Spatiotemporal dynamics of feature-based attention spread: evidence from combined electroencephalographic and magnetoencephalographic recordings.

Abstract

Attentional selection on the basis of nonspatial stimulus features induces a sensory gain enhancement by increasing the firing-rate of individual neurons tuned to the attended feature, while responses of neurons tuned to opposite feature-values are suppressed. Here we recorded event-related potentials (ERPs) and magnetic fields (ERMFs) in human observers to investigate the underlying neural correlates of feature-based attention at the population level. During the task subjects attended to a moving transparent surface presented in the left visual field, while task-irrelevant probe stimuli executing brief movements into varying directions were presented in the opposite visual field. ERP and ERMF amplitudes elicited by the unattended task-irrelevant probes were modulated as a function of the similarity between their movement direction and the task-relevant movement direction in the attended visual field. These activity modulations reflecting globally enhanced processing of the attended feature were observed to start not before 200 ms poststimulus and were localized to the motion-sensitive area hMT. The current results indicate that feature-based attention operates in a global manner but needs time to spread and provide strong support for the feature-similarity gain model.

Full text

Page 1

Behavioral/Systems/Cognitive
SpatiotemporalDynamicsofFeature-BasedAttention
Spread:EvidencefromCombinedElectroencephalographic
andMagnetoencephalographicRecordings
ChristianMichaelStoppel,1CarstenNicolasBoehler,2,3HendrikStrumpf,1RuthMarieKrebs,1,3Hans-JochenHeinze,1,2
Jens-MaxHopf,1,2andMirceaArielSchoenfeld1,2,4
1DepartmentofNeurologyandCenterforAdvancedImaging,Otto-von-Guericke-University,39120Magdeburg,Germany,2Leibniz-Institutefor
Neurobiology,39118Magdeburg,Germany,3DepartmentofExperimentalPsychology,GhentUniversity,9000Ghent,Belgium,and4KlinikenSchmieder,
78476Allensbach,Germany
Attentional selection on the basis of nonspatial stimulus features induces a sensory gain enhancement by increasing the firing-rate of
individualneuronstunedtotheattendedfeature,whileresponsesofneuronstunedtooppositefeature-valuesaresuppressed.Herewe
recorded event-related potentials (ERPs) and magnetic fields (ERMFs) in human observers to investigate the underlying neural corre-
latesoffeature-basedattentionatthepopulationlevel.Duringthetasksubjectsattendedtoamovingtransparentsurfacepresentedinthe
left visual field, while task-irrelevant probe stimuli executing brief movements into varying directions were presented in the opposite
visualfield.ERPandERMFamplitudeselicitedbytheunattendedtask-irrelevantprobesweremodulatedasafunctionofthesimilarity
between their movement direction and the task-relevant movement direction in the attended visual field. These activity modulations
reflectinggloballyenhancedprocessingoftheattendedfeaturewereobservedtostartnotbefore200mspoststimulusandwerelocalized
tothemotion-sensitiveareahMT.Thecurrentresultsindicatethatfeature-basedattentionoperatesinaglobalmannerbutneedstimeto
spreadandprovidestrongsupportforthefeature-similaritygainmodel.
Introduction
The capacity for perceptual analysis of incoming sensory infor-
mation is limited within the human brain. Consequently, when
multiple stimuli compete for representation, our visual system
hastoselectwhichinputshouldbepreferentiallyprocessedatthe
expenseofotherinformation(DesimoneandDuncan,1995;Ser-
ences and Yantis, 2006). Psychophysical and neurophysiological
evidence indicates that this selection can be accomplished by fo-
cusing attention to particular spatial locations (Posner, 1980;
Heinzeetal.,1994).Furthermore,attentionmayalsobedeployed
tothenonspatialattributesofastimulus,suchasitscolor,shape,
ormotion(HillyardandMu ¨nte,1984;Corbettaetal.,1990;Brit-
ten et al., 1992), which results in increased activity within those
cortical modules that are specialized in processing the respective
features(O’Cravenetal.,1997;Chawlaetal.,1999;Schoenfeldet
al., 2007).
Beyond this response enhancement for attended compared
with unattended stimulus-features, pioneering studies by Treue
and coworkers (Treue and Martínez Trujillo, 1999; Treue and
Maunsell, 1999; Martinez-Trujillo and Treue, 2004) demon-
stratedthatfeature-basedattentionmodifiestheresponseprofile
of feature-selective neurons in a multiplicative manner. This
“feature-similarity gain” mechanism operates by amplifying the
firing-rate of neurons whose feature-preference closely matches
the attended feature value (e.g., one specific motion direction
of a stimulus), while the firing of neurons tuned to opposite
feature-values (e.g., other motion directions) is suppressed.
This multiplicative gain enhancement observed by single-
neuron recordings in primates has been proposed to result in
improved feature selectivity at the level of an integrated pop-
ulation response. Recently this assumption has been con-
firmed using fMRI in human observers (Kamitani and Tong,
2006; Stoppel et al., 2011), with stimuli that were presented
within the focus of spatial attention.
In addition to these feature-based effects within the focus of
attention, feature-selection has also been shown to modulate the
firing-rate of neurons in an entirely location-independent man-
ner. An enhanced response of neurons tuned to a specific feature
couldbeobservedalthoughstimuliwerepresentedoutsideofthe
neuron’s receptive field (Treue and Martínez Trujillo, 1999;
Martinez-TrujilloandTreue,2004;Bichotetal.,2005).Recently,
suchspatiallyglobalfeature-selectivemodulationshavealsobeen
described at the population-level using fMRI (Saenz et al., 2002;
Kamitani and Tong, 2006; Serences and Boynton, 2007) and re-
cordingsofsteady-statevisualevokedpotentialsinhumans(An-
dersen et al., 2009, 2011). Nevertheless, little is known about the
timing of this type of modulation.
ReceivedJan.31,2012;revisedApril28,2012;acceptedMay4,2012.
Authorcontributions:C.M.S.,C.N.B.,J.-M.H.,andM.A.S.designedresearch;C.M.S.andH.S.performedresearch;
C.M.S.,H.S.,andM.A.S.analyzeddata;C.M.S.,C.N.B.,R.M.K.,H.-J.H.,J.-M.H.,andM.A.S.wrotethepaper.
Thisworkwassupportedbythefollowinggrants:Scho1217/1-2andSFB779-A1fromtheDeutscheForschungs-
gemeinschaft(DFG)awardedtoM.A.S.
Correspondence should be addressed to Christian Stoppel, Department of Neurology, Otto-von-Guericke-
UniversityMagdeburg,LeipzigerStrasse44,39120Magdeburg,Germany.E-mail:christian.stoppel@med.ovgu.de.
DOI:10.1523/JNEUROSCI.0439-12.2012
Copyright©2012theauthors 0270-6474/12/329671-06$15.00/0
TheJournalofNeuroscience,July11,2012 • 32(28):9671–9676 • 9671

