Neural mechanisms of attentional modulation ofperceptual grouping by collinearity

Yanhong Wu, Jun Chen and Shihui HanCA

Department of Psychology, Peking University, Beijing100871, PRChina

CACorresponding Author: shan@pku.edu.cn

Received 4 February 2005; accepted 22 February 2005

Psychophysical research showed that detection of an oriented vi-sual target is facilitated when the target is grouped with collinearvisual £ankers. However, this collinear grouping e¡ect is evidentonly when the £ankers are attended. This study examined neuralmechanisms underlying the interaction between attention andgrouping by collinearity. Event-related potentials were recordedfrom study participants who judged whether oriented Gaborpatches (i.e. visual elements consisting of a sinusoidal contrastmodulation convolvedwith aGaussian function) along the cued or-ientation were collinear or orthogonal. Event-related potentialsshowed an enhancednegativity over the posterior occipital cortex

at 48^72ms when collinear patches were congruent ratherthan incongruent with the cued orientation. A negative shift be-tween 260 and 380ms was observed over the occipital^parietalareas in the congruent rather than incongruent conditions.The long-latency e¡ect, however, was evident only when the colli-near patches were allocated along 451.The event-related potentialresults suggest that the interaction between attention andcollinear grouping may take place as early as in the primaryvisual cortex and is independent of global orientations ofperceptual groups. NeuroReport 16:567^570 �c 2005 LippincottWilliams &Wilkins.

Key words: Attention; Collinear grouping; Event-relatedpotential; Orientation

INTRODUCTIONOur perception of an ordered visual world is based ongrouping processes that integrate local elements into globalconfigurations. It is widely accepted that such groupingoperations occur at early stages of visual perception [1].Human neural correlates of grouping processes have beeninvestigated in neuroimaging studies. For example, event-related brain potential (ERP) studies found that perceptualgrouping defined by proximity (i.e. integration of spatiallyclose objects into a whole) results in enhanced neuralactivities at 100ms after sensory stimulation [2]. Functionalmagnetic resonance imaging (fMRI) studies further identi-fied generators of such early grouping activity in the humanprimary visual cortex [3]. Similarly, there has been evidencefor the engagement of the primary visual cortex in thegrouping process defined by collinearity (i.e. integration oforiented visual elements of which orientations match theglobal orientation of a virtual line drawn through allelements). Neuronal responses in monkeys’ primary visualcortex are enhanced when stimuli inside and outside theneurons’ receptive fields form a collinear group [4].Collinear grouping also facilitates the detection of centralGabor patches (a Gabor patch consists of a sinusoidalcontrast modulation convolved with a Gaussian function) inhumans [5] and is associated with an increased activity at80–140ms over the middle occipital area [6], suggesting theinvolvement of the human primary visual cortex in collineargrouping.It has been argued that grouping processes may take

place independent of focal attention [7]. However, recentresearch has shown evidence for the interaction between

attention and perceptual grouping in the human primaryvisual cortex [3]. Neural activities in the primary visualcortex around the calcarine sulcus associated with proxi-mity grouping are strengthened when stimulus arrays are ofhigh task relevance and fall inside attended areas. Similarly,a recent psychophysical study showed that collinearflankers facilitate the detection of a central Gabor patchwhen flankers are attended but not when flankers areignored [8]. This lateral interaction possibly reflects atten-tional modulation of flanker-target grouping defined bycollinearity. The current study investigated neural mechan-isms underlying the interaction between attention andgrouping by collinearity using a cueing paradigm similarto that used in prior psychophysical experiments [8,9].Study participants were cued to allocate their attentionalong a specific orientation (451 or 1351) before they wereshown a stimulus array of Gabor patches, as illustrated inFig. 1. Neural mechanisms underlying attentional modula-tion of grouping by collinearity were investigated bycomparing ERPs with stimulus arrays between the condi-tions when attentional allocation was either congruent orincongruent with the orientation of perceptual groupsformed by collinear Gabor patches.

MATERIALS AND METHODSStudy participants: Sixteen undergraduate and graduatestudents (nine women, seven men; 18–27 years of age, mean22 years) from Peking University participated in this study.All participants were right-handed and had normal orcorrected-to-normal vision. Informed consent was obtained

VISION,CENTRAL NEUROREPORT

0959-4965�c LippincottWilliams &Wilkins Vol 16 No 6 25 April 2005 567

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

from all the participants. Four of them were excluded fromdata analyses because of excessive eye blinks duringelectrophysiological data recording.

