Cross-Modal Interactions between Audition Touch andVision in Endogenous Spatial Attention ERP Evidence on

Preparatory States and Sensory Modulations

Martin Eimer1 Jose van Velzen1 and Jon Driver2

Abstract

amp Recent behavioral and event-related brain potential (ERP)studies have revealed cross-modal interactions in endogenousspatial attention between vision and audition plus vision andtouch The present ERP study investigated whether theseinteractions reflect supramodal attentional control mecha-nisms and whether similar cross-modal interactions also existbetween audition and touch Participants directed attention tothe side indicated by a cue to detect infrequent auditory ortactile targets at the cued side The relevant modality (auditionor touch) was blocked Attentional control processes werereflected in systematic ERP modulations elicited during cuedshifts of attention An anterior negativity contralateral to thecued side was followed by a contralateral positivity at posterior

sites These effects were similar whether the cue signaledwhich side was relevant for audition or for touch They alsoresembled previously observed ERP modulations for shifts ofvisual attention thus implicating supramodal mechanisms inthe control of spatial attention Following each cue singleauditory tactile or visual stimuli were presented at the cued oruncued side Although stimuli in task-irrelevant modalitiescould be completely ignored visual and auditory ERPs werenevertheless affected by spatial attention when touch wasrelevant revealing cross-modal interactions When auditionwas relevant visual ERPs but not tactile ERPs were affected byspatial attention indicating that touch can be decoupled fromcross-modal attention when task-irrelevant amp

INTRODUCTION

In everyday life attention is often directed to informationcoming from different sensory modalities but from thesame location in space When trying to follow a conver-sation in a noisy room we may attend to seen lip move-ments as well as to the speakerrsquos voice coming from thesame location When exploring an object manually wemay attend both to its visual appearance and to thetactile information produced at the same location Ingeneral adaptive control of behavior requires the inte-gration and coordination of information originating fromdifferent input modalities but from corresponding spa-tial locations This may involve spatial synergies in selec-tive attention across different sensory modalities

Selective attention has traditionally been studied sep-arately for different sensory modalities with little directcontact between traditional research on lsquolsquovisual atten-tionrsquorsquo lsquolsquoauditory attentionrsquorsquo or lsquolsquotactile attentionrsquorsquo Theissue of cross-modal interactions in spatial attentionbetween vision audition and touch has been addressedonly more recently (see Driver amp Spence 1998 forreview) A few studies have also begun to examineneural correlates of such interactions using event-re-lated brain potentials (ERP) (eg Eimer amp Schroger

1998 Hillyard Simpson Woods Van Voorhis amp Munte1984) or functional imaging measures (eg MacalusoFrith amp Driver 2000a 2000b) Several recent behavioralstudies (eg Spence amp Driver 1996 Spence Pavani ampDriver 2000) found evidence for cross-modal interac-tions in endogenous (voluntary) spatial attention be-tween vision and audition and vision and touch Inthese experiments participants had to direct theirattention to the expected location of target stimuliwithin one (primary) modality On a few trials stimuliof a different (secondary) modality were presented butwere somewhat more likely on the side opposite theexpected location in the primary modality Superiorperformance for stimuli at the location that was ex-pected for the primary modality was observed not onlyfor that modality but also for stimuli in the secondarymodality thus demonstrating a spatial synergy in atten-tion between the two modalities

Such behavioral results demonstrate that cross-modalinteractions exist between vision audition and touch inendogenous spatial attention However the behavioralresults alone cannot directly address two importantquestions First do such cross-modal interactions reflecta supramodal control system for spatial attention whichdirects endogenous spatial attention together across thedifferent modalities (Farah Wong Monheit amp Morrow1989) or do they instead reflect lsquolsquohorizontalrsquorsquo links1 Bickbeck College London 2University College London

copy 2002 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 142 pp 254- 271

between otherwise separate systems (Spence amp Driver1996) Second what level(s) of processing (eg sensorydecision response selection) are affected by cross-mo-dal interactions in endogenous spatial attention

The present study used ERP measures to investigateboth of these issues In different experimental blocksparticipants were instructed to direct attention to theleft or right side in order to detect infrequent targets onthat side only within the relevant modality This relevantmodality (audition or touch) was blocked The relevantside for that modality was indicated on a trial-by-trialbasis with a symbolic central cue Single auditorytactile or visual stimuli were then presented at the cuedor uncued side Stimuli in the two currently irrelevantmodalities had to be entirely ignored regardless of theirposition To investigate whether control processes in-volved in shifts of spatial attention are supramodal ormodality-specific we measured and compared ERPselicited by cues directing attention to the left or rightin blocks where audition or touch was task-relevant Tostudy whether cross-modal interactions in spatial atten-tion affect sensory-specific perceptual processes or onlylater postperceptual stages we recorded ERPs to stimuliin task-irrelevant modalities at locations that were at-tended or unattended within the currently relevantmodality

ERP Correlates of Attentional Control Processes

Some previous studies have investigated ERP correlates ofattentional control in the interval between a cue stimulusthat indicates the direction of an attentional shift and asubsequent target stimulus However all such studies todate have been unimodal focusing exclusively on pro-cesses underlying the control of visual- spatial attentionA major aim of the present study was to assess whethercued attentional shifts may be controlled by supramodalmechanisms To this end we investigated ERPs elicitedduring cued attentional shifts to one side or the other forauditory and for tactile tasks If preparatory shifts ofspatial attention for visual auditory and tactile tasks allreflect a supramodal control system one could expect tofind similar ERP correlates in the cue- target interval forthese different modalities By contrast if separate modal-ity-specific control systems were involved this should bereflected in systematic differences in the ERPs recordedfor the cue- target interval when the subject prepares for astimulus in a different modality on the cued side

In a pioneering ERP study of the control of spatialattention in a visual task Harter Miller Price LaLondeand Keyes (1989) measured ERPs during leftward versusrightward shifts of visual attention These were trig-gered by central arrow cues that indicated the side ofan upcoming visual event They found an early negativedeflection at posterior electrodes contralateral to thedirection of the induced attentional shift (early-direct-ing attention negativity or EDAN see also Nobre

Sebestyen amp Miniussi 2000 Yamaguchi Tsuchiya ampKobayashi 1994 1995) Subsequent to this an en-hanced contralateral positivity at posterior electrodes(late-directing attention positivity LDAP) was foundduring later phases of the cue- target interval In addi-tion Nobre et al (2000) and Mangun (1994 see alsoHopfinger Jha Hopf Girelli amp Mangun 2000) ob-served enhanced negativities at frontal electrodes con-tralateral to the direction of attentional shifts between300 and 500 msec after onset of the central cue(anterior-directing attention negativity ADAN) All theseeffects (ie the EDAN LDAP and ADAN) were assumedto reflect successive phases in the control of covertvisual- spatial orienting The EDAN has been linked(Harter et al 1989) to the encoding of the spatialinformation provided by the cue andor to the initiationof an attentional shift The ADAN (Nobre et al 2000Mangun 1994) has been linked to the activation offrontal structures involved in the control of spatialattentional shifts within visual space (Corbetta MiezinShulman amp Petersen 1993 Posner amp Petersen 1990)Finally the LDAP has been linked to preparatory mod-ulations in the excitability of sensory areas (Harter et al1989) The present study investigated whether similarERP effects can be observed in the cue- target intervalduring shifts of attention to the location of relevantauditory or tactile events This could be expected if thecontrol of endogenous spatial attention is based onsupramodal mechanisms

ERP Effects of Cross-Modal Interactions inEndogenous Spatial Attention on SensoryProcessing

While previous behavioral studies have uncovered theexistence of cross-modal interactions in endogenousspatial attention between vision and audition and be-tween vision and touch (Spence amp Driver 1996 Spenceet al 2000) recent electrophysiological studies haveprovided additional insights into the processing stagesaffected by such cross-modal interactions Eimer andSchroger (1998) investigated ERP effects of cross-modalattention for vision and audition Attention had to bedirected to the left or right to perform a task within oneprimary modality (vision or audition) while the othermodality could be entirely ignored Effects of spatialattention on sensory-specific N1 components were ob-served not only for the currently relevant primary mo-dality but also for stimuli in the currently irrelevantsecondary modality although these effects were typi-cally somewhat larger in amplitude for the primarymodality (see also Teder-Salejarvi Munte Sperlich ampHillyard 1999 Hillyard et al 1984 for similar results)Analogous ERP results were obtained by Eimer andDriver (2000) in a study of vision and touch with theexception that touch can apparently be lsquolsquodecoupledrsquorsquofrom the direction of endogenous spatial attention in

Eimer van Velzen and Driver 255

vision provided that all tactile events are entirely task-irrelevant throughout a block of trials (see Ward McDo-nald amp Lin 2000 for further arguments about possiblelsquolsquodecouplingrsquorsquo based on behavioral data)

These ERP results suggest that relatively early stages ofvisual auditory and tactile information processing (iestages traditionally considered to be lsquolsquounimodalrsquorsquo) can beaffected by cross-modal interactions between vision andaudition and between vision and touch (see also Ken-nett Eimer Spence amp Driver 2001 McDonald amp Ward2000 for related ERP evidence concerning exogenousrather than endogenous cross-modal spatial attention)This pattern of cross-modal interactions apparently in-fluencing sensory-specific responses suggests that loca-tions may initially be selected for attention at amultimodal level of spatial representation with thisselection then feeding down to influence lsquolsquounimodalrsquorsquosensory processes for incoming stimuli If spatial selec-tion does indeed operate supramodally similar cross-modal interactions should also be found betweenaudition and touch The second major aim of the presentstudy was to examine with ERP measures interactions inendogenous spatial attention between these two modal-ities A final question was whether such interactionsmight be asymmetric such that touch may be lsquolsquode-coupledrsquorsquo from multimodal influences when entirelytask-irrelevant (as suggested by Eimer amp Driverrsquos [2000]visual- tactile ERP study) In addition to studying cross-modal interactions between audition and touch we alsoincluded visual events to confirm and replicate previ-ously observed cross-modal interactions involving vision

The Present Study

Single visual auditory or tactile stimuli were presentedon the left or right side Each was preceded by a centralsymbolic precue that instructs participants to attendcovertly to one side in the relevant modality In theaudition-relevant condition the task was to detect in-frequently presented auditory target stimuli (17 lsquolsquoodd-ballsrsquorsquo which were a double-pulse rather than the usualsingle-pulse stimulus) at the cued side In the touch-relevant condition participants had to detect infrequenttactile targets (ie double pulses) at the cued sideStimuli at uncued locations were to be ignored as wereall irrelevant modality stimuli regardless of their location(vision was irrelevant throughout)

To study preparatory states involved in the control ofcovert attentional shifts we examined ERP waveforms inresponse to cues directing attention to the left or rightside separately for the audition- and touch-relevantconditions In order to measure ERP correlates of atten-tional shifts it is important that the direction of theattentional shift (left vs right) is not confounded withsensory differences in the cue Therefore the presentcentral cues were designed to be physically equivalentacross the different conditions (see also Hopfinger

Buonocore amp Mangun 2000 Nobre et al 2000 forsimilar procedures in unimodal visual studies) Two smalltriangles (one red one blue) were presented near centralfixation to produce the symbolic cue one to the imme-diate left and one to the immediate right of fixation Toobtain a symmetrical cue display triangles alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thedirection to be attended for the task-relevant modalitywas indicated by the direction of pointing for the trianglein one particular color (red or blue) This relevant colorwas counterbalanced across subjects

To investigate effects of cross-modal interactions inspatial attention on the processing of stimuli at attendedand unattended locations we measured ERPs to audi-tory tactile and visual nontarget stimuli This was doneat cued and uncued locations separately for the audi-tion- and vision-relevant conditions Because visual stim-uli were always irrelevant any attentional modulationsof visual ERPs would provide additional evidence for theexistence of cross-modal interactions in endogenousspatial attention either audio- visual (as in Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) or tactile- visual (as in Eimer amp Driver 2000)More importantly by examining tactile ERPs in theaudition-relevant condition and auditory ERPs in thetactile-relevant condition we could investigate auditory-tactile and tactile- auditory interactions for the first timeAttentional modulation of auditory ERPs in the touch-relevant blocks (where auditory stimuli were irrelevant)would reveal tactile- auditory interactions in spatial at-tention A similar logic applies to tactile ERPs in theaudition-relevant blocks