Page 2

Inthepresentstudywesimultaneously
recorded electroencephalographic and
magnetoencephalographic activity time-
locked to the motion onset of a spatially
unattended random-dot kinematogram
toinvestigatethetimecourseandtheneu-
ral substrates of global feature-based at-
tentionalselectionatthepopulationlevel.
Participants were required to attend to a
moving transparent surface to perform in
a motion discrimination task, while a
second surface presented to the oppo-
site visual field performed brief transient
movements into varying directions. This
design permitted us to quantify the mag-
nitudeandlatencyofevent-relatedpoten-
tials (ERPs) and event-related magnetic
fields (ERMFs) evoked by the unattended
surface, in dependence of the similarity
between its motion direction and that of
the attended surface.
MaterialsandMethods
Subjects. Sixteen right-handed healthy subjects
(mean age: 27.0 ? 3.7 SD years, 4 males), all
with normal or corrected-to-normal visual acu-
ity, participated as paid volunteers in the study.
The local ethics committee of the Otto-von-
Guericke University Magdeburg approved the
experiment and all subjects gave written in-
formedconsentbeforeparticipation.
Stimuliandexperimentaldesign.Stimuliwerepresentedagainstadark
background (0.5 cd/m2) within two square apertures (4.2° ? 4.2°) cen-
tered 5.7° to the left and right of a central fixation cross (0.8° ? 0.8°; Fig.
1).Eachaperturecontained100randomlydistributedisoluminantwhite
dots (brightness 200 cd/m2; dot size 0.08°). All dots within the left aper-
ture moved either coherently upward (during even runs) or downward
(during odd runs; velocity: 10°/s) and were perceived as a transparent
surface. The subjects’ task was to attend to this surface and to make a
speeded button-press response after detecting an accelerated movement
of the attended surface (velocity: 22°/s for 300 ms). Within the right
aperture all dots remained stationary throughout the experiment except
during probe trials in which all dots coherently performed a short dis-
placement into one of the eight cardinal or ordinal directions (velocity:
10°/s for 200 ms). These probe movements deviated from the motion
directionoftheattendedsurfaceby0°,45°,90°,135°,or180°(Fig.1)and
were completely task-irrelevant. During target trials (accelerated move-
ment of the attended surface in the left hemifield) the dots in the unat-
tended visual field always remained stationary. Conversely, on probe
trials (short displacement of the dots in the unattended aperture in the
right hemifield) there was never an accelerated movement of the at-
tended surface. Thus, probe and target trials always occurred indepen-
dentofeachotherinapredefinedpseudo-randomsequence,andalltrial
types(targetsandeachoftheprobetypesofdifferentdirectionality)were
presented equally often throughout the experiment. All trials (targets, as
well as probes) were separated by an intertrial interval that varied ran-
domly between 1250 and 1750 ms (mean 1500 ms). The experiment was
performedinsevenrunsof385seachwithrestperiodsin-between.Each
run consisted of 252 trials, resulting in a total of 294 trials per condition.
Throughout the experiment subjects were instructed to keep accurate
fixation, which was monitored by electro-oculogram (EOG; see below).
Data acquisition. ERPs and ERMFs were simultaneously recorded us-
ing a Magnes 3600 whole-head magnetoencephalographic (MEG) sys-
tem with 248 magnetometer and 32 electroencephalographic (EEG)
channels (4-D Neuroimaging/Biomagnetic Technologies Inc.). The sig-
nals were digitized at a rate of 508 Hz with an online bandpass of DC to
200 Hz. The horizontal EOG was recorded using a bipolar montage with
2 electrodes behind the lateral orbital angles, whereas the vertical EOG
wasrecordedfromanelectrodebelowtherightorbitallimb.Impedances
were kept below 5 k? and a midline fronto-polar electrode served as
ground. MEG signals were submitted to online and offline noise reduc-
tion (Robinson, 1989), and an artifact rejection was applied with peak-
to-peaklimitsof2–4pTfortheMEGand80–200?VfortheEOGsignal
(thresholds were adjusted individually for each subject, but were con-
stant over all experimental conditions). Individual head shapes were
coregistered with the sensor coordinate system by digitizing (Polhemus
3SpaceFastraksystem)skulllandmarks(nasion,left,andrightpreauric-
ular points) and determining their locations relative to sensor and elec-
trodepositionsusingsignalsfrom5spatiallydistributedcoilsattachedto
the subjects’ heads. To account for differences in head positions relative
to the MEG sensors, the individual subjects’ MEG data were aligned
using the following procedures. First a lead field was computed (using
Curry6.01CompumedicsNeuroscan)foreachsubjectusingthesubjects
individual sensor-configuration. Then an inverse solution for this lead
field was computed (using generalized least-squares methods) without
applying any sensor- or dipol-weighting. Finally regularization by a
modeling-parameter lambda was performed to prevent from ghost-
activity in source-space. Lambda was chosen to make a compromise
between largefull-width-half-maximum-values
function and noise-increment by ghost-activity. The result was an in-
verse solution that is independent from the measured sensor-data. In a
second step a forward solution was computed for a reference sensor grid
of248sensors.Theindividualsubject’sdata,nowinthereferencesensor
grid, were averaged together to obtain the grand average.
Dataanalysis.SeparateERPandERMFaveragewaveformswerecom-
putedtime-lockedtothemotiononsetforeachofthe5probeconditions.
Attention effects were quantified in these average waveforms as mean
amplitudemeasureswithinlatencyintervalsof110–210and210–310ms
poststimulusonset(withrespecttoa200msprestimulusbaseline)atthe
sensor/electrode sites showing the largest amplitudes. Statistical analysis
of the data was performed using within-subjects repeated ANOVAs
(Greenhouse-Geisser correction was applied when necessary). To deter-
mine the time of onset of the attention effects, amplitude measures were
takenoversuccessive10msintervalsandtestedforsignificantdifferences
ofpoint-spread-
135°
ISI
90°
0°
Target
(fast movement)
ISI (1250-1750 ms)
ISI
Figure1.
leftandrightvisualfield.Intheleftaperturealldotsmovedeithercoherentlyupward(duringevenruns)ordownward(duringodd
runs)andthuswereperceivedasatransparentsurface.Onsometrials,thissurfacemovedwithahighervelocity,andsubjects
respondedtothoseastargets.Withintherightaperturealldotsremainedstationaryduringtheintertrialinterval(ISI)andduring
targettrials,whileonprobetrialstheyperformedashortcoherentdisplacementintooneoftheeightcardinalorordinaldirections.
Whilethesemovementscoulddeviatefromthemotiondirectionoftheattendedsurfaceby0°,45°,90°,135°,or180°,theywere
completelyirrelevanttothetaskandhadtobeignoredbythesubjects.
Schematicillustrationoftheexperimentaldesign.Subjectswerepresentedwithtwosquaredapertureslocatedinthe
9672 • J.Neurosci.,July11,2012 • 32(28):9671–9676 Stoppeletal.•SpatiotemporalDynamicsofFeature-BasedAttention