Stimuli and procedure: Stimuli were displayed on a graybackground (25.1 cd/m2). Each stimulus array consisted of aconfiguration of Gabor patches (see Fig. 1). The centralGabor patch was orientated either 451 or 1351 and wasflanked by two pairs of patches in an ‘X’ configuration. Theflankers were either collinear with or orthogonal with thecentral Gabor patch. At a viewing distance of 120 cm, eachGabor patch had a wavelength (l) and Gaussian distribu-tion equal to 0.451 of visual angle (spatial frequency, 2.2cycles per degree), with center-to-center separation of 4.2lbetween the central Gabor and each flanker. Target stimulusarrays consisted of Gabor patches with carrier wavelengthand Gaussian distribution of contrast envelope both equal to0.361 of the visual angle (spatial frequency, 2.8 cycles perdegree). Center-to-center separation between center andflanker was 3.7l. The target stimulus arrays were 70%smaller than the nontarget stimulus arrays.Participants pressed a button with the left or right index

finger to start each block of trials. On each trial, the fixationcross (0.61�0.61) and two peripheral bars (each with lengthof 0.61) were presented for 300ms. The two bars cued theorientation (451 or 1351) along which the Gabor patcheswere congruent or incongruent. After an interstimulusinterval that varied randomly between 300 and 600ms, astimulus array of Gabor patches was shown for 200ms,which was followed by a blank screen that varied randomlybetween 1000 and 1500ms. After two blocks of 48 trials forpractice, each participant was presented with 10 blocks of 48trials. There were 25% of targets in each block of trials.Participants were asked to decide whether the Gaborpatches in target stimuli at cued orientation were collinearor orthogonal by pressing one of the buttons with the left orright index finger. The assignment of the left and rightfinger to ‘yes’ and ‘no’ response was counterbalanced acrossparticipants.

Electrophysiological data recording and analysis: Theelectroencephalogram (EEG) was recorded from electrodesat 10–20 standard positions and five other pairs ofnonstandard sites. An electrode at the right mastoid wasused as reference. The electrode impedance was kept lessthan 5 kO. The EEG was amplified by using a band pass of0.1–75Hz (1/2 amplitude cutoffs), digitized at 250Hz/channel. Eye blinks were monitored with an electrodelocated below the right eye. The horizontal electrooculo-gram (EOG) was recorded from electrodes placed about1.5 cm lateral to the left and right external canthi. ERPs wereaveraged offline using a computer program that extractedepochs of EEG beginning 200ms before stimulus onset andcontinuing for 1200ms. Trials containing eye blinks, eyemovement deflections exceeding 750mv at any electrode,or incorrect behavioral responses were excluded from theERP averages. The baseline for ERP measurements wasthe mean voltage of a 200-ms prestimulus interval and thelatency was measured relative to the stimulus onset.Reaction times were submitted to repeated measures

analyses of variance (ANOVAs) with response types (‘yes’vs. ‘no’ response) and attended orientation (451 vs. 1351) asindependent variables. Mean ERP amplitudes at specific

time windows elicited by nontargets were submitted toANOVAs with congruency (the collinear group wascongruent or incongruent with the orientation along whichattention was allocated) and orientation of the collineargroup (451 vs. 1351) as independent variables.

RESULTSBehavioral data: Response accuracy did not differ be-tween ‘yes’ and ‘no’ responses [95.0% vs. 92.3%, t(11)¼1.53,p40.1]. Significant main effects of response type andattended orientation were observed on reaction times [F(1,11)¼21.44 and 7.00, respectively, both po0.03]. ‘yes’response (728724ms) were faster than ‘no’ response(810728ms). Responses were faster when attention wasallocated along 451 rather than along 1351. However, theinteraction between response types and attended orientationwas not significant [F(1,11)¼1.13, p40.3].

