Cognitive Brain Research 13 (2002) 241–247 www.elsevier.com / locate / bres
Research report
Is experimental motion blindness due to sensory suppression? An ERP approach a, b a b Michael Niedeggen *, Arash Sahraie , Guido Hesselmann , Maarten Milders , c Colin Blakemore a
Institute of Experimental Psychology II, Heinrich-Heine-University, D-40225 Duesseldorf, Germany b Vision Research Laboratories, Department of Psychology, University of Aberdeen, Aberdeen, UK c University Laboratory of Physiology, University of Oxford, Oxford, UK Accepted 8 November 2001
Abstract Recent psychophysical studies have revealed attentional modulation of visual motion perception and interest now focuses on the locus of this interaction. Using event-related brain potentials (ERPs) we examined whether transient motion blindness evoked in a dual task [Vision Res. 41 (2001) 1613–1617] is related to a selection process occurring at the stage of sensory processing or at a higher level. In our paradigm, a particular change of colour of the fixation point cued the subject to detect a brief episode of coherent random dot motion embedded in a succession of episodes of incoherent motion. Detection of the coherent motion was significantly impaired when it occurred simultaneously with the colour cue, and recovered over the subsequent 300 ms. This functional relationship was reflected in the amplitude of a sensory, motion-evoked component (N200), and in a late positive complex (P300). However, a direct comparison of ERPs produced by stimuli that were detected or missed revealed differences only in the P300 component. These results indicate that attenuation of sensory motion processing does not account for this transient, attention-induced deficit in visual motion perception. 2002 Elsevier Science B.V. All rights reserved. Theme: Sensory systems Topic: Visual cortex: extrastriate Keywords: Motion perception; Attention; Event-related brain potential; Experimental blindness
1. Introduction According to the traditional view, a number of visual features, such as orientation, colour, and size difference, are thought to be processed pre-attentively, i.e., without the aid of attention. Since a moving target can be easily identified among stationary distractors [6], motion has been added to this group. However, recent psychophysical studies have questioned whether all kinds of motion stimuli are analysed in parallel, pre-attentively. In the case of isoluminant motion [11], or long-range motion [8], the search for dynamic targets has been found to depend on *Corresponding author. Tel.: 149-211-8112-011; fax: 149-211-8114522. E-mail address:
[email protected] (M. Niedeggen).
focused attention. Comparable effects were confirmed in studies using higher-order motion stimuli, such as textureor stereo-defined motion ([4,10], for a review see Ref. [17]). A recent study by Valdes-Sosa et al. (2000) extended these findings from space- to time-based attention, demonstrating a temporary inability to shift perception between two surfaces defined by transparent motion [21]. These findings lead to two questions: (1) Does attention also affect the processing of luminance-based, first-order motion? (2) Is the limitation measured behaviourally due to suppression in the sensory cortex, or is it related to an inhibition at higher-order processing stages? The first evidence for an interaction between attentional mechanisms and those involved in processing of motion stimuli came from adaptation studies [1,5], which indicated that the strength of the motion after-effect was modulated
0926-6410 / 02 / $ – see front matter 2002 Elsevier Science B.V. All rights reserved. PII: S0926-6410( 01 )00122-7
242
M. Niedeggen et al. / Cognitive Brain Research 13 (2002) 241 – 247