Page 3

between conditions with a criterion of p ? 0.05. The earliest significant
intervalfollowedby5(ormore)successivesignificantintervalswastaken
astheonsetlatency(GuthrieandBuchwald,1991;Schoenfeldetal.,2003,
2007).
Source localization. For source localization, current source density
estimates were computed by means of standardized low-resolution
electromagnetic tomography (sLORETA; Pascual-Marqui, 2002) as
implemented in the neuroimaging software Curry 6.01 (Compumedics
Neuroscan). The sLORETA represents an extension of the minimum
norm least square method (Ha ¨ma ¨la ¨inen and Ilmoniemi, 1994; Fuchs et
al., 1999), where current estimates at each source location are weighted
by their measurement error, yielding a pseudo-F-value distribution of
currentsoverthecorticalsurface,calledsourcedensityestimates(SDEs).
All source localization results provided in Figures 2 and 3 represent such
SDEs. Since the distribution of magnetic fields measured by MEG is
oriented perpendicularly to the concurrent voltage field distribution as-
sessed by EEG, the surface topographies of both fields elicited by a given
dipolar source display a nearly orthogonal surface topography. There-
[µV]
1
-2.5
55
[fT]
-55
ERPERMF
-3 µV-50 fT
110-210 ms
500 ms
500 ms
110-210 ms
0°
45°
90°
135°
180°
0
-1
-2
-3
0° 45° 90° 135° 180°
mean amplitude [µV]
0
-10
-20
-50
-40
0°45° 90° 135° 180°
mean amplitude [fT]
0°
45°
90°
135°
180°
Source
CDR [F-distributed]
20004000
160 ms
0°
45°
90°
135°
180°
-30
0°
45°
90°
135°
180°
Figure2.
column)andERMF(rightcolumn)responsesinthetimerangebetween110and210msafteronsetoftheprobestimuli.Recordingsitesareindicatedasblackdotswithinthefielddistributionmaps.
ThemagnitudeofERPandERMFamplitudesshowsnodependencyonthemotiondirectionoftheattendedsurface.AminimumintheERPfielddistributioncanbeenseenoverleftparieto-occipital
electrodesites(lefttopographymaps),accompaniedbyanefflux-influxdistributionoftheERMFs,whichislocatedoverleftoccipitalsensors(righttopographymaps).Theestimatedcurrentsource
densitydistribution160mspoststimulusonsetshowsonemaximumlocatedinleftposteriorlateralextrastriatecortex.
Noglobalfeature-basedattentioneffectsinthetimerangeoftheN1component(110–210mspoststimulusonset).Timecoursesandmeanamplitudesoftheprobe-relatedERP(left
[µV]
3
-1
36
[fT]
-36
ERPERMF
0°
45°
90°
135°
180°
-2 µV
-20 fT
210-310 ms
500 ms
500 ms
210-310 ms
0°
45°
90°
135°
180°
0
1
2
3
0°45° 90° 135° 180°
mean amplitude [µV]
0
10
20
30
0°45° 90° 135° 180°
mean amplitude [fT]
0°
45°
90°
135°
180°
Source
CDR [F-distributed]
7001400
250 ms
0°
45°
90°
135°
180°
Figure3.
ERP(leftcolumn)andERMF(rightcolumn)responses.Recordingsitesareindicatedasblackdotswithinthefielddistributionmaps.NotethatthemagnitudeoftheERPandERMFamplitudes
parametricallydependsonthedeviationoftheprobes’motiondirectionfromthedirectionoftheattendedsurface(indicatedindegreesofvisualanglebywhichtheprobesdeviatedfromthe
attended motion direction). The topographical field distributions (averaged over the time range 210–310 ms) show a maximal positivity over midline central electrode sites for the ERPs (left
topographymaps)andanefflux-influxfieldtransitionlocatedoverleftoccipitotemporalsensorsfortheERMFs(righttopographymaps).Theestimatedcurrentsourcedensitydistribution250ms
poststimulusonset(displayedinthemiddleofthefigure)showsonemaximumlocatedintheleftmiddleoccipitotemporalcortex.
Globalfeature-basedattentionalmodulationsbetween210and310mspoststimulusonset.Timecoursesandmeanamplitudes(210–310mspoststimulusonset)oftheprobe-related
Stoppeletal.•SpatiotemporalDynamicsofFeature-BasedAttention J.Neurosci.,July11,2012 • 32(28):9671–9676 • 9673