RESULTS

Behavioral Performance

Vocal responses were faster to tactile double-pulsetargets in the touch-relevant condition than to auditorydouble-pulse targets in the audition-relevant condition[548 vs 598 msec F(115) = 156 p lt 001] Responseswere also somewhat faster to target stimuli presented onthe right versus left side [568 vs 579 msec F(115) =126 p lt 003] Participants missed more auditorytargets than tactile targets (43 vs 27) but thisdifference was not significant False alarms to nontargetsoccurred on less than 03 of all trials

ERPs in the Interval Between Central Cue andPeripheral Stimulus Preparatory States

Figure 1 shows ERPs elicited in the interval betweencue onset to 100 msec after the onset of the subse-quent peripheral stimulus These are shown for electro-des over the left and right hemisphere in response tocues directing attention to the left versus right sideThey are displayed separately for the audition- (A) and

256 Journal of Cognitive Neuroscience Volume 14 Number 2

the touch-relevant conditions (B) The presence of ERPlateralizations sensitive to the direction of attentionalshifts can be seen more clearly in the difference wave-forms shown in Figure 2 These difference waves weregenerated in two stages First we subtracted ERPsrecorded during attentional shifts to the right fromERPs elicited during leftward attentional shifts Secondwe subtracted the resulting difference waveforms atright electrodes from the difference waveforms emerg-ing at corresponding electrodes over the left hemi-sphere These double subtractions were conductedsolely to simplify graphical presentation not for formalstatistics (see below) A net negativity contralateral tothe direction of attentional shifts is reflected by positiveamplitude values (downward-going deflections) whilea net positivity at contralateral sites is reflected bynegative values (upward deflections) Figures 1 and 2

suggest that the direction of cued attentional shifts(ie indicating the left or right hemifield as relevant inaudition or touch) had systematic effects on ERPselicited in the interval between central cue and sub-sequent peripheral stimulus Critically as regards ourmultimodal concerns these effects were remarkablysimilar regardless of whether audition or touch wastask-relevant Between about 350 and 500 msec aftercue onset frontocentral ERPs were more negative atcontralateral electrodes (ADAN) while no such effectwas present at posterior sites Starting about 500 msecafter cue onset posterior ERPs became more positiveat contralateral sites (LDAP) whereas no such effectwas visible at anterior electrodes (Figure 2)

Statistical analyses were used to confirm that theADAN and LDAP effects apparent in Figure 2 were indeedrelated to systematically different ERP patterns over the

Cue-Target Interval - Audition Relevant

Attend LeftAttend Right

F7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

-4microV

3microV

800msec

(A)

-4microV

3microVF7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

Attend LeftAttend Right

800msec

Cue-Target Interval - Touch Relevant

(B)

Figure 1 Grand-averaged ERPs elicited at lateral electrodes in the interval between cue onset and 800 msec after cue onset (100 msec after theonset of a peripheral stimulus) relative to a 100-msec baseline in response to central cues directing attention to the left side (solid lines) or cuesdirecting attention to the right side (dashed lines) (A) ERPs elicited in the audition-relevant condition (B) ERPs elicited in the touch-relevantcondition

Eimer van Velzen and Driver 257

left and right hemisphere during leftward and rightwardattentional shifts In an initial analysis ERPs elicitedduring leftward versus rightward attentional shifts werecompared using unpaired t tests for each electrode siteand sampling point throughout the cue- target intervalThis was done separately for the audition- and touch-relevant conditions These analyses revealed no system-

atic ERP modulations related to the direction of atten-tional shifts within the first 350 msec after cue onset (seeFigures 1 and 2) In other words there was no evidencewhatsoever for the presence of a posterior EDAN

Repeated measure ANOVAs were then conducted onERP mean amplitudes elicited at lateral recording sites inthe later cue- target intervals from 350 to 500 msec and

Figure 2 Difference wave-forms obtained at lateralanterior central and posteriorelectrodes in the interval be-tween cue onset and 800 msecafter cue onset (100 msec afterthe onset of a peripheralstimulus) in the audition- (solidlines) and touch-relevant con-ditions (dashed lines) reflectinglateralized ERP modulationssensitive to the direction ofattentional shifts Differencewaveforms were generated byfirst subtracting ERPs in re-sponse to cues directingattention to the right from ERPsin response to cues directingattention to the left and thensubtracting the resultingdifference waves at right elec-trodes from the differencewaveform obtained for thecorresponding left hemisphereelectrode Enlarged negativitiescontralateral to the direction ofattentional shifts are reflectedby positive amplitude valuesand larger positivities at con-tralateral sites are reflected bynegative values (see text)An ADAN at frontocentralcontralateral sites was followedby an LDAP at posteriorcontralateral electrodes forboth task conditions

ERP Lateralizations in the Cue-Target Interval

Negativity

Contralateral

F78 FC56 F34

T78 CP56 C34

P78 OLR P34

-2microV

2microV

800msec

ADAN

Audition Relevant Touch Relevant

Positivity

Negativity

Contralateral

Positivity

Negativity

Contralateral

Positivity

LDAP

258 Journal of Cognitive Neuroscience Volume 14 Number 2

from 500 to 700 msec with the factors of cued direction(left vs right) recording side (left vs right) and taskcondition (audition- vs touch-relevant) This was imple-mented separately for anterior central and posteriorelectrode pairs In the 350- 500-msec interval an anteriornegativity contralateral to the direction of attentionalshifts (ADAN) was reflected in highly significant Record-ing Side pound Cued Direction interactions at anterior sites[F(115) = 569 p lt 001] Further analyses revealedsignificant Recording Side pound Cued Direction interactionsin the audition- and touch-relevant condition at allanterior sites as well as at C34 and T78 but not atCP56 or any posterior site To assess whether an ADANwas elicited over both hemispheres the effects of cuedirection were analyzed separately for left and for rightfrontocentral electrodes A reliable main effect of cueddirection was present over the left hemisphere [F(115)= 61 p lt 026] and this effect was nearly significantover the right hemisphere [F(115) = 43 p lt 055]Importantly no three-way interactions (task condition byrecording side by cued direction) were obtained indicat-ing that the ADAN was elicited similarly in the tactile- andaudition-relevant conditions (see Figure 2) This wasconfirmed by the fact that reliable Frontal RecordingSide pound Cued Direction interactions were present forboth task conditions when considered separately [bothF(115) gt 274 both p lt 001]

In the 500- 700-msec interval the LDAP was reflectedin a highly significant Recording Side pound Cued Directioninteraction at posterior sites [F(115) = 236 p lt 001]When analyzed separately for left and right posteriorelectrodes main effects of cued direction were foundin both analyses [both F(115) gt 108 both p lt 005]This indicates that an LDAP was present over the left aswell as over the right hemisphere Subsequent analysesrevealed highly significant Recording Side pound CuedDirection interactions for both the audition- and thetouch-relevant conditions [both F(115) gt 201 bothp lt 001] This demonstrates that an LDAP was elicitedboth when the relevant side was cued for audition andwhen it was cued for touch with a very similar patternbeing apparent for both modalities (see Figure 2) Areliable LDAP was also elicited at CP56 [F(115) = 80p lt 013] but not at T78 and C34 or at any frontalelectrode

ERPs in Response to Subsequent PeripheralStimuli Attentional Modulation of SensoryResponses

Auditory ERPs

Figure 3 shows ERPs elicited by auditory stimuli at cuedand uncued locations when audition was task-relevant(A) or when touch was relevant instead (B) The corre-sponding cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure 5

(left) As expected spatial attention resulted in en-hanced negativities for auditory stimuli presented atcued relative to uncued locations This effect startedabout 160 msec poststimulus overlapping with thedescending flank of the auditory N1 component Con-sequently no effects of spatial attention were obtainedin an analysis window centered on the peak latency ofthe N1 (120- 160 msec poststimulus) Between 160 and200 msec the main effects of spatial attention werepresent at lateral anterior central and posterior electro-des as well as at midline sites [all F(115) gt 53 all p lt037] Most importantly there was no sign of any spatialattention by relevant modality interactions (all Fs lt 1)indicating that attentional negativities for auditory ERPswere elicited not only in the audition-relevant conditionbut also for the touch-relevant condition (see Figure 3Band Figure 5 left) That is modulation of auditory ERPsby spatial attention were found even when audition wasentirely task-irrelevant with the central cue indicatingwhich side should be attended for a tactile task insteadIn analyses conducted separately for just the touch-relevant condition effects of spatial attention on audi-tory ERPs were obtained at lateral central and posteriorsites as well as at midline electrodes [all F(115) gt 55all p lt 033] and this effect approached significance atanterior electrodes [F(115) = 40 p lt 064] Thusattentional negativities were elicited for auditory ERPseven when attention was cued to one side for touchthus revealing cross-modal tactile- auditory interactionsin spatial attention

In a later time window between 200 and 280 msecpoststimulus attentional negativities were present atlateral frontal and central electrodes as well as at midlinesites [all F(115) gt 72 all p lt 017] but not at posteriorelectrodes At lateral central electrodes plus Fz and Czspatial attention affected auditory ERPs in the audition-relevant condition [all Fs(115) gt 47 all p lt 045] butnot in the tactile-relevant condition (F lt 1) At anteriorelectrodes reliable attentional negativities were elicitedin the audition-relevant condition [F(115) = 65 p lt022] This effect approached significance in the touch-relevant condition [F(115) = 39 p lt 067]

As the difference waveforms in Figure 5 (left) suggestthat additional attentional negativities were elicited atfrontal electrodes beyond 300 msec poststimulus fur-ther analyses were carried out on ERP mean amplitudesmeasured between 300 and 380 msec after auditorystimulus onset Effects of spatial attention were foundat frontal recording sites in the audition-relevant as wellas in the touch-relevant condition [both F(115) gt 50both p lt 041]

Tactile ERPs

Figure 4 shows ERPs elicited at lateral electrodes con-tralateral and ipsilateral to the stimulated hand (as wellas at midline electrodes) by tactile stimuli at cued and

Eimer van Velzen and Driver 259

uncued locations These are shown separately for thetouch- (A) and the audition-relevant conditions (B) Theresulting cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure5 (right) Similarly to the attentional modulations ofauditory ERPs effects of spatial attention on somatosen-sory ERPs started around 160 msec poststimulus Whentouch was relevant enhanced negativities were elicitedby tactile stimuli at cued versus uncued locations andthese effects tended to be largest at sites close tosomatosensory areas (electrodes Cz and C34 see Figure

5 right) In contrast attentional effects on tactile ERPsappear to be absent in the audition-relevant condition

Statistical comparisons confirmed this pattern Be-tween 160 and 200 msec effects of spatial attention ontactile ERPs were present at lateral central electrodes inthe touch-relevant condition [F(115) = 98 p lt 007]but not for the audition-relevant condition (F lt 1) Thisdifference was reflected in a Spatial Attention pound RelevantModality interaction [F(115) = 85 p lt 011] Between200 and 280 msec poststimulus attentional negativitieswere present at midline electrodes and at lateral anterior

Figure 3 Grand-averagedERPs elicited by auditorynontarget stimuli at cued loca-tions (solid lines) and uncuedlocations (dashed lines) in the500-msec interval following sti-mulus onset (A) Audition-rele-vant condition (B) Touch-relevant condition

F7 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

CuedUncued

260 Journal of Cognitive Neuroscience Volume 14 Number 2

(A)

Auditory ERPsmdashAudition Relevant

500msec

6microV

F3

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

between otherwise separate systems (Spence amp Driver1996) Second what level(s) of processing (eg sensorydecision response selection) are affected by cross-mo-dal interactions in endogenous spatial attention

The present study used ERP measures to investigateboth of these issues In different experimental blocksparticipants were instructed to direct attention to theleft or right side in order to detect infrequent targets onthat side only within the relevant modality This relevantmodality (audition or touch) was blocked The relevantside for that modality was indicated on a trial-by-trialbasis with a symbolic central cue Single auditorytactile or visual stimuli were then presented at the cuedor uncued side Stimuli in the two currently irrelevantmodalities had to be entirely ignored regardless of theirposition To investigate whether control processes in-volved in shifts of spatial attention are supramodal ormodality-specific we measured and compared ERPselicited by cues directing attention to the left or rightin blocks where audition or touch was task-relevant Tostudy whether cross-modal interactions in spatial atten-tion affect sensory-specific perceptual processes or onlylater postperceptual stages we recorded ERPs to stimuliin task-irrelevant modalities at locations that were at-tended or unattended within the currently relevantmodality