by the focus of attention. We extended these findings using rapid serial visual presentation (RSVP) of dynamic random dots [18]. We demonstrated that the detection of a brief episode of coherent motion (the ‘target’) within a stream of incoherent motion is impaired if it occurs within 200 ms after presentation, at the fixation point, of a coloured cue, signalling attention to the motion. Although the strength of this ‘transient motion blindness’ weakens with the temporal separation between the cue and the target, and therefore relates to the attentional shift, the phenomenon does not occur without the occasional presentation of similar coherent motion episodes (‘distractors’) preceding the cue. We suggested that the transient motion blindness results not simply from the time take to shift attention from cue to motion, but from a delay in releasing suppression of responses to the coherent motion distractors. This effect of attention on first-order motion perception seems to be compatible with evidence for the modulation of motion detection at the level of early sensory analysis. Neuroimaging studies have suggested that the neuronal processing of various motion signals, including first-order (i.e., luminance-defined) motion, may be modulated by attention in relevant early visual cortical areas [3,15]. Some psychophysical experiments have also pointed towards such sensory modulation (for a review see Ref. [17]). And a number of event-related brain potential (ERP) studies [20,23] reported a significant reduction in the amplitude of the motion evoked N200 component (sometimes referred to as N1, N170, or N2) if attention was directed to invalid locations or objects. Considering the number of findings supporting an ‘early’ or ‘sensory’ locus of attentional selection, one might expect that the transient ‘motion blindness’ observed in our paradigm [18] would also be associated with suppression at a sensory level. On the other hand, the effect occurs in a time-based attentional selection task (RSVP). RSVPs are known to elicit the ‘attentional blink’ [19], which ERP studies indicate operates at a level of information processing following the primary sensory analysis. Early visual processing, and even semantic categorisation, remain unaffected by the attentional blink [22]. In our psychophysiological approach we therefore asked whether early ERP components evoked by the onset of a relevant coherent motion will reflect the hypothesized release from attentional suppression. A second analysis focused on the relationship between the ERP components and the subject’s perceptual state, i.e., whether they actually detected the coherent target motion.
2. Methods
2.1. Subjects and stimuli EEG recording was conducted at the University of
¨ Duesseldorf. Data were obtained in 17 normal naıve subjects aged between 21 and 37 (10 female, 7 male). Four subjects had to be discarded because of eye movement, or psychophysical performance that left an insufficient number of trials for ERP analysis. All subjects had normal or corrected-to-normal visual acuity, and had no history of neurological disorders. The stimulus display, created on an sVGA monitor using a VSG stimulus generator (Cambridge Research Systems), contained a ‘global stream’ and a ‘local stream’. The global stream, which occupied an annulus (outer diameter 24.58; inner diameter 6.08), consisted of 157 randomly distributed white dots (0.330.38; 80 cd m 22 ), moving at 78 s 21 , on a light grey background (19.2 cd m 22 ). The local stream consisted of changes, every 100 ms, of the colour (green, blue or red) or brightness (five grey levels) of a central fixation point (0.58 in diameter). Between the fixation point and the dot array, the screen was uniformly grey (31 cd m 22 ). Within each trial, lasting 4 s, the dots were constantly in motion. During each successive fixation colour, lasting 100 ms, the motion of the dots was either random or coherent In case of the coherent motion, all dots moved correlated for 100 ms in a randomly selected horizontal or vertical direction. Random motion was defined by a displacement of each dot in a direction that was randomly assigned (a random walk) every 10 ms (total of 10 changes of direction in each 100 ms interval).