Page 4

fore, both the ERP and ERMF distributions were concurrently fit by
sLORETA to obtain maximal localization power (Fuchs et al., 1998;
Schoenfeldetal.,2003).Oneprerequisitefortheconcomitantuseofboth
the ERP and ERMF field distributions is that the conductivities of the
volumeconductormodelarematchedfortheEEGandMEGrecordings.
Therefore,aconductivityfactorwasdeterminedtoscaletheEEGrelative
to the MEG data based on a tangential dipole evoked by tactile stimula-
tion of the index finger by an air puff at 30–40 ms latency (Fuchs et al.,
1998).Thisconductivityfactorcouldbereliablyapproximatedtoavalue
of 0.8 and was used for estimation of the source localization on the
average data across all subjects. All inverse computations were con-
strained by realistic anatomical models of volume conductor and source
compartment derived by 3-dimensional surface reconstructions of the
head, CSF space, and cortical surface, respectively (boundary element
method; Ha ¨ma ¨la ¨inen and Sarvas, 1989). The anatomical basis for the
sourceanalysiswastheMontrealNeurologicalInstitutebrain(averageof
152 T1-weighted stereotaxic volumes).
Results
Behavioral results
Subject were accurate at detecting the faster moving targets, with
a mean hit rate of 95.5% (SD: ? 4.1%) and a false alarm rate of
2.9%(SD:?2.6%).Meanreactiontimesrangedfrom414to501
ms (mean ? SD: 458 ? 31 ms).
ERP/ERMF results
Theeffectsoffeature-basedattentionondirection-selectiveneu-
ral activity were assessed, by comparing the ERP/ERMF wave-
forms elicited by the different probe stimuli. The ERP/ERMF
amplitudes within the time-range of the N1-component (110–
210 ms) were not significantly modulated by the similarity be-
tweenthemotiondirectionoftheattendedsurfaceandthatofthe
moving probe stimuli (ERPs, F(4,60)? 0.99, p ? 0.4, Fig. 2, left
column;ERMFs,F(4,60)?0.97;p?0.4,Fig.2,rightcolumn).The
topographicalfielddistributionsforallprobeconditionsshowed
a maximal negativity over left parieto-occipital electrodes in the
ERPs (Fig. 2, left topography maps) and one efflux-influx field
configuration located over left occipital sensors for the ERMFs
withonlyminimalvariationsinamplitude(Fig.2,righttopogra-
phy maps). The corresponding current source distribution
withintheN1time-range(at160mspoststimulusonset)revealed
source-activity estimates located in the left posterior lateral ex-
trastriate cortex (Talairach coordinates: ?35/?93/7) most likely
corresponding to visual areas V2 and V3.
In contrast, the ERP/ERMF amplitudes in the time-range
between 210 and 310 ms were modulated as a function of the
similarity between the motion direction of the probe and the
direction of the attended surface, with more negative ERP
(F(4,60)? 10.77; p ? 0.0001; Fig. 3, left column) and ERMF am-
plitudes (F(4,60)? 3.57; p ? 0.05; Fig. 3, right column) for probe
stimuli matching more closely the attended direction. Note that
the magnitude of the ERP and ERMF amplitudes between 210
and310msdecreasesasafunctionofthedeviationoftheprobes’
motion direction from that of the attended surface. Statistical
comparison in successive 10 ms epochs indicated that these dif-
ferences between probe conditions became significant at ?200
ms post-probe. In the subsequent interval between 210 and 310
ms, the ERP field distribution map showed a maximal positivity
overmidlinecentralelectrodesites(Fig.3,lefttopographymaps),
accompanied by an efflux-influx field configuration located over
left occipitotemporal sensors for the ERMFs (Fig. 3, right topog-
raphy maps). The source analysis revealed estimates of activity
(sLORETA estimates, see Materials and Methods) located in left
middle occipitotemporal cortex (Talairach coordinates: ?46/
?77/?1), most likely corresponding to region V5/hMT.
Discussion
The present study used simultaneous recordings of ERPs and
ERMFs in human observers to investigate the spatiotemporal
correlates of feature-based attentional selection at the neural
population level. During the task subjects attended to a moving
transparentsurfaceintheleftvisualfieldandperformedamotion
discrimination task, while a task-irrelevant second surface lo-
cated in the opposite visual field moved into different directions.
Thisexperimentaldesignpermittedtoquantifythemagnitudeof
ERPs and ERMFs evoked by the unattended surface while sys-
tematically varying the similarity between its motion direction
andthatoftheattendedsurface.Ourcurrentresultsdemonstrate
a parametric feature-based attentional modulation of ERP and
ERMF amplitudes, as a function of the similarity between the
motiondirectionsofthespatiallyattendedandunattendedstim-
uli. The time courses of the ERP and ERMF waveforms indicate
that this attentional enhancement starts not before 200 ms post-
stimulus onset and originates from left middle occipitotemporal
cortex, most likely corresponding to area V5/hMT. These find-
ingsprovidestrongsupportforthefeature-similaritygainmodel
by demonstrating that feature-based attention parametrically
modulates direction-selective population activity within V5/
hMT in a global manner. The timing indicates that the global
spread of attention toward spatially unattended locations does
not occur immediately after the selection of the attended feature
pointing out to the time-consuming nature of this process.
Theobservedfeature-basedmodulationswerelocalizedtothe
lateral middle occipitotemporal cortex, which corresponds well
toareaV5/hMT.Thisregion,consideredtobethehomologofthe
well described MT region in non-human primates, is specialized
for the processing of motion information (Zeki et al., 1991; Ahl-
fors et al., 1999). Beyond purely sensory-driven effects, activity
within this region can be markedly affected by attention. For
example, V5/MT neurons increase their firing rate to attended
comparedwithunattendedmotionstimuli(CookandMaunsell,
2002, 2004), which is consistent with findings in human subjects
(Corbetta et al., 1990, 1991; O’Craven et al., 1997; Schoenfeld et
al., 2007). In addition to this global response enhancement, re-
cent observations indicated that V5/MT activity could also be
affected in a direction-selective manner. Feature-based attention
modifies the response profile of direction-selective neurons
within V5/MT multiplicatively: neurons whose feature prefer-
ence closely match an attended motion direction increase their
firing rate, while the firing of neurons tuned to opposite direc-
tionsissuppressed(TreueandMartínezTrujillo,1999;Martinez-
TrujilloandTreue,2004).Asaconsequenceofthesefindingsthe
‘feature-similarity gain model’ has been formulated. This model
posits that an individual neuron’s response depends on the fea-
turesimilaritybetweenacurrentbehaviorallyrelevanttargetand
the feature preference of that neuron, resulting in an improved
selectivity for the attended feature at the population level. Re-
centlythishypothesishasalsobeenconfirmedusingfMRIexper-
iments in human observers with stimuli that were presented
within the focus of spatial attention (Kamitani and Tong, 2006;
Stoppel et al., 2011). Another fMRI study indicated that feature-
based attention might also operate in a spatially global manner,
i.e.,forstimulioccurringatspatiallyunattendedlocations(Saenz
et al., 2002). However, this study only compared stimuli that
moved into the attended versus opposed to the attended direc-
tion. More importantly, due to the nature of the fMRI technique
no inferences could be made on the timing. The current results
extend these findings (Saenz et al., 2002) by demonstrating that
9674 • J.Neurosci.,July11,2012 • 32(28):9671–9676 Stoppeletal.•SpatiotemporalDynamicsofFeature-BasedAttention