ERP Correlates of Attentional Control Processes

Some previous studies have investigated ERP correlates ofattentional control in the interval between a cue stimulusthat indicates the direction of an attentional shift and asubsequent target stimulus However all such studies todate have been unimodal focusing exclusively on pro-cesses underlying the control of visual- spatial attentionA major aim of the present study was to assess whethercued attentional shifts may be controlled by supramodalmechanisms To this end we investigated ERPs elicitedduring cued attentional shifts to one side or the other forauditory and for tactile tasks If preparatory shifts ofspatial attention for visual auditory and tactile tasks allreflect a supramodal control system one could expect tofind similar ERP correlates in the cue- target interval forthese different modalities By contrast if separate modal-ity-specific control systems were involved this should bereflected in systematic differences in the ERPs recordedfor the cue- target interval when the subject prepares for astimulus in a different modality on the cued side

In a pioneering ERP study of the control of spatialattention in a visual task Harter Miller Price LaLondeand Keyes (1989) measured ERPs during leftward versusrightward shifts of visual attention These were trig-gered by central arrow cues that indicated the side ofan upcoming visual event They found an early negativedeflection at posterior electrodes contralateral to thedirection of the induced attentional shift (early-direct-ing attention negativity or EDAN see also Nobre

Sebestyen amp Miniussi 2000 Yamaguchi Tsuchiya ampKobayashi 1994 1995) Subsequent to this an en-hanced contralateral positivity at posterior electrodes(late-directing attention positivity LDAP) was foundduring later phases of the cue- target interval In addi-tion Nobre et al (2000) and Mangun (1994 see alsoHopfinger Jha Hopf Girelli amp Mangun 2000) ob-served enhanced negativities at frontal electrodes con-tralateral to the direction of attentional shifts between300 and 500 msec after onset of the central cue(anterior-directing attention negativity ADAN) All theseeffects (ie the EDAN LDAP and ADAN) were assumedto reflect successive phases in the control of covertvisual- spatial orienting The EDAN has been linked(Harter et al 1989) to the encoding of the spatialinformation provided by the cue andor to the initiationof an attentional shift The ADAN (Nobre et al 2000Mangun 1994) has been linked to the activation offrontal structures involved in the control of spatialattentional shifts within visual space (Corbetta MiezinShulman amp Petersen 1993 Posner amp Petersen 1990)Finally the LDAP has been linked to preparatory mod-ulations in the excitability of sensory areas (Harter et al1989) The present study investigated whether similarERP effects can be observed in the cue- target intervalduring shifts of attention to the location of relevantauditory or tactile events This could be expected if thecontrol of endogenous spatial attention is based onsupramodal mechanisms

ERP Effects of Cross-Modal Interactions inEndogenous Spatial Attention on SensoryProcessing

While previous behavioral studies have uncovered theexistence of cross-modal interactions in endogenousspatial attention between vision and audition and be-tween vision and touch (Spence amp Driver 1996 Spenceet al 2000) recent electrophysiological studies haveprovided additional insights into the processing stagesaffected by such cross-modal interactions Eimer andSchroger (1998) investigated ERP effects of cross-modalattention for vision and audition Attention had to bedirected to the left or right to perform a task within oneprimary modality (vision or audition) while the othermodality could be entirely ignored Effects of spatialattention on sensory-specific N1 components were ob-served not only for the currently relevant primary mo-dality but also for stimuli in the currently irrelevantsecondary modality although these effects were typi-cally somewhat larger in amplitude for the primarymodality (see also Teder-Salejarvi Munte Sperlich ampHillyard 1999 Hillyard et al 1984 for similar results)Analogous ERP results were obtained by Eimer andDriver (2000) in a study of vision and touch with theexception that touch can apparently be lsquolsquodecoupledrsquorsquofrom the direction of endogenous spatial attention in

Eimer van Velzen and Driver 255

vision provided that all tactile events are entirely task-irrelevant throughout a block of trials (see Ward McDo-nald amp Lin 2000 for further arguments about possiblelsquolsquodecouplingrsquorsquo based on behavioral data)

These ERP results suggest that relatively early stages ofvisual auditory and tactile information processing (iestages traditionally considered to be lsquolsquounimodalrsquorsquo) can beaffected by cross-modal interactions between vision andaudition and between vision and touch (see also Ken-nett Eimer Spence amp Driver 2001 McDonald amp Ward2000 for related ERP evidence concerning exogenousrather than endogenous cross-modal spatial attention)This pattern of cross-modal interactions apparently in-fluencing sensory-specific responses suggests that loca-tions may initially be selected for attention at amultimodal level of spatial representation with thisselection then feeding down to influence lsquolsquounimodalrsquorsquosensory processes for incoming stimuli If spatial selec-tion does indeed operate supramodally similar cross-modal interactions should also be found betweenaudition and touch The second major aim of the presentstudy was to examine with ERP measures interactions inendogenous spatial attention between these two modal-ities A final question was whether such interactionsmight be asymmetric such that touch may be lsquolsquode-coupledrsquorsquo from multimodal influences when entirelytask-irrelevant (as suggested by Eimer amp Driverrsquos [2000]visual- tactile ERP study) In addition to studying cross-modal interactions between audition and touch we alsoincluded visual events to confirm and replicate previ-ously observed cross-modal interactions involving vision

The Present Study

Single visual auditory or tactile stimuli were presentedon the left or right side Each was preceded by a centralsymbolic precue that instructs participants to attendcovertly to one side in the relevant modality In theaudition-relevant condition the task was to detect in-frequently presented auditory target stimuli (17 lsquolsquoodd-ballsrsquorsquo which were a double-pulse rather than the usualsingle-pulse stimulus) at the cued side In the touch-relevant condition participants had to detect infrequenttactile targets (ie double pulses) at the cued sideStimuli at uncued locations were to be ignored as wereall irrelevant modality stimuli regardless of their location(vision was irrelevant throughout)

To study preparatory states involved in the control ofcovert attentional shifts we examined ERP waveforms inresponse to cues directing attention to the left or rightside separately for the audition- and touch-relevantconditions In order to measure ERP correlates of atten-tional shifts it is important that the direction of theattentional shift (left vs right) is not confounded withsensory differences in the cue Therefore the presentcentral cues were designed to be physically equivalentacross the different conditions (see also Hopfinger

Buonocore amp Mangun 2000 Nobre et al 2000 forsimilar procedures in unimodal visual studies) Two smalltriangles (one red one blue) were presented near centralfixation to produce the symbolic cue one to the imme-diate left and one to the immediate right of fixation Toobtain a symmetrical cue display triangles alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thedirection to be attended for the task-relevant modalitywas indicated by the direction of pointing for the trianglein one particular color (red or blue) This relevant colorwas counterbalanced across subjects

To investigate effects of cross-modal interactions inspatial attention on the processing of stimuli at attendedand unattended locations we measured ERPs to audi-tory tactile and visual nontarget stimuli This was doneat cued and uncued locations separately for the audi-tion- and vision-relevant conditions Because visual stim-uli were always irrelevant any attentional modulationsof visual ERPs would provide additional evidence for theexistence of cross-modal interactions in endogenousspatial attention either audio- visual (as in Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) or tactile- visual (as in Eimer amp Driver 2000)More importantly by examining tactile ERPs in theaudition-relevant condition and auditory ERPs in thetactile-relevant condition we could investigate auditory-tactile and tactile- auditory interactions for the first timeAttentional modulation of auditory ERPs in the touch-relevant blocks (where auditory stimuli were irrelevant)would reveal tactile- auditory interactions in spatial at-tention A similar logic applies to tactile ERPs in theaudition-relevant blocks

RESULTS

Behavioral Performance

Vocal responses were faster to tactile double-pulsetargets in the touch-relevant condition than to auditorydouble-pulse targets in the audition-relevant condition[548 vs 598 msec F(115) = 156 p lt 001] Responseswere also somewhat faster to target stimuli presented onthe right versus left side [568 vs 579 msec F(115) =126 p lt 003] Participants missed more auditorytargets than tactile targets (43 vs 27) but thisdifference was not significant False alarms to nontargetsoccurred on less than 03 of all trials

ERPs in the Interval Between Central Cue andPeripheral Stimulus Preparatory States

Figure 1 shows ERPs elicited in the interval betweencue onset to 100 msec after the onset of the subse-quent peripheral stimulus These are shown for electro-des over the left and right hemisphere in response tocues directing attention to the left versus right sideThey are displayed separately for the audition- (A) and

256 Journal of Cognitive Neuroscience Volume 14 Number 2

the touch-relevant conditions (B) The presence of ERPlateralizations sensitive to the direction of attentionalshifts can be seen more clearly in the difference wave-forms shown in Figure 2 These difference waves weregenerated in two stages First we subtracted ERPsrecorded during attentional shifts to the right fromERPs elicited during leftward attentional shifts Secondwe subtracted the resulting difference waveforms atright electrodes from the difference waveforms emerg-ing at corresponding electrodes over the left hemi-sphere These double subtractions were conductedsolely to simplify graphical presentation not for formalstatistics (see below) A net negativity contralateral tothe direction of attentional shifts is reflected by positiveamplitude values (downward-going deflections) whilea net positivity at contralateral sites is reflected bynegative values (upward deflections) Figures 1 and 2

suggest that the direction of cued attentional shifts(ie indicating the left or right hemifield as relevant inaudition or touch) had systematic effects on ERPselicited in the interval between central cue and sub-sequent peripheral stimulus Critically as regards ourmultimodal concerns these effects were remarkablysimilar regardless of whether audition or touch wastask-relevant Between about 350 and 500 msec aftercue onset frontocentral ERPs were more negative atcontralateral electrodes (ADAN) while no such effectwas present at posterior sites Starting about 500 msecafter cue onset posterior ERPs became more positiveat contralateral sites (LDAP) whereas no such effectwas visible at anterior electrodes (Figure 2)

Statistical analyses were used to confirm that theADAN and LDAP effects apparent in Figure 2 were indeedrelated to systematically different ERP patterns over the

Cue-Target Interval - Audition Relevant

Attend LeftAttend Right

F7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

-4microV

3microV

800msec

(A)

-4microV

3microVF7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

Attend LeftAttend Right

800msec

Cue-Target Interval - Touch Relevant

(B)

Figure 1 Grand-averaged ERPs elicited at lateral electrodes in the interval between cue onset and 800 msec after cue onset (100 msec after theonset of a peripheral stimulus) relative to a 100-msec baseline in response to central cues directing attention to the left side (solid lines) or cuesdirecting attention to the right side (dashed lines) (A) ERPs elicited in the audition-relevant condition (B) ERPs elicited in the touch-relevantcondition

Eimer van Velzen and Driver 257

left and right hemisphere during leftward and rightwardattentional shifts In an initial analysis ERPs elicitedduring leftward versus rightward attentional shifts werecompared using unpaired t tests for each electrode siteand sampling point throughout the cue- target intervalThis was done separately for the audition- and touch-relevant conditions These analyses revealed no system-

atic ERP modulations related to the direction of atten-tional shifts within the first 350 msec after cue onset (seeFigures 1 and 2) In other words there was no evidencewhatsoever for the presence of a posterior EDAN

Repeated measure ANOVAs were then conducted onERP mean amplitudes elicited at lateral recording sites inthe later cue- target intervals from 350 to 500 msec and

Figure 2 Difference wave-forms obtained at lateralanterior central and posteriorelectrodes in the interval be-tween cue onset and 800 msecafter cue onset (100 msec afterthe onset of a peripheralstimulus) in the audition- (solidlines) and touch-relevant con-ditions (dashed lines) reflectinglateralized ERP modulationssensitive to the direction ofattentional shifts Differencewaveforms were generated byfirst subtracting ERPs in re-sponse to cues directingattention to the right from ERPsin response to cues directingattention to the left and thensubtracting the resultingdifference waves at right elec-trodes from the differencewaveform obtained for thecorresponding left hemisphereelectrode Enlarged negativitiescontralateral to the direction ofattentional shifts are reflectedby positive amplitude valuesand larger positivities at con-tralateral sites are reflected bynegative values (see text)An ADAN at frontocentralcontralateral sites was followedby an LDAP at posteriorcontralateral electrodes forboth task conditions