2.2. Task and design The subjects were instructed to fixate on the fixation point for the duration of each trial, and to watch attentively for the single appearance of a red fixation (cue), which occurred at random between 1.5 and 3 s after the start of the local RSVP (Fig. 1). The red cue was the signal to shift attention to the global stream (the moving dots) and to watch for an episode of coherent motion (the target). Coherent motion episodes prior to the cue had to be ignored. To reduce selection errors, the interval between the last irrelevant coherent motion episode and the red fixation cue was at least 400 ms and no further episode of coherent motion occurred after the target motion. In half the trials the colour cue was not accompanied or followed by a coherent motion target (catch trials). In the other trials the coherent target motion occurred with a randomly chosen stimulus onset asynchrony (SOA) either simultaneously with the red cue (SOA 0 ms), or during one of two 100 ms epochs thereafter (SOA 100 ms, and SOA 300 ms). Following each trial, subjects were required to signal whether they detected a coherent motion target and to report (if necessary by guessing) if the target motion was in vertical or horizontal direction (two-alternativeforced-choice). No further coherent motion episode occurred after the target motion. Each subject received six blocks, each containing 120 trials. Sixty trials contained coherent target motion, 20 for each of the three SOAs. The
M. Niedeggen et al. / Cognitive Brain Research 13 (2002) 241 – 247
243
Fig. 1. A schematic diagram of the stimulus configuration. White arrows indicate the motion direction (random or coherent) of the dots in the global RSVP, which was defined in 100 ms episodes. Episodes of coherent motion were synchronized with changes of the colour or grey level of the fixation point (local stream). Attention had to be directed to the global RSVP when a red fixation (cue) was presented, and the subject’s task was to detect an episode of coherent motion (target). When present, the onset of this target motion was simultaneous with, 100 ms or 300 ms after the cue. Episodes of coherent motion occurring prior to the cue (‘distractors’) had to be ignored.
sequence of conditions (target present or absent; SOAs 0, 100, or 300 ms) was randomised in each block.
2.3. EEG recording and data analysis Electrodes were attached to the scalp according to the 10-20-system at frontal (Fz), central (Cz), temporal (T5, T6), parietal (P3, Pz, P4), and occipital (O1, O2) positions, and referenced to linked mastoids. Each recording epoch, triggered 200 ms prior to the onset of the red cue, lasted for 1500 ms. Data were sampled at 250 Hz, and band-pass filtered online (0.3–40 Hz). Single EEG sweeps containing muscular or occular (vEOG, hEOG) artefacts were excluded from analysis. Remaining sweeps were separately averaged according to the target presence (catch vs. target trials), the temporal separation between cue and target (SOA 0, 100 and 300 ms), and the subject’s performance (target trials: hits and misses; catch trials: correct rejections). The number of false alarms was insufficient to allow ERP data analysis for that condition. Behavioural results (correct detection of target motion at different SOAs) and ERP data (mean amplitude in the time ranges 160–240 and 400–480 ms, respectively, after the cue) were analysed using a repeated measure analysis of variance. In addition to the factor ‘SOA’, ERP analysis
considered the factors ‘spatial position’ (parietal, temporal, occipital) and ‘laterality’ (left vs. right hemisphere). The midline leads (Fz, Cz, Pz) were tested in a separate analysis. F-values were Greenhouse–Geisser corrected if indicated by tests for sphericity.
3. Results Behavioural data for 13 subjects are shown in Table 1. The subjects’ criterion in the motion detection task was quite conservative: false alarm rate, that is reporting coherent motion in the catch trials, was very low (in 3.7% of the catch trials). In accordance with our previous psychophysical study [18], correct detection of the target motion was significantly reduced at SOA 0 ms. Here, the coherent motion was detected in only 50% of the trials. Increasing the SOA resulted in a significant improvement of detection performance [F(2,24)559.01, P,0.001]. Pairwise comparison of the single temporal delays indicated significant differences between SOA 0 and 100 ms, as well as between SOA 100 and 300 ms (P,0.001, in both cases). A qualitatively similar relationship was obtained for direction discrimination (of detected targets), which was 70.5% at SOA 0 ms (chance level 50%), and then
M. Niedeggen et al. / Cognitive Brain Research 13 (2002) 241 – 247
244
Table 1 Behavioural performance in relation to the cue-target stimulus onset asynchrony (SOA) SOA (ms)
F(2,24)
P
0
100
300
Detection
52.3% (20.8)
72.2% (15.8)
93.2% (7.7)
59.01
,0.001
Direction
70.5% (15.3)
84.0% (10.7)
95.0% (6.4)
32.43
,0.001
False alarms
3.67% (1.54)
Means and standard deviations (in parentheses) for the correct detection of the target motion are given, as well as for correct direction discrimination (horizontal vs. vertical). False alarms refer to erroneous motion detection in case of catch trials (cue only). Significance of the effect of SOA is indicated in the right column (Greenhouse–Geisser corrected).