Event-related potential data: The grand-average ERPsover the posterior electrodes elicited by nontarget stimuliin each stimulus condition are illustrated in Fig. 1. ERPs tonontargets were characterized with a negativity peakingbetween 40 and 80ms (C1), which was followed by apositivity at 80–130ms (P1) and a negativity at 140–190ms(N1) over the lateral occipital sites. The mean P1 and N1amplitudes did not vary as a function of congruency andorientation of the collinear group (p40.05). However, therewas a main effect of congruency on the mean amplitudesbetween 48 and 72ms at occipital electrodes [O1:F(1,11)¼5.94, po0.04; O2: F(1,11)¼7.74, po0.02; OZ:F(1,11)¼4.92, po0.05] suggesting that the mean amplitudesin the time window were larger in the congruent than in theincongruent conditions. Also, a main effect of congruencyon the mean amplitudes at 200–420ms at posterior electro-des [O1: F(1,11)¼6.21, po0.03; O2: F(1,11)¼5.51, po0.04; OZ:F(1,11)¼7.24, po0.02; PZ: F(1,11)¼4.52, po0.057; P3:F(1,11)¼5.33, po0.042; P4: F(1,11)¼5.25, po0.04] was ob-served, indicating that there was a negative shift in thecongruent rather than incongruent conditions at theoccipital–parietal areas. Moreover, there was a significantinteraction between congruency and orientation of thecollinear group at 260–340ms [F(1,11)45.48, po0.04]. Post-hoc analyses confirmed that the congruency effect wassignificant only when the collinear patches were arrangedalong 451 [451 O1: F(1,11)¼20.93, po0.001; O2: F(1,11)¼7.76,po0.018; OZ: F(1,11)¼11.69, po0.006; PZ: F(1,11)¼7.82,po0.017; P3: F(1,11)¼29.33, po0.001; P4: F(1,11)¼5.59,po0.038; 1351 F(1,11)o3.79, p40.08 for all posteriorelectrodes].

DISCUSSIONWe used a cueing paradigm to examine neural mechanismsunderlying the interaction between attention and groupingby collinearity. Participants were able to allocate theirattention along the global cued orientation, as indicated byhigh response accuracy. Responses to collinear groups ofGabor patches (‘yes’ responses) were faster than those tononcollinear groups of Gabor patches (‘no’ responses),reflecting the fact that identification of a collinear groupwas either faster or earlier than identification of a noncol-linear group. Moreover, the advantage of ‘yes’ response over‘no’ responses did not differ between conditions when the

56 8 Vol 16 No 6 25 April 2005

NEUROREPORT Y.WU, J.CHENAND S.HAN

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

collinear group was allocated along 451 or 1351, indicatingthat response speeds were independent of the globalorientations of perceptual groups.The neural mechanisms of attentional modulation of

grouping by collinearity were indexed by the differences inERPs between the conditions when attentional allocationwas either congruent or incongruent with the globalorientations of perceptual groups composed of collinearGabor patches. We found that nontarget stimulus arrayselicited an early negative wave peaking between 40 and80ms after stimulus onset, which was enlarged by attentionallocated along the collinear group in stimulus displays.Both the time course and morphology suggest that thisnegativity is the C1 component that has been identified to

have neural generators in the human primary visual cortexaround the calcarine sulcus [10,11]. Because stimulus arrayswere identical in the congruent and incongruent conditions,the C1 effect could not arise from any difference in stimulusfeatures. Thus, our ERP results suggest that attention alongthe collinear group resulted in enhancement of neuralactivities in the primary visual cortex as early as 50ms afterstimulus onset, providing electrophysiological evidence forthe interaction between attention and grouping by colli-nearity in the primary visual cortex. Because the task usedin the current study emphasized the global orientation of aperceptual group rather than the orientation of the centralGabor patch, the C1 effect suggests that the integration ofcollinear Gabor patches involved neural mechanisms in the

−12

−6

0C1 C1

N1

(a)

(b)

N1O1 O2µV

6

−100 100 300

P1 P1

500 (ms)

−12

−6

0

C1 C1

N1N1

O1 O2µV

µV

µV

6

−100 100 300

P1 P1

500 (ms)

45° Collinear groupattend 45°

45° Collinear groupattend 135°

135° Collinear groupattend 135°

135° Collinear groupattend 45°

Fig. 1. Illustrations of stimulus arrays and grand average event-related potentials (ERPs) elicited by 451 and 1351 collinear groupings under di¡erentconditions recorded at occipital electrode (O1,O2). (a) ERPs elicitedby 451 collinear group under congruent and incongruent conditions. (b) ERPs elicitedby1351 collinear grouping under congruent and incongruent conditions.