Page 5

electromagnetic population activity (i.e., ERP/ERMF ampli-
tudes)tomovingstimulipresentedatunattendedlocationsscales
parametrically in dependence of their feature-similarity with re-
spect to the attended stimulus.
Although fMRI studies provided a detailed picture on the an-
atomical structures modulated by feature-based selection, their
temporal resolution is too limited to reveal the timing of the
underlying attentional modulations. Fine-grained information
about the time course of feature-based selection has therefore
beendeterminedprimarilybasedondatafromnoninvasiveEEG/
MEG recordings in humans. By this means, previous studies in-
dicated that the selection of task-relevant features is initiated in
the time range of the N1-component, i.e., between 100 and 180
ms poststimulus onset (Harter and Aine, 1984; Kenemans et al.,
1993; Motter, 1994; Anllo-Vento and Hillyard, 1996; Smid et al.,
1999; Torriente et al., 1999; Kenemans et al., 2000; Martínez et
al., 2001; Beer and Ro ¨der, 2004, 2005; Schoenfeld et al., 2007),
whichinmostcasesisreflectedbyabroadnegativityovercentro-
posterior electrodes in the ERP (the so-called selection-negativ-
ity; Harter and Aine, 1984; Hillyard and Anllo-Vento, 1998). A
common feature of these studies was that the stimulus eliciting
the neurophysiological response was located in the attended part
of the space.
Thus, the feature-based selection occurred at the spatially at-
tended location. In the present study we observed an enhanced
negativity over centro-posterior electrodes in the EEG, whose
magnitude parametrically depended on the similarity between
the motion directions of the attended and the unattended sur-
faces(Fig.3).Thecorrespondingmodulationswerealsoobserved
in the simultaneously recorded ERMF over occipitotemporal
sensors. Importantly, the onset latency of these modulations was
later (?200 ms) than previously reported (Hillyard and Mu ¨nte,
1984; Anllo-Vento and Hillyard, 1996; Karayanidis and Michie,
1996;Langeetal.,1998;Schoenfeldetal.,2007).Giventhatinthe
present experiment the motion probes were located in the unat-
tended visual field, this latency difference is likely to reflect the
temporal costs underlying the spread of feature-selective modu-
lations toward spatially unattended locations.
However,therearealsootherexplanationsthatmightaccount
fortheobservedlatencydifference.Onepossibilitywouldbethat
the delay results mainly from time costs related to involuntary
attention capture processes triggered by the task-irrelevant stim-
uli (Egeth and Yantis, 1997). On the other hand the initiation of
an involuntary attention shift towards the task-irrelevant stimu-
lus could be regarded as an integral process of the feature-
selective attention spread that might be used to determine the
most useful direction of propagation.
A second possibility would be that the observed latency differ-
ence rather reflects a delay in the general processing of the task-
irrelevant stimulus than the time costs of the feature-based
attentionalspread.However,theprocessingoftask-relevantstimuli
within the same feature value (e.g., one color from another) at an
attendedlocationistypicallyenhancedat?100–110mspoststimu-
lus (Anllo-Vento et al., 1998). The processing of a task-irrelevant
feature at the attended location is typically delayed by ?50 ms
(Schoenfeldetal.,2003).Ageneraldelayduetothetaskirrelevance
persewouldthereforebeexpectedtobereflectedintheelectrophys-
iological recordings ?150–160 ms poststimulus, which is different
from the timing observed in the present study (?200 ms). Task
relevance is certainly a key factor with regard to the timing differ-
ences,butisunlikelytocausetheentiredelayobserved.
Together with other findings from the literature, the current
results point to a more general framework of the temporal dy-
namicsofattentionalspreading.Withinanattendedspatialloca-