recovered to 95% correct discrimination at lag 300 ms [F(2,24)532.43, P,0.001]. Pairwise comparison of the single SOAs confirmed significant differences between the single temporal delays (0 ms vs. 100 ms: P,0.001; 100 ms vs. 300 ms: P,0.005). For both correct detection and discrimination, analysis of variance revealed a significant linear trend with increasing SOA [detection: F(1,12)5 67.18, P,0.001; discrimination: F(1,12)535.05, P, 0.001]. No evidence was found for covert or implicit ability to discriminate direction in the absence of the explicit perception of coherent motion detection (Table 2). For SOA 0 ms, a correct motion detection did enable subjects to discriminate motion direction (92.7% correct), with this hit rate clearly exceeding the chance level of 50% [t(12)5 19.713, P,0.001]. In cases that the target motion was missed, the proportion of correct direction discrimination (53.1%) did not differ significantly from chance level [t(12)51.250, P50.235]. The ERPs contained signals relating to the processing of the coloured fixation cue, as well as the motion target (when it occurred). The two processes affected the ERPs differentially, particularly when the two events were temporally separated. In line with previous experiments using the RSVP technique [16,22], we extracted the ERP components related to the processing of the coherent motion targets by subtracting the averaged ERP for presentation of the cue alone from ERPs generated by cue plus target. Fig. 2 exemplifies this subtraction procedure Table 2 Forced-choice direction discrimination at SOA 0 separated for detected (det.) and non-detected (non-det.) target motions Target detection
Discrimination (%)
t(12)
P
Direction (det.) Direction (non-det.)
92.7 (2.1) 53.1 (2.5)
19.713 1.250
,0.001 0.235
Means and standard deviations (in parentheses) are given. Direction discrimination was significantly different from chance level (50%) in case of target detection, but not in case of misses (paired t-test).
Fig. 2. Computation of difference ERPs. ERPs evoked by coherent motion target onset (a) are superimposed on brain activity associated with processing of the cue alone (b). Since the SOA between these events was varied, the superimposition affected the various SOAs differentially. Assuming an additive effect of the processing of cue and target, pure motion-related ERPs were identified by subtracting the effect of cue processing alone (a–b).
for grand-average ERPs: the transient early positive and negative deflections as well as the later slow positive wave that characterized the cue-related ERP (b) obviously affected the time course of ERPs evoked by both cue and motion targets (a). The difference potentials (a–b) reflecting coherent motion processing alone are characterized by a transient negativity peaking at about 200 ms after motion onset (labelled as N200), followed by a slow positive shift reaching its maximum between 400 and 500 ms after motion onset (labelled as P300). Our first analysis focused on the effect of SOA, which clearly affected the subject’s detection performance. Fig. 3 shows the grand-average difference ERPs, for different delays, and different posterior electrode positions. Only correct detections were considered in this averaging procedure, since the ERP correlates of ‘motion blindness’ were the subject of the second analysis. The analysis of the early N200 negativity (temporal window 160–240 ms) revealed a significant effect with SOA [F(2,24)53.33, P,0.05], which was not affected by the laterality of electrode placement but by spatial region [‘SOA’3‘spatial region’ F(2,24)52.995, P,0.05]. Post-hoc tests indicated that a significant increase of the N200 amplitude was restricted to temporal and parietal, but not occipital leads. The increase in amplitude with increasing lag was restricted to the temporo-parietal electrodes, and did not spread to the frontocentral regions (Fz, Cz). In contrast, effects were more clearly expressed and more widespread in the late temporal window (400–480 ms). A main effect of SOA was obtained [F(2,24)529.25, P,0.001], which described a significant increase of the late positivity with increasing lag. The difference between the single SOAs was more pronounced at parietal leads
M. Niedeggen et al. / Cognitive Brain Research 13 (2002) 241 – 247
245
Fig. 3. Grand-average of the difference wave ERPs (n513) separated for the effect of SOA (0, 100, and 300 ms). Onset of single traces (at 0 ms) was aligned to the onset of the relevant coherent motion stimulus in the RSVP. The ERPs shown for the posterior electrode leads are based only on target motion that was correctly detected. ERPs for missed targets were not considered. At temporal (T5, T6) and lateral parietal (P3, P4) electrodes the N200 component increased with increasing SOA after the cue. The mean amplitude effect is exemplified for the electrode P3, in the left upper part of the figure. For the late P300-like component a main effect of SOA can be seen, independent of the electrode position. As shown in the right upper diagram, the amplitude increase was strongly expressed at SOA 300 ms compared to the early delays 0, and 100 ms, respectively.