ERP Lateralizations in the Cue-Target Interval

Negativity

Contralateral

F78 FC56 F34

T78 CP56 C34

P78 OLR P34

-2microV

2microV

800msec

ADAN

Audition Relevant Touch Relevant

Positivity

Negativity

Contralateral

Positivity

Negativity

Contralateral

Positivity

LDAP

258 Journal of Cognitive Neuroscience Volume 14 Number 2

from 500 to 700 msec with the factors of cued direction(left vs right) recording side (left vs right) and taskcondition (audition- vs touch-relevant) This was imple-mented separately for anterior central and posteriorelectrode pairs In the 350- 500-msec interval an anteriornegativity contralateral to the direction of attentionalshifts (ADAN) was reflected in highly significant Record-ing Side pound Cued Direction interactions at anterior sites[F(115) = 569 p lt 001] Further analyses revealedsignificant Recording Side pound Cued Direction interactionsin the audition- and touch-relevant condition at allanterior sites as well as at C34 and T78 but not atCP56 or any posterior site To assess whether an ADANwas elicited over both hemispheres the effects of cuedirection were analyzed separately for left and for rightfrontocentral electrodes A reliable main effect of cueddirection was present over the left hemisphere [F(115)= 61 p lt 026] and this effect was nearly significantover the right hemisphere [F(115) = 43 p lt 055]Importantly no three-way interactions (task condition byrecording side by cued direction) were obtained indicat-ing that the ADAN was elicited similarly in the tactile- andaudition-relevant conditions (see Figure 2) This wasconfirmed by the fact that reliable Frontal RecordingSide pound Cued Direction interactions were present forboth task conditions when considered separately [bothF(115) gt 274 both p lt 001]

In the 500- 700-msec interval the LDAP was reflectedin a highly significant Recording Side pound Cued Directioninteraction at posterior sites [F(115) = 236 p lt 001]When analyzed separately for left and right posteriorelectrodes main effects of cued direction were foundin both analyses [both F(115) gt 108 both p lt 005]This indicates that an LDAP was present over the left aswell as over the right hemisphere Subsequent analysesrevealed highly significant Recording Side pound CuedDirection interactions for both the audition- and thetouch-relevant conditions [both F(115) gt 201 bothp lt 001] This demonstrates that an LDAP was elicitedboth when the relevant side was cued for audition andwhen it was cued for touch with a very similar patternbeing apparent for both modalities (see Figure 2) Areliable LDAP was also elicited at CP56 [F(115) = 80p lt 013] but not at T78 and C34 or at any frontalelectrode

ERPs in Response to Subsequent PeripheralStimuli Attentional Modulation of SensoryResponses

Auditory ERPs

Figure 3 shows ERPs elicited by auditory stimuli at cuedand uncued locations when audition was task-relevant(A) or when touch was relevant instead (B) The corre-sponding cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure 5

(left) As expected spatial attention resulted in en-hanced negativities for auditory stimuli presented atcued relative to uncued locations This effect startedabout 160 msec poststimulus overlapping with thedescending flank of the auditory N1 component Con-sequently no effects of spatial attention were obtainedin an analysis window centered on the peak latency ofthe N1 (120- 160 msec poststimulus) Between 160 and200 msec the main effects of spatial attention werepresent at lateral anterior central and posterior electro-des as well as at midline sites [all F(115) gt 53 all p lt037] Most importantly there was no sign of any spatialattention by relevant modality interactions (all Fs lt 1)indicating that attentional negativities for auditory ERPswere elicited not only in the audition-relevant conditionbut also for the touch-relevant condition (see Figure 3Band Figure 5 left) That is modulation of auditory ERPsby spatial attention were found even when audition wasentirely task-irrelevant with the central cue indicatingwhich side should be attended for a tactile task insteadIn analyses conducted separately for just the touch-relevant condition effects of spatial attention on audi-tory ERPs were obtained at lateral central and posteriorsites as well as at midline electrodes [all F(115) gt 55all p lt 033] and this effect approached significance atanterior electrodes [F(115) = 40 p lt 064] Thusattentional negativities were elicited for auditory ERPseven when attention was cued to one side for touchthus revealing cross-modal tactile- auditory interactionsin spatial attention

In a later time window between 200 and 280 msecpoststimulus attentional negativities were present atlateral frontal and central electrodes as well as at midlinesites [all F(115) gt 72 all p lt 017] but not at posteriorelectrodes At lateral central electrodes plus Fz and Czspatial attention affected auditory ERPs in the audition-relevant condition [all Fs(115) gt 47 all p lt 045] butnot in the tactile-relevant condition (F lt 1) At anteriorelectrodes reliable attentional negativities were elicitedin the audition-relevant condition [F(115) = 65 p lt022] This effect approached significance in the touch-relevant condition [F(115) = 39 p lt 067]

As the difference waveforms in Figure 5 (left) suggestthat additional attentional negativities were elicited atfrontal electrodes beyond 300 msec poststimulus fur-ther analyses were carried out on ERP mean amplitudesmeasured between 300 and 380 msec after auditorystimulus onset Effects of spatial attention were foundat frontal recording sites in the audition-relevant as wellas in the touch-relevant condition [both F(115) gt 50both p lt 041]

Tactile ERPs

Figure 4 shows ERPs elicited at lateral electrodes con-tralateral and ipsilateral to the stimulated hand (as wellas at midline electrodes) by tactile stimuli at cued and

Eimer van Velzen and Driver 259

uncued locations These are shown separately for thetouch- (A) and the audition-relevant conditions (B) Theresulting cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure5 (right) Similarly to the attentional modulations ofauditory ERPs effects of spatial attention on somatosen-sory ERPs started around 160 msec poststimulus Whentouch was relevant enhanced negativities were elicitedby tactile stimuli at cued versus uncued locations andthese effects tended to be largest at sites close tosomatosensory areas (electrodes Cz and C34 see Figure

5 right) In contrast attentional effects on tactile ERPsappear to be absent in the audition-relevant condition

Statistical comparisons confirmed this pattern Be-tween 160 and 200 msec effects of spatial attention ontactile ERPs were present at lateral central electrodes inthe touch-relevant condition [F(115) = 98 p lt 007]but not for the audition-relevant condition (F lt 1) Thisdifference was reflected in a Spatial Attention pound RelevantModality interaction [F(115) = 85 p lt 011] Between200 and 280 msec poststimulus attentional negativitieswere present at midline electrodes and at lateral anterior

Figure 3 Grand-averagedERPs elicited by auditorynontarget stimuli at cued loca-tions (solid lines) and uncuedlocations (dashed lines) in the500-msec interval following sti-mulus onset (A) Audition-rele-vant condition (B) Touch-relevant condition

F7 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

CuedUncued

260 Journal of Cognitive Neuroscience Volume 14 Number 2

(A)

Auditory ERPsmdashAudition Relevant

500msec

6microV

F3

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

vision provided that all tactile events are entirely task-irrelevant throughout a block of trials (see Ward McDo-nald amp Lin 2000 for further arguments about possiblelsquolsquodecouplingrsquorsquo based on behavioral data)

These ERP results suggest that relatively early stages ofvisual auditory and tactile information processing (iestages traditionally considered to be lsquolsquounimodalrsquorsquo) can beaffected by cross-modal interactions between vision andaudition and between vision and touch (see also Ken-nett Eimer Spence amp Driver 2001 McDonald amp Ward2000 for related ERP evidence concerning exogenousrather than endogenous cross-modal spatial attention)This pattern of cross-modal interactions apparently in-fluencing sensory-specific responses suggests that loca-tions may initially be selected for attention at amultimodal level of spatial representation with thisselection then feeding down to influence lsquolsquounimodalrsquorsquosensory processes for incoming stimuli If spatial selec-tion does indeed operate supramodally similar cross-modal interactions should also be found betweenaudition and touch The second major aim of the presentstudy was to examine with ERP measures interactions inendogenous spatial attention between these two modal-ities A final question was whether such interactionsmight be asymmetric such that touch may be lsquolsquode-coupledrsquorsquo from multimodal influences when entirelytask-irrelevant (as suggested by Eimer amp Driverrsquos [2000]visual- tactile ERP study) In addition to studying cross-modal interactions between audition and touch we alsoincluded visual events to confirm and replicate previ-ously observed cross-modal interactions involving vision

The Present Study

Single visual auditory or tactile stimuli were presentedon the left or right side Each was preceded by a centralsymbolic precue that instructs participants to attendcovertly to one side in the relevant modality In theaudition-relevant condition the task was to detect in-frequently presented auditory target stimuli (17 lsquolsquoodd-ballsrsquorsquo which were a double-pulse rather than the usualsingle-pulse stimulus) at the cued side In the touch-relevant condition participants had to detect infrequenttactile targets (ie double pulses) at the cued sideStimuli at uncued locations were to be ignored as wereall irrelevant modality stimuli regardless of their location(vision was irrelevant throughout)

To study preparatory states involved in the control ofcovert attentional shifts we examined ERP waveforms inresponse to cues directing attention to the left or rightside separately for the audition- and touch-relevantconditions In order to measure ERP correlates of atten-tional shifts it is important that the direction of theattentional shift (left vs right) is not confounded withsensory differences in the cue Therefore the presentcentral cues were designed to be physically equivalentacross the different conditions (see also Hopfinger

Buonocore amp Mangun 2000 Nobre et al 2000 forsimilar procedures in unimodal visual studies) Two smalltriangles (one red one blue) were presented near centralfixation to produce the symbolic cue one to the imme-diate left and one to the immediate right of fixation Toobtain a symmetrical cue display triangles alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thedirection to be attended for the task-relevant modalitywas indicated by the direction of pointing for the trianglein one particular color (red or blue) This relevant colorwas counterbalanced across subjects

To investigate effects of cross-modal interactions inspatial attention on the processing of stimuli at attendedand unattended locations we measured ERPs to audi-tory tactile and visual nontarget stimuli This was doneat cued and uncued locations separately for the audi-tion- and vision-relevant conditions Because visual stim-uli were always irrelevant any attentional modulationsof visual ERPs would provide additional evidence for theexistence of cross-modal interactions in endogenousspatial attention either audio- visual (as in Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) or tactile- visual (as in Eimer amp Driver 2000)More importantly by examining tactile ERPs in theaudition-relevant condition and auditory ERPs in thetactile-relevant condition we could investigate auditory-tactile and tactile- auditory interactions for the first timeAttentional modulation of auditory ERPs in the touch-relevant blocks (where auditory stimuli were irrelevant)would reveal tactile- auditory interactions in spatial at-tention A similar logic applies to tactile ERPs in theaudition-relevant blocks

RESULTS

Behavioral Performance

Vocal responses were faster to tactile double-pulsetargets in the touch-relevant condition than to auditorydouble-pulse targets in the audition-relevant condition[548 vs 598 msec F(115) = 156 p lt 001] Responseswere also somewhat faster to target stimuli presented onthe right versus left side [568 vs 579 msec F(115) =126 p lt 003] Participants missed more auditorytargets than tactile targets (43 vs 27) but thisdifference was not significant False alarms to nontargetsoccurred on less than 03 of all trials

ERPs in the Interval Between Central Cue andPeripheral Stimulus Preparatory States

Figure 1 shows ERPs elicited in the interval betweencue onset to 100 msec after the onset of the subse-quent peripheral stimulus These are shown for electro-des over the left and right hemisphere in response tocues directing attention to the left versus right sideThey are displayed separately for the audition- (A) and

256 Journal of Cognitive Neuroscience Volume 14 Number 2

the touch-relevant conditions (B) The presence of ERPlateralizations sensitive to the direction of attentionalshifts can be seen more clearly in the difference wave-forms shown in Figure 2 These difference waves weregenerated in two stages First we subtracted ERPsrecorded during attentional shifts to the right fromERPs elicited during leftward attentional shifts Secondwe subtracted the resulting difference waveforms atright electrodes from the difference waveforms emerg-ing at corresponding electrodes over the left hemi-sphere These double subtractions were conductedsolely to simplify graphical presentation not for formalstatistics (see below) A net negativity contralateral tothe direction of attentional shifts is reflected by positiveamplitude values (downward-going deflections) whilea net positivity at contralateral sites is reflected bynegative values (upward deflections) Figures 1 and 2

suggest that the direction of cued attentional shifts(ie indicating the left or right hemifield as relevant inaudition or touch) had systematic effects on ERPselicited in the interval between central cue and sub-sequent peripheral stimulus Critically as regards ourmultimodal concerns these effects were remarkablysimilar regardless of whether audition or touch wastask-relevant Between about 350 and 500 msec aftercue onset frontocentral ERPs were more negative atcontralateral electrodes (ADAN) while no such effectwas present at posterior sites Starting about 500 msecafter cue onset posterior ERPs became more positiveat contralateral sites (LDAP) whereas no such effectwas visible at anterior electrodes (Figure 2)