tion feature selection can be very fast (?100 ms poststimulus
onset). This is especially the case when an entire feature-
dimension can be selected from another one (e.g., attending a
stimulus’motionvsitscolor;Schoenfeldetal.,2007).Additional
time costs are observed when attentional selection operates
within a single feature-dimension (e.g., selecting one particular
motion direction from another; Hillyard and Mu ¨nte, 1984;
Anllo-Vento and Hillyard, 1996; Karayanidis and Michie, 1996;
Lange et al., 1998). Depending on the stimuli and task used,
attentional selection occurs ?110–160 ms poststimulus. In the
current experiment the probe was located in the spatially unat-
tended visual field. In this case the attentional selection occurred
?50 ms later (?200 ms poststimulus onset) indicating that the
spreadofattentionacrossspatiallocationstakes?50msoftime.
Importantly, not only the spread of attention over space takes
time. Studies on object-based attention could show that the at-
tentional spread from an attended to an unattended feature of
thatsameobjecttook?40–50ms.Inthiscasebothfeatureswere
presentatthesamespatiallocationthatwasattended(Schoenfeld
et al., 2003). A recent study showed that attention not only
spreads across the same objects but also to other objects at spa-
tially unattended locations if they share a task-irrelevant feature
ofanattendedobject(Boehleretal.,2011).Inthiscasetheatten-
tional boost occurs even later, at ?270 ms poststimulus. This
suggeststhatthetemporalcostsfortheattentionspreadfromthe
task-relevant to the task-irrelevant object feature sum up to the
cost for spreading from attended to unattended locations.
Around 150 ms are needed for the selection of the task-relevant
feature of a spatially attended object (Hillyard and Mu ¨nte, 1984;
Anllo-Vento and Hillyard, 1996; Karayanidis and Michie, 1996;
Lange et al., 1998). Another 50–60 ms are needed for spreading
tothetask-irrelevantfeatureofthesameobject(Schoenfeldetal.,
2003). The following spread toward an unattended spatial loca-
tion would take another 50–60 ms (result of the current study).
The resulting total time would be in the range between 250 and
270 ms, which is well in line with the aforementioned findings
(Boehler et al., 2011).
In conclusion, the current results show that feature-based at-
tentionisassociatedwithaglobalprocessingenhancementofthe
attendedfeature(TreueandMartínezTrujillo,1999;Saenzetal.,
2002; Kamitani and Tong, 2006). Neural responses elicited by
spatially unattended task-irrelevant probes were modulated as a
function of the degree of similarity between their movement di-
rectionandthetask-relevantmovementdirectionintheattended
visual field. This provides strong support for the ‘‘feature simi-
larity gain model” at the level of integrated population responses
(Treue and Martínez Trujillo, 1999; Stoppel et al., 2011). Impor-
tantly,theglobalspreadoffeature-basedattentiondoesnotoccur
suddenly following feature selection but rather appears to be a
dynamic time-consuming process.
References
Ahlfors SP, Simpson GV, Dale AM, Belliveau JW, Liu AK, Korvenoja A,
VirtanenJ,HuotilainenM,TootellRB,AronenHJ,IlmoniemiRJ (1999)
Spatiotemporalactivityofacorticalnetworkforprocessingvisualmotion
revealed by MEG and fMRI. J Neurophysiol 82:2545–2555.
Andersen SK, Muller MM, Hillyard SA (2009) Color-selective attention
need not be mediated by spatial attention. J Vis 9:2.1–2.7.
Andersen SK, Fuchs S, Mu ¨ller MM (2011) Effects of feature-selective and
spatial attention at different stages of visual processing. J Cogn Neurosci
23:238–246.
Anllo-Vento L, Hillyard SA (1996) Selective attention to the color and di-
rection of moving stimuli: electrophysiological correlates of hierarchical
feature selection. Percept Psychophys 58:191–206.
Stoppeletal.•SpatiotemporalDynamicsofFeature-BasedAttentionJ.Neurosci.,July11,2012 • 32(28):9671–9676 • 9675