compared to temporal and occipital leads [‘SOA’3‘spatial region’: F(2,24)520.75, P,0.001]. An increase of the P300 was also found at fronto-central sites [‘SOA’: F(2,24)529.06, P,0.001]. The ERP effects described above demonstrate the impact of the temporal separation between cue and target, but are not related to the subjects’ detection performance, since only correct responses were considered in this analysis. In order to analyse the correlates of motion blindness we compared the ERP responses associated with hits and with misses, at SOA 0 ms (Fig. 4). For larger delays, sweep numbers for misses were insufficient to allow separate analysis of hits and misses. Focusing on the early negativity (160–240 ms), analysis of variance failed to show significant differences between hits and misses at posterior and fronto-central sites. In contrast, the late positivity, starting at about 300 ms, at parietal leads was clearly related to the observers’ detection performance. Whereas no clear P300 shift was observed for misses, it was significantly expressed for hits [main effect ‘Detection’: F(1,12)521.72, P,0.005]. The topography of the
effect was shifted towards parietal positions [‘Detection’3 ‘spatial region’: F(2,24)56.62, P,0.01], and also affected the fronto-central electrodes [F(1,12)54.901, P,0.05].
4. Discussion Our experimental results can be summarised as follows. With increasing delay between the cue and the motion target, the probability of missing the target dropped significantly. ERP analysis for correctly detected motion targets showed a concordant increase in amplitude with increasing SOA, for both the early posterior negativity, classically related to motion onset (N200), and a late centro-parietal positivity (P300). When ERPs were separated according to the subject’s detection performance, clear differences in amplitude were found in the late positive component, but not in the early negativity. Indeed, the distinct clear positive peak seen for correctly detected targets was completely absent for misses. The behavioural data replicate our previously report
246
M. Niedeggen et al. / Cognitive Brain Research 13 (2002) 241 – 247
Fig. 4. Grand-average of difference wave ERPs (n513) separated for correct motion detection and incorrect rejections, obtained at SOA 0 ms. The negativity at about 200 ms was not affected by the subject’s perceptive state, neither at the temporo-parietal nor at central electrodes. The lack of differential effects is represented in the left upper inset for the mean amplitudes obtained at electrode P3. However, the sustained, late positive shift, mostly expressed at centro-parietal leads, was significantly enhanced for correct detections. The right upper inset, referring to the mean amplitude shift obtained at electrode Cz, indicates that ERPs associated with missed targets lacked this brain activation.