Statistical analyses were used to confirm that theADAN and LDAP effects apparent in Figure 2 were indeedrelated to systematically different ERP patterns over the

Cue-Target Interval - Audition Relevant

Attend LeftAttend Right

F7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

-4microV

3microV

800msec

(A)

-4microV

3microVF7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

Attend LeftAttend Right

800msec

Cue-Target Interval - Touch Relevant

(B)

Figure 1 Grand-averaged ERPs elicited at lateral electrodes in the interval between cue onset and 800 msec after cue onset (100 msec after theonset of a peripheral stimulus) relative to a 100-msec baseline in response to central cues directing attention to the left side (solid lines) or cuesdirecting attention to the right side (dashed lines) (A) ERPs elicited in the audition-relevant condition (B) ERPs elicited in the touch-relevantcondition

Eimer van Velzen and Driver 257

left and right hemisphere during leftward and rightwardattentional shifts In an initial analysis ERPs elicitedduring leftward versus rightward attentional shifts werecompared using unpaired t tests for each electrode siteand sampling point throughout the cue- target intervalThis was done separately for the audition- and touch-relevant conditions These analyses revealed no system-

atic ERP modulations related to the direction of atten-tional shifts within the first 350 msec after cue onset (seeFigures 1 and 2) In other words there was no evidencewhatsoever for the presence of a posterior EDAN

Repeated measure ANOVAs were then conducted onERP mean amplitudes elicited at lateral recording sites inthe later cue- target intervals from 350 to 500 msec and

Figure 2 Difference wave-forms obtained at lateralanterior central and posteriorelectrodes in the interval be-tween cue onset and 800 msecafter cue onset (100 msec afterthe onset of a peripheralstimulus) in the audition- (solidlines) and touch-relevant con-ditions (dashed lines) reflectinglateralized ERP modulationssensitive to the direction ofattentional shifts Differencewaveforms were generated byfirst subtracting ERPs in re-sponse to cues directingattention to the right from ERPsin response to cues directingattention to the left and thensubtracting the resultingdifference waves at right elec-trodes from the differencewaveform obtained for thecorresponding left hemisphereelectrode Enlarged negativitiescontralateral to the direction ofattentional shifts are reflectedby positive amplitude valuesand larger positivities at con-tralateral sites are reflected bynegative values (see text)An ADAN at frontocentralcontralateral sites was followedby an LDAP at posteriorcontralateral electrodes forboth task conditions

ERP Lateralizations in the Cue-Target Interval

Negativity

Contralateral

F78 FC56 F34

T78 CP56 C34

P78 OLR P34

-2microV

2microV

800msec

ADAN

Audition Relevant Touch Relevant

Positivity

Negativity

Contralateral

Positivity

Negativity

Contralateral

Positivity

LDAP

258 Journal of Cognitive Neuroscience Volume 14 Number 2

from 500 to 700 msec with the factors of cued direction(left vs right) recording side (left vs right) and taskcondition (audition- vs touch-relevant) This was imple-mented separately for anterior central and posteriorelectrode pairs In the 350- 500-msec interval an anteriornegativity contralateral to the direction of attentionalshifts (ADAN) was reflected in highly significant Record-ing Side pound Cued Direction interactions at anterior sites[F(115) = 569 p lt 001] Further analyses revealedsignificant Recording Side pound Cued Direction interactionsin the audition- and touch-relevant condition at allanterior sites as well as at C34 and T78 but not atCP56 or any posterior site To assess whether an ADANwas elicited over both hemispheres the effects of cuedirection were analyzed separately for left and for rightfrontocentral electrodes A reliable main effect of cueddirection was present over the left hemisphere [F(115)= 61 p lt 026] and this effect was nearly significantover the right hemisphere [F(115) = 43 p lt 055]Importantly no three-way interactions (task condition byrecording side by cued direction) were obtained indicat-ing that the ADAN was elicited similarly in the tactile- andaudition-relevant conditions (see Figure 2) This wasconfirmed by the fact that reliable Frontal RecordingSide pound Cued Direction interactions were present forboth task conditions when considered separately [bothF(115) gt 274 both p lt 001]

In the 500- 700-msec interval the LDAP was reflectedin a highly significant Recording Side pound Cued Directioninteraction at posterior sites [F(115) = 236 p lt 001]When analyzed separately for left and right posteriorelectrodes main effects of cued direction were foundin both analyses [both F(115) gt 108 both p lt 005]This indicates that an LDAP was present over the left aswell as over the right hemisphere Subsequent analysesrevealed highly significant Recording Side pound CuedDirection interactions for both the audition- and thetouch-relevant conditions [both F(115) gt 201 bothp lt 001] This demonstrates that an LDAP was elicitedboth when the relevant side was cued for audition andwhen it was cued for touch with a very similar patternbeing apparent for both modalities (see Figure 2) Areliable LDAP was also elicited at CP56 [F(115) = 80p lt 013] but not at T78 and C34 or at any frontalelectrode

ERPs in Response to Subsequent PeripheralStimuli Attentional Modulation of SensoryResponses

Auditory ERPs

Figure 3 shows ERPs elicited by auditory stimuli at cuedand uncued locations when audition was task-relevant(A) or when touch was relevant instead (B) The corre-sponding cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure 5

(left) As expected spatial attention resulted in en-hanced negativities for auditory stimuli presented atcued relative to uncued locations This effect startedabout 160 msec poststimulus overlapping with thedescending flank of the auditory N1 component Con-sequently no effects of spatial attention were obtainedin an analysis window centered on the peak latency ofthe N1 (120- 160 msec poststimulus) Between 160 and200 msec the main effects of spatial attention werepresent at lateral anterior central and posterior electro-des as well as at midline sites [all F(115) gt 53 all p lt037] Most importantly there was no sign of any spatialattention by relevant modality interactions (all Fs lt 1)indicating that attentional negativities for auditory ERPswere elicited not only in the audition-relevant conditionbut also for the touch-relevant condition (see Figure 3Band Figure 5 left) That is modulation of auditory ERPsby spatial attention were found even when audition wasentirely task-irrelevant with the central cue indicatingwhich side should be attended for a tactile task insteadIn analyses conducted separately for just the touch-relevant condition effects of spatial attention on audi-tory ERPs were obtained at lateral central and posteriorsites as well as at midline electrodes [all F(115) gt 55all p lt 033] and this effect approached significance atanterior electrodes [F(115) = 40 p lt 064] Thusattentional negativities were elicited for auditory ERPseven when attention was cued to one side for touchthus revealing cross-modal tactile- auditory interactionsin spatial attention

In a later time window between 200 and 280 msecpoststimulus attentional negativities were present atlateral frontal and central electrodes as well as at midlinesites [all F(115) gt 72 all p lt 017] but not at posteriorelectrodes At lateral central electrodes plus Fz and Czspatial attention affected auditory ERPs in the audition-relevant condition [all Fs(115) gt 47 all p lt 045] butnot in the tactile-relevant condition (F lt 1) At anteriorelectrodes reliable attentional negativities were elicitedin the audition-relevant condition [F(115) = 65 p lt022] This effect approached significance in the touch-relevant condition [F(115) = 39 p lt 067]

As the difference waveforms in Figure 5 (left) suggestthat additional attentional negativities were elicited atfrontal electrodes beyond 300 msec poststimulus fur-ther analyses were carried out on ERP mean amplitudesmeasured between 300 and 380 msec after auditorystimulus onset Effects of spatial attention were foundat frontal recording sites in the audition-relevant as wellas in the touch-relevant condition [both F(115) gt 50both p lt 041]

Tactile ERPs

Figure 4 shows ERPs elicited at lateral electrodes con-tralateral and ipsilateral to the stimulated hand (as wellas at midline electrodes) by tactile stimuli at cued and

Eimer van Velzen and Driver 259

uncued locations These are shown separately for thetouch- (A) and the audition-relevant conditions (B) Theresulting cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure5 (right) Similarly to the attentional modulations ofauditory ERPs effects of spatial attention on somatosen-sory ERPs started around 160 msec poststimulus Whentouch was relevant enhanced negativities were elicitedby tactile stimuli at cued versus uncued locations andthese effects tended to be largest at sites close tosomatosensory areas (electrodes Cz and C34 see Figure

5 right) In contrast attentional effects on tactile ERPsappear to be absent in the audition-relevant condition

Statistical comparisons confirmed this pattern Be-tween 160 and 200 msec effects of spatial attention ontactile ERPs were present at lateral central electrodes inthe touch-relevant condition [F(115) = 98 p lt 007]but not for the audition-relevant condition (F lt 1) Thisdifference was reflected in a Spatial Attention pound RelevantModality interaction [F(115) = 85 p lt 011] Between200 and 280 msec poststimulus attentional negativitieswere present at midline electrodes and at lateral anterior

Figure 3 Grand-averagedERPs elicited by auditorynontarget stimuli at cued loca-tions (solid lines) and uncuedlocations (dashed lines) in the500-msec interval following sti-mulus onset (A) Audition-rele-vant condition (B) Touch-relevant condition

F7 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

CuedUncued

260 Journal of Cognitive Neuroscience Volume 14 Number 2

(A)

Auditory ERPsmdashAudition Relevant

500msec

6microV

F3

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

the touch-relevant conditions (B) The presence of ERPlateralizations sensitive to the direction of attentionalshifts can be seen more clearly in the difference wave-forms shown in Figure 2 These difference waves weregenerated in two stages First we subtracted ERPsrecorded during attentional shifts to the right fromERPs elicited during leftward attentional shifts Secondwe subtracted the resulting difference waveforms atright electrodes from the difference waveforms emerg-ing at corresponding electrodes over the left hemi-sphere These double subtractions were conductedsolely to simplify graphical presentation not for formalstatistics (see below) A net negativity contralateral tothe direction of attentional shifts is reflected by positiveamplitude values (downward-going deflections) whilea net positivity at contralateral sites is reflected bynegative values (upward deflections) Figures 1 and 2

suggest that the direction of cued attentional shifts(ie indicating the left or right hemifield as relevant inaudition or touch) had systematic effects on ERPselicited in the interval between central cue and sub-sequent peripheral stimulus Critically as regards ourmultimodal concerns these effects were remarkablysimilar regardless of whether audition or touch wastask-relevant Between about 350 and 500 msec aftercue onset frontocentral ERPs were more negative atcontralateral electrodes (ADAN) while no such effectwas present at posterior sites Starting about 500 msecafter cue onset posterior ERPs became more positiveat contralateral sites (LDAP) whereas no such effectwas visible at anterior electrodes (Figure 2)

Statistical analyses were used to confirm that theADAN and LDAP effects apparent in Figure 2 were indeedrelated to systematically different ERP patterns over the

Cue-Target Interval - Audition Relevant

Attend LeftAttend Right

F7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

-4microV

3microV

800msec

(A)

-4microV

3microVF7 F3 F4 F8

FC5 FC6

T7 C3 C4 T8

CP5 CP6

P7 P3 P4 P8

OL OR

Attend LeftAttend Right

800msec

Cue-Target Interval - Touch Relevant

(B)

Figure 1 Grand-averaged ERPs elicited at lateral electrodes in the interval between cue onset and 800 msec after cue onset (100 msec after theonset of a peripheral stimulus) relative to a 100-msec baseline in response to central cues directing attention to the left side (solid lines) or cuesdirecting attention to the right side (dashed lines) (A) ERPs elicited in the audition-relevant condition (B) ERPs elicited in the touch-relevantcondition

Eimer van Velzen and Driver 257

left and right hemisphere during leftward and rightwardattentional shifts In an initial analysis ERPs elicitedduring leftward versus rightward attentional shifts werecompared using unpaired t tests for each electrode siteand sampling point throughout the cue- target intervalThis was done separately for the audition- and touch-relevant conditions These analyses revealed no system-

atic ERP modulations related to the direction of atten-tional shifts within the first 350 msec after cue onset (seeFigures 1 and 2) In other words there was no evidencewhatsoever for the presence of a posterior EDAN

Repeated measure ANOVAs were then conducted onERP mean amplitudes elicited at lateral recording sites inthe later cue- target intervals from 350 to 500 msec and