Page 6

Anllo-Vento L, Luck SJ, Hillyard SA (1998) Spatio-temporal dynamics of
attention to color: evidence from human electrophysiology. Hum Brain
Mapp 6:216–238.
Beer AL, Ro ¨der B (2004) Attention to motion enhances processing of both
visual and auditory stimuli: an event-related potential study. Brain Res
Cogn Brain Res 18:205–225.
Beer AL, Ro ¨der B (2005) Attending to visual or auditory motion affects
perceptionwithinandacrossmodalities:anevent-relatedpotentialstudy.
Eur J Neurosci 21:1116–1130.
Bichot NP, Rossi AF, Desimone R (2005) Parallel and serial neural mecha-
nisms for visual search in macaque area V4. Science 308:529–534.
Boehler CN, Schoenfeld MA, Heinze HJ, Hopf JM (2011) Object-based se-
lectionofirrelevantfeaturesisnotconfinedtotheattendedobject.JCogn
Neurosci 23:2231–2239.
Britten KH, Shadlen MN, Newsome WT, Movshon JA (1992) The analysis
of visual motion: a comparison of neuronal and psychophysical perfor-
mance. J Neurosci 12:4745–4765.
Chawla D, Rees G, Friston KJ (1999) The physiological basis of attentional
modulation in extrastriate visual areas. Nat Neurosci 2:671–676.
CookEP,MaunsellJH (2002) Attentionalmodulationofbehavioralperfor-
manceandneuronalresponsesinmiddletemporalandventralintrapari-
etal areas of macaque monkey. J Neurosci 22:1994–2004.
Cook EP, Maunsell JH (2004) Attentional modulation of motion integra-
tion of individual neurons in the middle temporal visual area. J Neurosci
24:7964–7977.
Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Petersen SE (1990) At-
tentionalmodulationofneuralprocessingofshape,color,andvelocityin
humans. Science 248:1556–1559.
Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Petersen SE (1991) Se-
lective and divided attention during visual discriminations of shape,
color, and speed: functional anatomy by positron emission tomography.
J Neurosci 11:2383–2402.
Desimone R, Duncan J (1995) Neural mechanisms of selective visual atten-
tion. Annu Rev Neurosci 18:193–222.
Egeth HE, Yantis S (1997) Visual attention: control, representation, and
time course. Annu Rev Psychol 48:269–297.
Fuchs M, Wagner M, Wischmann HA, Ko ¨hler T, Theissen A, Drenckhahn R,
Buchner H (1998) Improving source reconstructions by combining
bioelectric and biomagnetic data. Electroencephalogr Clin Neurophysiol
107:93–111.
Fuchs M, Wagner M, Ko ¨hler T, Wischmann HA (1999) Linear and nonlin-
ear current density reconstructions. J Clin Neurophysiol 16:267–295.
Guthrie D, Buchwald JS (1991) Significance testing of difference potentials.
Psychophysiology 28:240–244.
Ha ¨ma ¨la ¨inen MS, Ilmoniemi RJ (1994) Interpreting magnetic fields of the
brain: minimum norm estimates. Med Biol Eng Comput 32:35–42.
Ha ¨ma ¨la ¨inen MS, Sarvas J (1989) Realistic conductivity geometry model of
the human head for interpretation of neuromagnetic data. IEEE Trans
Biomed Eng 36:165–171.
Harter MR, Aine C (1984) Brain mechanisms of visual selective attention.
In: Varieties of attention (Parasuraman R, Davies DR, eds), pp 293–321.
New York: Academic.
Heinze HJ, Mangun GR, Burchert W, Hinrichs H, Scholz M, Mu ¨nte TF, Go ¨s
A, Scherg M, Johannes S, Hundeshagen H, Gazzaniga MS, Hillyard SA
(1994) Combined spatial and temporal imaging of brain activity during
visual selective attention in humans. Nature 372:543–546.
Hillyard SA, Anllo-Vento L (1998) Event-related brain potentials in the
study of visual selective attention. Proc Natl Acad Sci U S A 95:781–787.
Hillyard SA, Mu ¨nte TF (1984) Selective attention to color and location:
an analysis with event-related brain potentials. Percept Psychophys
36:185–198.
Kamitani Y, Tong F (2006) Decoding seen and attended motion directions
