of visual awareness have to be explored by a post hoc analysis of prestimulus states (Busch et al., 2009; Hesselmann et al., 2008).
ERP components triggered by deviant or distracting events have also been identified in previous studies. A similar frontal ERP topography, for example, has been obtained for the “processing negativity” (Karayanidis & Michie, 1996) or as part of the no-go response (Kopp, Mattler, Goertz, & Rist, 1996). In contrast to these components, however, the distractor-evoked FN is not related to an enhancement in attention or to the suppression of a motor response. As shown in Figure 3B, the ERP responses to the early congruent distractor are characterized by a positive shift that resembles the characteristics of the P3a (Polich & Comerchero, 2003). A corresponding reduction of amplitude might therefore indicate that the activation of a frontal attention system is decreased (Nieuwenhuis, Aston-Jones, & Cohen, 2005). Although this approach is in agreement with the cumulative disengagement of attention triggered by distractors (Belopolsky et al., 2010), we assume that the FN is a distinct process. First, a P3a-like effect has not been obtained in our previous experiment (Niedeggen et al., 2004). Second, the ERP related to the processing of late distractors is significantly negative-going as compared with baseline (F(1, 12) = 8.07, p < .05, η2 = 0.41). Finally, a “novelty” P3a is known to respond to perceptually novel distractors that differ from the target (Polich & Comerchero, 2003; Knight & Scabini, 1998; Knight, 1984). Accordingly, the ERP component is more likely to be triggered by incongruent distractors in the motion task, which is obviously not the case (see Figure 4B). P3a-like responses to nonnovel repeated distractor stimuli (no-go P300) are characterized by a centro-parietal topography (Falkenstein, Hoormann, & Hohnsbein, 1999; Katayama & Polich, 1998). In contrast, the topography of the FN is clearly focused at frontal leads (see Figure 4B).
Evidence in favor of a unique distractor-evoked ERP effect was recently provided in a study on the attentional blink (Zhang et al., 2009). In the study by Zhang and colleagues, the presentation of a single distractor immediately preceding T1 evoked a frontally located negativity peaking at about 250 msec. According to the authors, the ERP indicates the activation of a negative attentional set that prevents the response to the distractor. Because the set continues to function over time, the processing of the upcoming T2 is affected, too. Despite of the similar experimental approaches and results, we doubt whether the ERP negativity observed by Zhang and colleagues is comparable with the FN obtained in our study: On the basis of our finding of a cumulative increase of the FN, we assume that the activation of a single distractor is probably not sufficient to activate the negative attentional set. Furthermore, the presentation of a single distractor, even in temporal proximity to the cue, was not sufficient to induce distractor-induced blindness (Hesselmann et al., 2006). We conclude that the two tasks, attentional blink and distractor-induced blindness, differ with respect to the attentional demands and that distractor effects are therefore differently expressed.
We tentatively suggest that the FN could be linked to a dynamic filtering process originating in pFC and affecting only later, that is, postsensory processing stages (Shimamura, 2000; Chao & Knight, 1995). The activation of the process shares the prerequisites of contingent attentional capture—as far as a match between the features of the distractor and the target is required. In line with our behavioral findings (Michael et al., 2011), the pattern of neural activity is independent of the visual feature defining the distractor. Similar ERP results have been obtained for orientation changes and for motion episodes (Niedeggen et al., 2004). According to our explorative source modeling, the origin of the FN triggered by distractors can be found in the pFC. As we have shown, this filtering process dynamically builds up in relation to distracting and to-be-ignored items on an item-by-item basis.
Question 3: Can We Observe Differences in the Sensory Processing of Distractors?
The activation of the FN is probably part of a higher-order neural network, which enables the transition from unconscious to conscious visual processing. As stated before, it has been assumed that the activation of such a network also affects the sensory processing of stimuli via top–down control mechanism (Del Cul et al., 2007; Kessler et al., 2005; Sergent et al., 2005).
In our paradigm, sensory processing of the orientation flips is reflected in two ERP components with a posterior topography, the P1 and the N2. This ERP complex is usually triggered by pattern reversal stimulation or by motion onset (Heinrich & Bach, 2001; Kubova, Kuba, Spekreijse, & Blakemore, 1995). Whereas the primary visual cortex primarily contributes to the expression of the P1, the N2 is closely linked to the extrastriate activity in the motion processing system (Barnikol et al., 2006). The P1 component can be assumed to reflect the processing of local luminance changes associated with the orientation flip. The N2 component very likely reflects the motion signal included in an orientation flip. Following our source modeling approach, the N2 is linked to neural activation in the middle temporal area. The same area has been previously associated with the processing of motion onset (Probst et al., 1993).
Although participants are instructed to ignore distractors in the precue epoch, different neural processes were obtained for the processing of incongruent and congruent distractors. With respect to the N2, the consecutive presentation of incongruent distractors leads to a significant amplitude reduction. In contrast, the N2 remained stable if congruent distractors were presented repeatedly.
The interaction of the N2 amplitudes in the different experimental conditions as a function of probe position (see Figure 4A) is probably because of different neural effects: The gradual attenuation of the N2 observed for incongruent distractors can be attributed to an unspecific adaptation process (Bach & Ullrich, 1994) or to a more-active top–down suppression of irrelevant sensory events (Kastner & Ungerleider, 2001). In case of congruent distractors (orientation task), neither an adaptation process (see also Figure 3B) nor a top–down suppression is apparently activated. The results confirm that task instruction determines the activation of top–down suppression or enhancement and affects the gain setting in sensory processing (Zanto & Gazzaley, 2009; Gazzaley, Cooney, McEvoy, Knight, & D'Esposito, 2005). Moreover, the result provides further evidence that the behavioral effect of distractor-induced blindness is not associated with a top–down suppression of the sensory signal (Niedeggen et al., 2002).
In summary, our findings show that, in the distractor-induced blindness paradigm, access to conscious awareness critically depends on a frontal gating system. The state of the gating system is directly modulated by events that share the target's features but are irrelevant at the time of presentation. Although the conscious representation of these events appears to be gradually prevented, sensory processing remains unaffected.
We thank three anonymous reviewers for their helpful and constructive comments. M. N. and G. H. were supported by the German Research Foundation (grants NI 513/8-1 and HE 6244/1-1, respectively).
Reprint requests should be sent to Michael Niedeggen, Free University Berlin, Habelschwerdter Allee 45, 14195 Berlin, Germany, or via e-mail: email@example.com.