Figure 2 Difference wave-forms obtained at lateralanterior central and posteriorelectrodes in the interval be-tween cue onset and 800 msecafter cue onset (100 msec afterthe onset of a peripheralstimulus) in the audition- (solidlines) and touch-relevant con-ditions (dashed lines) reflectinglateralized ERP modulationssensitive to the direction ofattentional shifts Differencewaveforms were generated byfirst subtracting ERPs in re-sponse to cues directingattention to the right from ERPsin response to cues directingattention to the left and thensubtracting the resultingdifference waves at right elec-trodes from the differencewaveform obtained for thecorresponding left hemisphereelectrode Enlarged negativitiescontralateral to the direction ofattentional shifts are reflectedby positive amplitude valuesand larger positivities at con-tralateral sites are reflected bynegative values (see text)An ADAN at frontocentralcontralateral sites was followedby an LDAP at posteriorcontralateral electrodes forboth task conditions

ERP Lateralizations in the Cue-Target Interval

Negativity

Contralateral

F78 FC56 F34

T78 CP56 C34

P78 OLR P34

-2microV

2microV

800msec

ADAN

Audition Relevant Touch Relevant

Positivity

Negativity

Contralateral

Positivity

Negativity

Contralateral

Positivity

LDAP

258 Journal of Cognitive Neuroscience Volume 14 Number 2

from 500 to 700 msec with the factors of cued direction(left vs right) recording side (left vs right) and taskcondition (audition- vs touch-relevant) This was imple-mented separately for anterior central and posteriorelectrode pairs In the 350- 500-msec interval an anteriornegativity contralateral to the direction of attentionalshifts (ADAN) was reflected in highly significant Record-ing Side pound Cued Direction interactions at anterior sites[F(115) = 569 p lt 001] Further analyses revealedsignificant Recording Side pound Cued Direction interactionsin the audition- and touch-relevant condition at allanterior sites as well as at C34 and T78 but not atCP56 or any posterior site To assess whether an ADANwas elicited over both hemispheres the effects of cuedirection were analyzed separately for left and for rightfrontocentral electrodes A reliable main effect of cueddirection was present over the left hemisphere [F(115)= 61 p lt 026] and this effect was nearly significantover the right hemisphere [F(115) = 43 p lt 055]Importantly no three-way interactions (task condition byrecording side by cued direction) were obtained indicat-ing that the ADAN was elicited similarly in the tactile- andaudition-relevant conditions (see Figure 2) This wasconfirmed by the fact that reliable Frontal RecordingSide pound Cued Direction interactions were present forboth task conditions when considered separately [bothF(115) gt 274 both p lt 001]

In the 500- 700-msec interval the LDAP was reflectedin a highly significant Recording Side pound Cued Directioninteraction at posterior sites [F(115) = 236 p lt 001]When analyzed separately for left and right posteriorelectrodes main effects of cued direction were foundin both analyses [both F(115) gt 108 both p lt 005]This indicates that an LDAP was present over the left aswell as over the right hemisphere Subsequent analysesrevealed highly significant Recording Side pound CuedDirection interactions for both the audition- and thetouch-relevant conditions [both F(115) gt 201 bothp lt 001] This demonstrates that an LDAP was elicitedboth when the relevant side was cued for audition andwhen it was cued for touch with a very similar patternbeing apparent for both modalities (see Figure 2) Areliable LDAP was also elicited at CP56 [F(115) = 80p lt 013] but not at T78 and C34 or at any frontalelectrode

ERPs in Response to Subsequent PeripheralStimuli Attentional Modulation of SensoryResponses

Auditory ERPs

Figure 3 shows ERPs elicited by auditory stimuli at cuedand uncued locations when audition was task-relevant(A) or when touch was relevant instead (B) The corre-sponding cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure 5

(left) As expected spatial attention resulted in en-hanced negativities for auditory stimuli presented atcued relative to uncued locations This effect startedabout 160 msec poststimulus overlapping with thedescending flank of the auditory N1 component Con-sequently no effects of spatial attention were obtainedin an analysis window centered on the peak latency ofthe N1 (120- 160 msec poststimulus) Between 160 and200 msec the main effects of spatial attention werepresent at lateral anterior central and posterior electro-des as well as at midline sites [all F(115) gt 53 all p lt037] Most importantly there was no sign of any spatialattention by relevant modality interactions (all Fs lt 1)indicating that attentional negativities for auditory ERPswere elicited not only in the audition-relevant conditionbut also for the touch-relevant condition (see Figure 3Band Figure 5 left) That is modulation of auditory ERPsby spatial attention were found even when audition wasentirely task-irrelevant with the central cue indicatingwhich side should be attended for a tactile task insteadIn analyses conducted separately for just the touch-relevant condition effects of spatial attention on audi-tory ERPs were obtained at lateral central and posteriorsites as well as at midline electrodes [all F(115) gt 55all p lt 033] and this effect approached significance atanterior electrodes [F(115) = 40 p lt 064] Thusattentional negativities were elicited for auditory ERPseven when attention was cued to one side for touchthus revealing cross-modal tactile- auditory interactionsin spatial attention

In a later time window between 200 and 280 msecpoststimulus attentional negativities were present atlateral frontal and central electrodes as well as at midlinesites [all F(115) gt 72 all p lt 017] but not at posteriorelectrodes At lateral central electrodes plus Fz and Czspatial attention affected auditory ERPs in the audition-relevant condition [all Fs(115) gt 47 all p lt 045] butnot in the tactile-relevant condition (F lt 1) At anteriorelectrodes reliable attentional negativities were elicitedin the audition-relevant condition [F(115) = 65 p lt022] This effect approached significance in the touch-relevant condition [F(115) = 39 p lt 067]

As the difference waveforms in Figure 5 (left) suggestthat additional attentional negativities were elicited atfrontal electrodes beyond 300 msec poststimulus fur-ther analyses were carried out on ERP mean amplitudesmeasured between 300 and 380 msec after auditorystimulus onset Effects of spatial attention were foundat frontal recording sites in the audition-relevant as wellas in the touch-relevant condition [both F(115) gt 50both p lt 041]

Tactile ERPs

Figure 4 shows ERPs elicited at lateral electrodes con-tralateral and ipsilateral to the stimulated hand (as wellas at midline electrodes) by tactile stimuli at cued and

Eimer van Velzen and Driver 259

uncued locations These are shown separately for thetouch- (A) and the audition-relevant conditions (B) Theresulting cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure5 (right) Similarly to the attentional modulations ofauditory ERPs effects of spatial attention on somatosen-sory ERPs started around 160 msec poststimulus Whentouch was relevant enhanced negativities were elicitedby tactile stimuli at cued versus uncued locations andthese effects tended to be largest at sites close tosomatosensory areas (electrodes Cz and C34 see Figure

5 right) In contrast attentional effects on tactile ERPsappear to be absent in the audition-relevant condition

Statistical comparisons confirmed this pattern Be-tween 160 and 200 msec effects of spatial attention ontactile ERPs were present at lateral central electrodes inthe touch-relevant condition [F(115) = 98 p lt 007]but not for the audition-relevant condition (F lt 1) Thisdifference was reflected in a Spatial Attention pound RelevantModality interaction [F(115) = 85 p lt 011] Between200 and 280 msec poststimulus attentional negativitieswere present at midline electrodes and at lateral anterior

Figure 3 Grand-averagedERPs elicited by auditorynontarget stimuli at cued loca-tions (solid lines) and uncuedlocations (dashed lines) in the500-msec interval following sti-mulus onset (A) Audition-rele-vant condition (B) Touch-relevant condition

F7 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

CuedUncued

260 Journal of Cognitive Neuroscience Volume 14 Number 2

(A)

Auditory ERPsmdashAudition Relevant

500msec

6microV

F3

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

left and right hemisphere during leftward and rightwardattentional shifts In an initial analysis ERPs elicitedduring leftward versus rightward attentional shifts werecompared using unpaired t tests for each electrode siteand sampling point throughout the cue- target intervalThis was done separately for the audition- and touch-relevant conditions These analyses revealed no system-

atic ERP modulations related to the direction of atten-tional shifts within the first 350 msec after cue onset (seeFigures 1 and 2) In other words there was no evidencewhatsoever for the presence of a posterior EDAN

Repeated measure ANOVAs were then conducted onERP mean amplitudes elicited at lateral recording sites inthe later cue- target intervals from 350 to 500 msec and

Figure 2 Difference wave-forms obtained at lateralanterior central and posteriorelectrodes in the interval be-tween cue onset and 800 msecafter cue onset (100 msec afterthe onset of a peripheralstimulus) in the audition- (solidlines) and touch-relevant con-ditions (dashed lines) reflectinglateralized ERP modulationssensitive to the direction ofattentional shifts Differencewaveforms were generated byfirst subtracting ERPs in re-sponse to cues directingattention to the right from ERPsin response to cues directingattention to the left and thensubtracting the resultingdifference waves at right elec-trodes from the differencewaveform obtained for thecorresponding left hemisphereelectrode Enlarged negativitiescontralateral to the direction ofattentional shifts are reflectedby positive amplitude valuesand larger positivities at con-tralateral sites are reflected bynegative values (see text)An ADAN at frontocentralcontralateral sites was followedby an LDAP at posteriorcontralateral electrodes forboth task conditions

ERP Lateralizations in the Cue-Target Interval

Negativity

Contralateral

F78 FC56 F34

T78 CP56 C34

P78 OLR P34

-2microV

2microV

800msec

ADAN

Audition Relevant Touch Relevant

Positivity

Negativity

Contralateral

Positivity

Negativity

Contralateral

Positivity

LDAP

258 Journal of Cognitive Neuroscience Volume 14 Number 2

from 500 to 700 msec with the factors of cued direction(left vs right) recording side (left vs right) and taskcondition (audition- vs touch-relevant) This was imple-mented separately for anterior central and posteriorelectrode pairs In the 350- 500-msec interval an anteriornegativity contralateral to the direction of attentionalshifts (ADAN) was reflected in highly significant Record-ing Side pound Cued Direction interactions at anterior sites[F(115) = 569 p lt 001] Further analyses revealedsignificant Recording Side pound Cued Direction interactionsin the audition- and touch-relevant condition at allanterior sites as well as at C34 and T78 but not atCP56 or any posterior site To assess whether an ADANwas elicited over both hemispheres the effects of cuedirection were analyzed separately for left and for rightfrontocentral electrodes A reliable main effect of cueddirection was present over the left hemisphere [F(115)= 61 p lt 026] and this effect was nearly significantover the right hemisphere [F(115) = 43 p lt 055]Importantly no three-way interactions (task condition byrecording side by cued direction) were obtained indicat-ing that the ADAN was elicited similarly in the tactile- andaudition-relevant conditions (see Figure 2) This wasconfirmed by the fact that reliable Frontal RecordingSide pound Cued Direction interactions were present forboth task conditions when considered separately [bothF(115) gt 274 both p lt 001]

In the 500- 700-msec interval the LDAP was reflectedin a highly significant Recording Side pound Cued Directioninteraction at posterior sites [F(115) = 236 p lt 001]When analyzed separately for left and right posteriorelectrodes main effects of cued direction were foundin both analyses [both F(115) gt 108 both p lt 005]This indicates that an LDAP was present over the left aswell as over the right hemisphere Subsequent analysesrevealed highly significant Recording Side pound CuedDirection interactions for both the audition- and thetouch-relevant conditions [both F(115) gt 201 bothp lt 001] This demonstrates that an LDAP was elicitedboth when the relevant side was cued for audition andwhen it was cued for touch with a very similar patternbeing apparent for both modalities (see Figure 2) Areliable LDAP was also elicited at CP56 [F(115) = 80p lt 013] but not at T78 and C34 or at any frontalelectrode

ERPs in Response to Subsequent PeripheralStimuli Attentional Modulation of SensoryResponses

Auditory ERPs

Figure 3 shows ERPs elicited by auditory stimuli at cuedand uncued locations when audition was task-relevant(A) or when touch was relevant instead (B) The corre-sponding cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure 5