[18], and confirm that attentional load can interact with the processing of first-order motion. The ERP results provide valuable information about the likely locus of this interaction. The N200 is known to be related to the processing of motion onset [2,9]. Although its topography is broadly distributed at posterior leads, temporal and parietal electrode positions have been found to be more sensitive to changes in physical motion parameters, i.e., to an increase in strength of the coherent motion signal [14]. In line with this topographical information, we found a significant amplitude modulation of the N200 as a function of SOA between cue and target. The N200 amplitude increased monotonically with increasing lag, and paralleled the improvement of behavioural detection performance. Although the analysis of SOA effects only considered correctly detected target motions, the amplitude modulation of the early ERP component appears to be related to the dynamics of an attentional mechanism. With increasing SOA, more capacity is available for visual processing of the relevant motion stimuli. Attention-sensitivity of the N200 component has previously been demonstrated for the onset of dynamic forms [23], and for transparent motion [20]. Recently, Vogel et al. [22] have shown suppression of the N200 in a time-based selective attention task, which was also based on the RSVP
technique. Comparable to the topography obtained in our experiment, the amplitude modulation of the N200 was more pronounced at parietal positions. Vogel et al. [22] related their observed suppression to the sluggish shift in attention between different surfaces defined by transparent motion. The prominent P300-like wave releasing the N200 was even more sensitive to increasing SOA. In contrast to the early sensory components, the late ERP positivity can be obtained at the fronto-central midline electrodes rather than at posterior-temporal leads. This finding is in line with previous experiments which relate the class of P300 components to higher-order, and amodal stimulus evaluation [7]. An attentional modulation in P300 amplitude as an effect of restricted processing capacity has been described in dual task performance [13], and in RSVP studies on the attentional blink [12,22]. In these studies, however, the reduction of the late components was not accompanied by an amplitude attenuation at a sensory level. Since we observed a reduction of both early and late components, it is not clear whether the reduction in P300 amplitude is a consequence of perceptual or post-perceptual filtering processes. It is important to note that we considered only correct responses in the analysis of the effect of SOA (Fig. 3). The increase in amplitudes of the early and late components
M. Niedeggen et al. / Cognitive Brain Research 13 (2002) 241 – 247
with increasing SOA can therefore be related to recovery of the attentional capacity, but not to the subject’s awareness of the target stimulus. As shown in Tables 1 and 2, motion detection appears to be a reliable indicator for conscious perception. False alarms occurred only rarely, and subjects were at chance level in reporting the direction of motion when they did not consciously perceive the motion. We have therefore averaged the ERP traces for those trials where the target motion was present and correctly detected (hits) and contrasted them with those in which the target motion was present but not detected (misses). The results showed differences in the late positive components only. In contrast, the N200 component was unaffected by detection performance. These results are in agreement with ERP studies on the attentional blink, which also observed a selective suppression of the P300 component, but no effect on early sensory evoked potentials [22]. Vogel et al. [22] concluded that the observer’s decision about the presence or absence of the target can be explained by a ‘late’ selection, following the sensory analysis. This conclusion appears to contradict the N200 findings in the previous study of Pinilla et al. [16] who found an amplitude modulation as a function of attentional object selection. We assume that our experiment bridges a gap between these experimental findings. Pinilla et al. [16] did not analyse the ERPs separately for hits and misses. Therefore, one cannot conclude whether the reduced attentional capacity, as reflected in N200 amplitude attenuation, affected the subjects’ detection ability. According to our data, attentional modulation of early ERP components is not necessarily related to visual target detection. Taken together, our results indicate a dissociation between attentional capacity in sensory processing and visual awareness. With respect to the SOA between cue and target, we have shown that the ERP correlates of sensory processing of first-order motion stimuli can be affected by time-based attentional selection in RSVP tasks. The attentional effect is probably related to the release of previous suppression, which affects integrative motion processing [17], and therefore the strength of the perceived motion coherence [14]. However, our experiments have also shown that the attentional modulation of early sensory processing—as reflected in the N200—is not sufficient to explain ‘motion blindness’. The detection of target motion was found to be related to a late P300-like component, and therefore, to further processing following sensory analysis. The absence of this component in incorrectly rejected targets may indicate that the relevant visual information is not updated in working memory [7].
Acknowledgements This work was supported by a grant from the VolkswagenStiftung (VW 2277) to M.N., A.S., M.M., and C.B.