from activity in the human visual cortex. Curr Biol 16:1096–1102.
Karayanidis F, Michie PT (1996) Frontal processing negativity in a visual
selectiveattentiontask.ElectroencephalogrClinNeurophysiol99:38–56.
Kenemans JL, Kok A, Smulders FT (1993) Event-related potentials to con-
junctions of spatial frequency and orientation as a function of stimulus
parameters and response requirements. Electroencephalogr Clin Neuro-
physiol 88:51–63.
KenemansJL,BaasJM,MangunGR,LijffijtM,VerbatenMN (2000) Onthe
processing of spatial frequencies as revealed by evoked-potential source
modeling. Clin Neurophysiol 111:1113–1123.
Lange JJ, Wijers AA, Mulder LJ, Mulder G (1998) Color selection and loca-
tion selection in ERPs: differences, similarities and ‘neural specificity’.
Biol Psychol 48:153–182.
Martínez A, Di Russo F, Anllo-Vento L, Sereno MI, Buxton RB, Hillyard SA
(2001) Putting spatial attention on the map: timing and localization of
stimulusselectionprocessesinstriateandextrastriatevisualareas.Vision
Res 41:1437–1457.
Martinez-Trujillo JC, Treue S (2004) Feature-based attention increases the
selectivity of population responses in primate visual cortex. Curr Biol
14:744–751.
Motter BC (1994) Neural correlates of attentive selection for color or lumi-
nance in extrastriate area V4. J Neurosci 14:2178–2189.
O’CravenKM,RosenBR,KwongKK,TreismanA,SavoyRL (1997) Volun-
tary attention modulates fMRI activity in human MT-MST. Neuron
18:591–598.
Pascual-Marqui RD (2002) Standardized low-resolution brain electromag-
netic tomography (sLORETA): technical details. Methods Find Exp Clin
Pharmacol 24 [Suppl D]:5–12.
Posner MI (1980) Orienting of attention. Q J Exp Psychol 32:3–25.
Robinson SE (1989) Environmental noise cancellation for biomagnetic
measurements. In: Advances in biomagnetism. (Williamson SJ, Hoke M,
Stroink G, Kotani M, eds), pp 721–724. New York: Plenum.
Saenz M, Buracas GT, Boynton GM (2002) Global effects of feature-based
attention in human visual cortex. Nat Neurosci 5:631–632.
SchoenfeldMA,TempelmannC,MartinezA,HopfJM,SattlerC,HeinzeHJ,
Hillyard SA (2003) Dynamics of feature binding during object-selective
attention. Proc Natl Acad Sci U S A 100:11806–11811.
Schoenfeld MA, Hopf JM, Martinez A, Mai HM, Sattler C, Gasde A, Heinze
HJ, Hillyard SA (2007) Spatio-temporal analysis of feature-based atten-
tion. Cereb Cortex 17:2468–2477.
SerencesJT,BoyntonGM (2007) Feature-basedattentionalmodulationsin
the absence of direct visual stimulation. Neuron 55:301–312.
SerencesJT,YantisS (2006) Selectivevisualattentionandperceptualcoher-
ence. Trends Cogn Sci 10:38–45.
SmidHG,JakobA,HeinzeHJ (1999) Anevent-relatedbrainpotentialstudy
of visual selective attention to conjunctions of color and shape. Psycho-
physiology 36:264–279.
Stoppel CM, Boehler CN, Strumpf H, Heinze HJ, Noesselt T, Hopf JM,
Schoenfeld MA (2011) Feature-based attention modulates direction-
selective hemodynamic activity within human MT. Hum Brain Mapp
32:2183–2192.
Torriente I, Valdes-Sosa M, Ramirez D, Bobes MA (1999) Visual evoked
potentialsrelatedtomotion-onsetaremodulatedbyattention.VisionRes
39:4122–4139.
Treue S, Martínez Trujillo JC (1999) Feature-based attention influences
motion processing gain in macaque visual cortex. Nature 399:575–579.
Treue S, Maunsell JH (1999) Effects of attention on the processing of mo-
tion in macaque middle temporal and medial superior temporal visual
cortical areas. J Neurosci 19:7591–7602.
Zeki S, Watson JD, Lueck CJ, Friston KJ, Kennard C, Frackowiak RS (1991)
A direct demonstration of functional specialization in human visual cor-
tex. J Neurosci 11:641–649.
9676 • J.Neurosci.,July11,2012 • 32(28):9671–9676Stoppeletal.•SpatiotemporalDynamicsofFeature-BasedAttention