(left) As expected spatial attention resulted in en-hanced negativities for auditory stimuli presented atcued relative to uncued locations This effect startedabout 160 msec poststimulus overlapping with thedescending flank of the auditory N1 component Con-sequently no effects of spatial attention were obtainedin an analysis window centered on the peak latency ofthe N1 (120- 160 msec poststimulus) Between 160 and200 msec the main effects of spatial attention werepresent at lateral anterior central and posterior electro-des as well as at midline sites [all F(115) gt 53 all p lt037] Most importantly there was no sign of any spatialattention by relevant modality interactions (all Fs lt 1)indicating that attentional negativities for auditory ERPswere elicited not only in the audition-relevant conditionbut also for the touch-relevant condition (see Figure 3Band Figure 5 left) That is modulation of auditory ERPsby spatial attention were found even when audition wasentirely task-irrelevant with the central cue indicatingwhich side should be attended for a tactile task insteadIn analyses conducted separately for just the touch-relevant condition effects of spatial attention on audi-tory ERPs were obtained at lateral central and posteriorsites as well as at midline electrodes [all F(115) gt 55all p lt 033] and this effect approached significance atanterior electrodes [F(115) = 40 p lt 064] Thusattentional negativities were elicited for auditory ERPseven when attention was cued to one side for touchthus revealing cross-modal tactile- auditory interactionsin spatial attention

In a later time window between 200 and 280 msecpoststimulus attentional negativities were present atlateral frontal and central electrodes as well as at midlinesites [all F(115) gt 72 all p lt 017] but not at posteriorelectrodes At lateral central electrodes plus Fz and Czspatial attention affected auditory ERPs in the audition-relevant condition [all Fs(115) gt 47 all p lt 045] butnot in the tactile-relevant condition (F lt 1) At anteriorelectrodes reliable attentional negativities were elicitedin the audition-relevant condition [F(115) = 65 p lt022] This effect approached significance in the touch-relevant condition [F(115) = 39 p lt 067]

As the difference waveforms in Figure 5 (left) suggestthat additional attentional negativities were elicited atfrontal electrodes beyond 300 msec poststimulus fur-ther analyses were carried out on ERP mean amplitudesmeasured between 300 and 380 msec after auditorystimulus onset Effects of spatial attention were foundat frontal recording sites in the audition-relevant as wellas in the touch-relevant condition [both F(115) gt 50both p lt 041]

Tactile ERPs

Figure 4 shows ERPs elicited at lateral electrodes con-tralateral and ipsilateral to the stimulated hand (as wellas at midline electrodes) by tactile stimuli at cued and

Eimer van Velzen and Driver 259

uncued locations These are shown separately for thetouch- (A) and the audition-relevant conditions (B) Theresulting cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure5 (right) Similarly to the attentional modulations ofauditory ERPs effects of spatial attention on somatosen-sory ERPs started around 160 msec poststimulus Whentouch was relevant enhanced negativities were elicitedby tactile stimuli at cued versus uncued locations andthese effects tended to be largest at sites close tosomatosensory areas (electrodes Cz and C34 see Figure

5 right) In contrast attentional effects on tactile ERPsappear to be absent in the audition-relevant condition

Statistical comparisons confirmed this pattern Be-tween 160 and 200 msec effects of spatial attention ontactile ERPs were present at lateral central electrodes inthe touch-relevant condition [F(115) = 98 p lt 007]but not for the audition-relevant condition (F lt 1) Thisdifference was reflected in a Spatial Attention pound RelevantModality interaction [F(115) = 85 p lt 011] Between200 and 280 msec poststimulus attentional negativitieswere present at midline electrodes and at lateral anterior

Figure 3 Grand-averagedERPs elicited by auditorynontarget stimuli at cued loca-tions (solid lines) and uncuedlocations (dashed lines) in the500-msec interval following sti-mulus onset (A) Audition-rele-vant condition (B) Touch-relevant condition

F7 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

CuedUncued

260 Journal of Cognitive Neuroscience Volume 14 Number 2

(A)

Auditory ERPsmdashAudition Relevant

500msec

6microV

F3

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

from 500 to 700 msec with the factors of cued direction(left vs right) recording side (left vs right) and taskcondition (audition- vs touch-relevant) This was imple-mented separately for anterior central and posteriorelectrode pairs In the 350- 500-msec interval an anteriornegativity contralateral to the direction of attentionalshifts (ADAN) was reflected in highly significant Record-ing Side pound Cued Direction interactions at anterior sites[F(115) = 569 p lt 001] Further analyses revealedsignificant Recording Side pound Cued Direction interactionsin the audition- and touch-relevant condition at allanterior sites as well as at C34 and T78 but not atCP56 or any posterior site To assess whether an ADANwas elicited over both hemispheres the effects of cuedirection were analyzed separately for left and for rightfrontocentral electrodes A reliable main effect of cueddirection was present over the left hemisphere [F(115)= 61 p lt 026] and this effect was nearly significantover the right hemisphere [F(115) = 43 p lt 055]Importantly no three-way interactions (task condition byrecording side by cued direction) were obtained indicat-ing that the ADAN was elicited similarly in the tactile- andaudition-relevant conditions (see Figure 2) This wasconfirmed by the fact that reliable Frontal RecordingSide pound Cued Direction interactions were present forboth task conditions when considered separately [bothF(115) gt 274 both p lt 001]

In the 500- 700-msec interval the LDAP was reflectedin a highly significant Recording Side pound Cued Directioninteraction at posterior sites [F(115) = 236 p lt 001]When analyzed separately for left and right posteriorelectrodes main effects of cued direction were foundin both analyses [both F(115) gt 108 both p lt 005]This indicates that an LDAP was present over the left aswell as over the right hemisphere Subsequent analysesrevealed highly significant Recording Side pound CuedDirection interactions for both the audition- and thetouch-relevant conditions [both F(115) gt 201 bothp lt 001] This demonstrates that an LDAP was elicitedboth when the relevant side was cued for audition andwhen it was cued for touch with a very similar patternbeing apparent for both modalities (see Figure 2) Areliable LDAP was also elicited at CP56 [F(115) = 80p lt 013] but not at T78 and C34 or at any frontalelectrode

ERPs in Response to Subsequent PeripheralStimuli Attentional Modulation of SensoryResponses

Auditory ERPs

Figure 3 shows ERPs elicited by auditory stimuli at cuedand uncued locations when audition was task-relevant(A) or when touch was relevant instead (B) The corre-sponding cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure 5

(left) As expected spatial attention resulted in en-hanced negativities for auditory stimuli presented atcued relative to uncued locations This effect startedabout 160 msec poststimulus overlapping with thedescending flank of the auditory N1 component Con-sequently no effects of spatial attention were obtainedin an analysis window centered on the peak latency ofthe N1 (120- 160 msec poststimulus) Between 160 and200 msec the main effects of spatial attention werepresent at lateral anterior central and posterior electro-des as well as at midline sites [all F(115) gt 53 all p lt037] Most importantly there was no sign of any spatialattention by relevant modality interactions (all Fs lt 1)indicating that attentional negativities for auditory ERPswere elicited not only in the audition-relevant conditionbut also for the touch-relevant condition (see Figure 3Band Figure 5 left) That is modulation of auditory ERPsby spatial attention were found even when audition wasentirely task-irrelevant with the central cue indicatingwhich side should be attended for a tactile task insteadIn analyses conducted separately for just the touch-relevant condition effects of spatial attention on audi-tory ERPs were obtained at lateral central and posteriorsites as well as at midline electrodes [all F(115) gt 55all p lt 033] and this effect approached significance atanterior electrodes [F(115) = 40 p lt 064] Thusattentional negativities were elicited for auditory ERPseven when attention was cued to one side for touchthus revealing cross-modal tactile- auditory interactionsin spatial attention

In a later time window between 200 and 280 msecpoststimulus attentional negativities were present atlateral frontal and central electrodes as well as at midlinesites [all F(115) gt 72 all p lt 017] but not at posteriorelectrodes At lateral central electrodes plus Fz and Czspatial attention affected auditory ERPs in the audition-relevant condition [all Fs(115) gt 47 all p lt 045] butnot in the tactile-relevant condition (F lt 1) At anteriorelectrodes reliable attentional negativities were elicitedin the audition-relevant condition [F(115) = 65 p lt022] This effect approached significance in the touch-relevant condition [F(115) = 39 p lt 067]

As the difference waveforms in Figure 5 (left) suggestthat additional attentional negativities were elicited atfrontal electrodes beyond 300 msec poststimulus fur-ther analyses were carried out on ERP mean amplitudesmeasured between 300 and 380 msec after auditorystimulus onset Effects of spatial attention were foundat frontal recording sites in the audition-relevant as wellas in the touch-relevant condition [both F(115) gt 50both p lt 041]

Tactile ERPs

Figure 4 shows ERPs elicited at lateral electrodes con-tralateral and ipsilateral to the stimulated hand (as wellas at midline electrodes) by tactile stimuli at cued and

Eimer van Velzen and Driver 259

uncued locations These are shown separately for thetouch- (A) and the audition-relevant conditions (B) Theresulting cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure5 (right) Similarly to the attentional modulations ofauditory ERPs effects of spatial attention on somatosen-sory ERPs started around 160 msec poststimulus Whentouch was relevant enhanced negativities were elicitedby tactile stimuli at cued versus uncued locations andthese effects tended to be largest at sites close tosomatosensory areas (electrodes Cz and C34 see Figure

5 right) In contrast attentional effects on tactile ERPsappear to be absent in the audition-relevant condition

Statistical comparisons confirmed this pattern Be-tween 160 and 200 msec effects of spatial attention ontactile ERPs were present at lateral central electrodes inthe touch-relevant condition [F(115) = 98 p lt 007]but not for the audition-relevant condition (F lt 1) Thisdifference was reflected in a Spatial Attention pound RelevantModality interaction [F(115) = 85 p lt 011] Between200 and 280 msec poststimulus attentional negativitieswere present at midline electrodes and at lateral anterior

Figure 3 Grand-averagedERPs elicited by auditorynontarget stimuli at cued loca-tions (solid lines) and uncuedlocations (dashed lines) in the500-msec interval following sti-mulus onset (A) Audition-rele-vant condition (B) Touch-relevant condition

F7 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

CuedUncued

260 Journal of Cognitive Neuroscience Volume 14 Number 2

(A)

Auditory ERPsmdashAudition Relevant

500msec

6microV

F3

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

uncued locations These are shown separately for thetouch- (A) and the audition-relevant conditions (B) Theresulting cued-minus-uncued difference waveforms atfrontal central and parietal sites are shown in Figure5 (right) Similarly to the attentional modulations ofauditory ERPs effects of spatial attention on somatosen-sory ERPs started around 160 msec poststimulus Whentouch was relevant enhanced negativities were elicitedby tactile stimuli at cued versus uncued locations andthese effects tended to be largest at sites close tosomatosensory areas (electrodes Cz and C34 see Figure

5 right) In contrast attentional effects on tactile ERPsappear to be absent in the audition-relevant condition

Statistical comparisons confirmed this pattern Be-tween 160 and 200 msec effects of spatial attention ontactile ERPs were present at lateral central electrodes inthe touch-relevant condition [F(115) = 98 p lt 007]but not for the audition-relevant condition (F lt 1) Thisdifference was reflected in a Spatial Attention pound RelevantModality interaction [F(115) = 85 p lt 011] Between200 and 280 msec poststimulus attentional negativitieswere present at midline electrodes and at lateral anterior

Figure 3 Grand-averagedERPs elicited by auditorynontarget stimuli at cued loca-tions (solid lines) and uncuedlocations (dashed lines) in the500-msec interval following sti-mulus onset (A) Audition-rele-vant condition (B) Touch-relevant condition

F7 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

CuedUncued

260 Journal of Cognitive Neuroscience Volume 14 Number 2

(A)

Auditory ERPsmdashAudition Relevant

500msec

6microV

F3

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

and central sites when touch was relevant [all F(115) gt56 all p lt 032] but not when audition was task-relevant instead (all Fs lt 1) Again this difference wasreflected in significant Spatial Attention pound RelevantModality interactions [all F(115) gt 77 all p lt 014]

As the difference waveforms in Figure 5 (right) in-dicate that tactile stimuli at cued versus uncued loca-tions elicited an additional negativity beyond 300 msecpoststimulus again in the touch-relevant conditionsonly further analyses were carried out on ERP meanamplitudes measured between 300 and 380 msec after

tactile stimulus onset Effects of spatial attention werefound for all recording sites in the touch-relevant con-dition [all F(115) gt 108 all p lt 005] whereas no sucheffects were present for the audition-relevant condition(all Fs lt 1)

Thus while auditory ERPs had shown spatial attentioneffects for cued versus uncued locations regardless ofwhether audition or touch was the task-relevant modal-ity by contrast tactile ERPs showed effects of spatialattention only when touch was task-relevant The factthat attending to one side versus the other for audition