247
References [1] D. Alais, R. Blake, Neural strength of visual attention gauged by motion adaptation, Nat. Neurosci. 2 (1999) 1015–1018. [2] M. Bach, D. Ullrich, Motion adaptation governs the shape of motion-evoked cortical potentials, Vision Res. 34 (1994) 1541– 1547. ¨ [3] C. Buchel, O. Josephs, G. Rees, R. Turner, C.D. Frith, K.J. Friston, The functional anatomy of attention to visual motion. A functional MRI study, Brain 121 (1998) 1281–1294. [4] P. Cavanagh, Attention-based motion perception, Science 257 (1992) 1563–1565. [5] A. Chaudhuri, Modulation of the motion aftereffect by selective attention, Nature 344 (1990) 60–62. [6] M. Dick, S. Ullman, D. Sagi, Parallel and serial processing in motion detection, Science 237 (1987) 400–402. [7] E. Donchin, M.G.H. Coles, Is the P300 component a manifestation of context updating?, Behav. Brain Sci. 11 (1988) 357–374. [8] T.S. Horowitz, A. Treisman, Attention and apparent motion, Spatial Vision 8 (1994) 193–219. ´ M. Kuba, H. Spekreijse, C. Blakemore, Contrast [9] Z. Kubova, dependence of motion-onset and pattern-reversal evoked potentials, Vision Res. 35 (1995) 197–205. [10] Z.L. Lu, G. Sperling, Attention-generated apparent motion, Nature 377 (1995) 237–239. [11] A. Luschow, H.C. Nothdurft, Pop-out of orientation but not pop-out of motion at isoluminance, Vision Res. 33 (1993) 91–104. [12] G. McArthur, T. Budd, P. Michie, The attentional blink and P300, NeuroReport 10 (1999) 3691–3695. [13] A.J. Nash, M. Fernandez, P300 and allocation of attention in dual-tasks, Int. J. Psychophysiol. 23 (1996) 171–180. [14] M. Niedeggen, E.R. Wist, Characteristics of visual evoked potentials generated by motion coherence onset, Cogn. Brain Res. 8 (1999) 95–105. [15] K.M. O’Craven, B.R. Rosen, K.K. Kwong, A. Treisman, R.L. Savoy, Voluntary attention modulates fMRI activity in human MTMST, Neuron 18 (1997) 591–598. [16] T. Pinnilla, A. Cobo, K. Torres, M. Valdes-Sosa, Attentional shifts between surfaces: effects on detection and early brain potentials, Vision Res. 41 (2001) 1619–1630. [17] J. Raymond, Attentional modulation of visual motion perception, Trends Cogn. Sci. 4 (2000) 42–50. [18] A. Sahraie, M. Milders, M. Niedeggen, Attention induced motion blindness, Vision Res. 41 (2001) 1613–1617. [19] K.L. Shapiro, J.E. Raymond, K.M. Arnell, Attention to visual pattern information produces the attentional blink in rapid serial visual presentation, J. Exp. Psychol.: Hum. 20 (1994) 357–371. [20] M. Valdes-Sosa, M.A. Bobes, V. Rodriguez, T. Pinilla, Switching attention without shifting the spotlight: object-based attentional modulation of brain potentials, J. Cogn. Neurosci. 10 (1998) 137– 151. [21] M. Valdes-Sosa, A. Cobo, T. Pinnilla, Attention to object files defined by transparent motion, J. Exp. Psychol.: Hum. 24 (2000) 1656–1674. [22] E. Vogel, S.J. Luck, K.L. Shapiro, Electrophysiological evidence for a postperceptual locus of suppression during the attentional blink, J. Exp. Psychol.: Hum. 24 (1998) 1656–1674. [23] J. Wang, Y. Jin, F. Xiao, S. Fan, L. Chen, Attention-sensitive visual event-related potentials elicited by kinetic forms, Clin. Neurophysiol. 110 (1999) 329–341.