Vol 16 No 6 25 April 2005 569

PERCEPTUALGROUPINGBYCOLLINEARITY NEUROREPORT

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

primary visual cortex, possibly through long-range hor-izontal connections linking neurons with common orienta-tion tunings [12].Prior ERP studies have shown that the C1 component

evoked by stimulus arrays is modulated by whether localelements in the stimulus display are grouped into columnsor rows by proximity [13]. In addition, the proximity-grouping-related activity in the calcarine cortex is modu-lated by whether stimulus arrays are of high task relevanceand are located inside an attended area [3]. In accordancewith the previous findings, the current ERP results suggestthat the interaction between attention and grouping opera-tions defined by different principles, such as proximity andcollinearity, may share a common neural mechanism in theprimary visual cortex.The current ERP study also showed evidence for a long-

latency effect of the interaction between attention andgrouping by collinearity. A negative shift was observed inthe congruent rather than incongruent conditions at theoccipital–parietal areas at 200–420ms. Because the long-latency effect was observed in ERPs to nontarget stimuli thatdid not require behavioral responses, it is unlikely that thiseffect reflected the process after perceptual processing suchas response selection or execution. Interestingly, the long-latency effect depended upon the global orientation ofcollinear groups, being significant only when collinearGabor patches were allocated along 451. This effect has notbeen reported in prior psychophysical studies [8,9] andcannot be simply accounted for by attentional allocation,which was decided by peripheral cues that appeared beforethe presentation of Gabor patch displays. A possibleinterpretation of this orientation-dependent long-latencyeffect is that the long-latency process of collinear groupingalong 1351 was less perceptually salient than that along 451and, thus, was less sensitive to the prior allocation of spatialattention. This proposal is consistent with the fact thatbehavioral responses to the perceptual groups were slowerwhen the perceptual group required to be identified wasalong 1351 than when along 451. However, this proposalneeds further evidence. Whatever the case, our ERP resultscomplement previous psychophysical research by showingthat there might be two distinct phases of interactionbetween spatial attention and grouping by collinearity.

CONCLUSIONThe early interaction between spatial attention and collineargrouping may have occurred in the primary visual cortexand was independent of global orientations of perceptualgroups. Also, a long-latency interaction occurred betweenattention and collinear grouping, which depended upon theglobal orientations of collinear groups.

REFERENCES1. Vecera PV, Behrmann M. Attention and unit formation: A biased com-

petition account of object-based attention. In: Shipley TF, Kellman PJ

(eds). From Fragments to ObjectsFSegmentation and Grouping in Vision.London: Elsevier; 2001. pp. 145–182.

2. Han S, Song Y, Ding Y, Yund EY, Woods DL. Neural substrates for visual

perceptual grouping in human. Psychophysiology 2001; 38:926–935.

3. Han S, Jiang Y, Mao L, Humphreys GW, Gu H. Attentional modulation of

perceptual grouping in human visual cortex: fMRI studies. Hum BrainMapp 2005 (in press).

4. Kapadia MK, Ito M, Gilbert CD, Westheimer G. Improvement in visual

sensitivity by changes in local context: parallel studies in human

observers and in V1 of alert monkeys. Neuron 1995; 15:843–856.

5. Polat U, Sagi D. Lateral interactions between spatial channels:

suppression and facilitation revealed by lateral masking experiments.

Vision Res 1993; 33:993–999.6. Khoe W, Freeman E, Woldorff MG, Mangun GR. Electrophysiological

correlates of lateral interactions in human visual cortex. Vision Res 2004;44:1695–1673.

7. Humphreys GW. Neural representation of objects in space: a dual coding

account. Philos Trans R Soc Lond B Biol Sci 1998; 353:1341–1351.

8. Freeman E, Sagi D, Driver J. Later interactions between targets and

flankers in low-level vision depend on attention to the flankers. NatNeurosci 2001; 4:1032–1036.

9. Freeman E, Sagi D, Driver J. Configuration-specific attentional

modulation of flanker-target lateral interactions. Perception 2004; 33:

181–194.

10. Clark VP, Fan S, Hillyard SA. Identification of early visual evoked

potential generators by retinotopic and topographic analyses. Hum BrainMapp 1995; 2:170–187.

11. Noesselt T, Hillyard SA, Woldorff MG, Schoenfeld A, Hagner T, Jancke L

et al. Delayed striate cortical activation during spatial attention. Neuron2002; 35:575–587.

12. Chisum HJ, Mooser F, Fitzpatrick D. Emergent properties of layer 2/3

neurons reflect the collinear arrangement of horizontal connections in

tree shrew visual cortex. J Neurosci 2003; 23:2947–2960.13. Mao L, Han S, Guo C, Jiang Y. Neural mechanisms of perceptual

grouping in human visual cortex. Chinese Sci Bull 2004; 49:

819–823.

Acknowledgements:This work was supportedby the National Natural Science Foundation of China (Project 30225026, 30328016,30470568, 30370476), the Ministry of Science and Technology of China (Project 2002CCA01000).

570 Vol 16 No 6 25 April 2005

NEUROREPORT Y.WU, J.CHENAND S.HAN

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.