F7 F3 Fz F4 F8

FC5 FC6

T7 C3 Cz C4 T8

CP5 CP6

P7 P3 Pz P4 P8

-7microV

6microV

500msec

Auditory ERPsmdashTouch Relevant

CuedUncued

(B)

Figure 3 (continued)

Eimer van Velzen and Driver 261

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

did not affect tactile ERPs is reminiscent of Eimer andDriverrsquos (2000) finding that touch can be lsquolsquodecoupledrsquorsquofrom spatial attention in another modality (vision intheir study but audition here) provided that touch istask-irrelevant throughout an entire block of trials Wediscuss this point later

Visual ERPs

Visual events were always irrelevant to the prescribedauditory or tactile tasks The results obtained for visualERPs in the audition- and touch-relevant conditions con-

firm previous cross-modal observations (Eimer ampSchroger 1998 Eimer amp Driver 2000 Hillyard et al1984) Accordingly they will be summarized only brieflyhere Figure 6 shows visual ERPs at lateral posteriorelectrodes contralateral to the side of visual stimuluspresentation (P34 and OLR) and at midline electrodesCz and Pz for visual stimuli at cued versus uncuedlocations in the audition- (left) and touch-relevant con-ditions (right) There were strong cross-modal influencesof spatial attention in both cases demonstrating audio-visual and tactile- visual interactions respectively A reli-able enhancement of the occipital P1 (90- 130-msec time

Figure 4 Grand-averagedERPs elicited by tactilenontarget stimuli at cuedlocations (solid lines) anduncued locations (dashed lines)in the 500-msec intervalfollowing stimulus onset(A) Touch-relevant condition(B) Audition-relevant conditionERPs elicited at electrodescontralateral to the stimulatedhand are shown on the leftERPs elicited at ipsilateralelectrodes are shown on theright

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashTouch Relevant

-7microV

6microV

500msec

CuedUncued

(A)

262 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

window) by spatial attention was present in the audition-relevant condition ( p lt 048) but not for the touch-relevant condition However modulations of posteriorN1 components (160- 200-msec time window) werefound at all posterior electrodes in the touch-relevantcondition (all ps lt 001) and at P3P4 in the audition-relevant condition also ( p lt 024) As in previous cross-modal studies significant attentional negativities werealso present in the N1 time range for visual ERPs atmidline electrodes (all ps lt 036) when attention was

cued to one side for audition (as in Eimer amp Schroger1998) or touch (as in Eimer amp Driver 2000) (see Figure 6)

DISCUSSION

The aim of the present ERP study was to investigatecross-modal interactions in endogenous spatial atten-tion both in terms of the control processes involvedand their effects on stimulus processing To identifyERP correlates of preparatory attentional control pro-

F78 F34 Fz F34 F78

FC56FC56

T78 C34 Cz C34 C78

CP56 CP56

P78 PzP34 P34 P78

CONTRALATERAL IPSILATERAL

Somatosensory ERPsmdashAudition Relevant

-7microV

6microV

500msec

CuedUncued

(B)

Figure 4 (continued)

Eimer van Velzen and Driver 263

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

cesses we measured ERPs in response to centralsymbolic cues that signaled the relevant location foran auditory or a tactile task In this way we could studywhether ERPs in the cue- target interval were similar ordifferent when shifting attention to the location ofrelevant auditory versus tactile stimuli Any close sim-ilarity for these two cases would implicate multimodalmechanisms in the control of endogenous spatial at-tention as would any similarity to previously studiedvisual cases To measure the effects of cross-modalinteractions in spatial attention on processing withincurrently irrelevant modalities we measured ERPs inresponse to auditory tactile and visual nontargetstimuli at locations that were attended or unattendedfor an auditory or tactile task Any ERP effects of spatialattention within currently irrelevant modalities shouldreflect the impact of attentional interactions betweensensory modalities Previous ERP experiments with thisgeneral logic had reported audio-visual (eg Teder-Salejarvi et al 1999 Eimer amp Schroger 1998 Hillyardet al 1984) and visual- tactile interactions (Eimer ampDriver 2000) Here we examined possible audio-tactileinteractions

Our results provide the first evidence on the multi-modal nature of preparatory attentional control mecha-nisms in the cue- target interval They also confirm andextend previous ERP observations related to the effectsof cross-modal interactions in spatial attention on pro-cessing within currently irrelevant modalities

ERP Evidence on Preparatory Attentional ControlProcesses in the CuendashTarget Interval

To investigate processes involved in the control ofspatial attention we compared ERPs elicited in thecue- target interval for cues directing attention to theleft versus right side separately for the audition- andtouch-relevant conditions Several previous studies haveexamined the ERP correlates of such preparatory atten-tional states but only for visual- spatial orienting (egNobre et al 2000 Mangun 1994 Yamaguchi et al 19941995 Harter et al 1989) If attentional shifts in differentsensory modalities are controlled by a supramodal sys-tem ERP correlates of covert attention shifts in the cue-target interval should be similar regardless of whetherattention is directed to one side for a visual auditory or

Figure 5 Differencewaveforms obtained at frontalcentral and parietal electrodesby subtracting ERPs to stimuli atuncued locations from ERPs tostimuli at cued locations LeftDifference waveforms obtainedfor auditory stimuli in theaudition- (solid lines) andtouch-relevant conditions(dashed lines) RightDifference waveforms obtainedfor tactile stimuli in the touch-(solid lines) and audition-rele-vant condition (dashed lines) atmidline electrode plus atelectrodes contralateral (leftside) or ipsilateral (right side)to the side of the stimulatedhand

F34 Fz F34

C34 Cz C34

P34 Pz P34

Fz

Cz

Pz

F3 F4

C3 C4

P3 P4

-3microV

2microV

500msec

-5microV

2microV

CONTRA IPSI

AUDITION TOUCH

TouchAudition

Relevant Modality

AuditionTouch

Relevant Modality

Difference Waveforms CuedndashUncued Locations

264 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

tactile task By contrast if separate modality-specificcontrol systems were involved there should be system-atic differences between ERPs recorded during shifts ofattention for vision audition or touch

ERP modulations sensitive to the cued direction of acovert attentional shift were strikingly similar in theaudition- and touch-relevant conditions Moreoverthese modulations also resembled those previouslyfound in unimodal visual studies1 Between 350 and500 msec after onset of the central cue a frontocentralnegativity was elicited contralateral to the direction ofthe attentional shift2 This ADAN was relatively small(but highly reliable) so it can be seen most clearly in thedifference waveforms of Figure 2 Lateralized ERP effectsthat are similar to the ADAN in terms of their latency andscalp distribution have previously been reported inunimodal studies of visual- spatial attention (eg Nobreet al 2000 Mangun 1994) In those studies they weretentatively interpreted as reflecting activation of frontalareas involved in the control of visual- spatial attentionThe fact that an ADAN was elicited here in both theaudition- and touch-relevant conditions in a similarmanner to previous visual studies suggests that theprocesses responsible for this effect are not modality-specific The ADAN may thus reflect supramodal controlprocesses possibly within an lsquolsquoanterior attention sys-temrsquorsquo (Posner amp Petersen 1990) that controls spatial

parameters of attentional shifts regardless of sensorymodality In line with such an interpretation physiolog-ical studies have identified cell populations in severalareas of frontal cortex that have a multimodal represen-tation of space (eg Graziano Yap amp Gross 1994)

Between 500 and 700 msec after cue onset a posteriorpositivity was elicited contralateral to the direction ofattentional shifts in both the audition- and the touch-relevant conditions (Figure 1) This LDAP appears asnegative-going deflections at lateral posterior sites in thedifference waveforms of Figure 2 Given the strikingsimilarity in terms of latencies and scalp distributionsthe LDAP observed in the present study almost certainlyreflects the same phenomenon as the LDAP observed inprevious unimodal studies of visual- spatial orienting(eg Harter et al 1989) This similarity of the LDAPfor auditory tactile and visual cases during later phasesof the cue- target interval is illustrated in Figure 7 Scalpdistribution maps of ERP lateralizations are shown forthe 500- 700-msec interval after cue onset These wereobtained by subtracting ERPs in response to cues direct-ing attention to the right side from ERPs to cues direct-ing attention to the left Enhanced positivities atelectrodes contralateral to an attentional shift are indi-cated by negative amplitude values over the left hemi-sphere and positive values over the right hemisphereResults obtained for the audition- and touch-relevant

Figure 6 Grand-averagedERPs elicited at midlineelectrodes Cz and Pz and atlateral parietal (P34) andoccipital (OLR) electrodescontralateral to the side ofpresentation for visual non-target stimuli at cued (solidlines) and uncued locations(dashed lines) in the 500-msec interval followingstimulus onset LeftAudition-relevant conditionRight Touch-relevantcondition

Audition Relevant Touch Relevant-7microV

6microV

Cz Pz

P34 OLR

Cz Pz

P34 OLR

500msec

CuedUncued

Eimer van Velzen and Driver 265

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

condition of the present experiment are shown Inaddition results from an unpublished follow-up studyin our laboratory are also presented Exactly the samecueing procedure was used in this study to directattention to the left or right side in a visual task (ieparticipants now had to detect visual gap targets on thecued side) As can be seen in Figure 7 the overalldistribution pattern of the LDAP is strikingly similar forattentional shifts cued for each of the three modalities

The fact that an LDAP is elicited during cued shifts ofspatial attention for an auditory task and also for atactile task may seem surprising given previous sugges-tions that this component is related to the spatiallyselective activation of modality-specific visual areas (Har-ter et al 1989) One possibility is that instead ofreflecting visual activations the LDAP may reflect supra-modal attentional control processes (as suggested abovefor the ADAN) Given the distinctive posterior scalpdistribution of the LDAP it could conceivably reflectactivity in posterior parietal areas although a moreventral source remains possible The posterior parietalcortex is involved in the control of visual- spatial atten-tion (eg LaBerge 1995) as well as in the integration ofinformation from different sense modalities (eg An-dersen Snyder Bradley amp Young 1997) The fact thatthe frontal ADAN precedes the posterior LDAP couldaccord with Posner and Petersenrsquos (1990) proposals thatanterior circuits control more posterior spatial attentioncircuits (see also Hopfinger Buonocore et al 2000Kastner Pinsk De Weerd Desimone amp Ungerleider1999 Rosen et al 1999 for recent evidence on thisfrom functional imaging albeit only in unimodal visualstudies)

A second possibility is that the LDAP may indeedreflect preparatory activation of modality-specific visualareas (as originally proposed by Harter et al 1989) butthat this can arise even for auditory or tactile tasksMultimodal spatial attention may be dominated by visualrepresentations of location perhaps because vision typ-ically has better spatial acuity than audition or touch(see Ward 1994) Moreover the present experimentalsituation provided many visible sources of informationabout the possible stimulus locations (eg the centralfixation cross visual cues the visible position of thearms on the table the visible locations of loudspeakersand tactile stimulators on the left and right side) Suchvisual information may have played a role in directingattention to the left or right side even for hearing ortouch That is visual information may have been used toanchor spatial selection by supramodal attentional con-trol processes even when relevant locations were se-lected for tactile or auditory tasks This possibility couldbe studied in further experiments by manipulatingwhether or not spatial information is available in vision(eg by repeating the study in complete darkness) In asimilar vein one could examine whether using nonvisualcentral cues would change the pattern of ERPs elicited inthe cue- target interval

Attentional Modulations of ERPs to PeripheralNontarget Stimuli

Modulations of ERPs elicited by irrelevant visual stimuliconfirmed previous findings (eg Eimer amp Schroger1998 Eimer amp Driver 2000) that directing attention toone side for audition or touch can affect modality-

Figure 7 Scalp topography maps of mean difference amplitudes obtained in the 500- 700-msec interval (relative to cue onset) by subtracting ERPselicited in response to cues directing attention to the right side from ERPs to cues directing attention to the left side An enhanced positivity atposterior electrodes contralateral to the direction of an attentional shift is reflected by negative amplitude values over the left hemisphere andpositive amplitude values over the right hemisphere Results obtained in the audition- (left map) and touch-relevant conditions (middle map) of thepresent study are presented together with results obtained in an unpublished study from our laboratory where an identical cueing procedure wasemployed to direct attention to the left or right side for a visual task (right map)

266 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

specific visual ERPs (with such effects arising as early asthe P1 andor N1 see Figure 6) More importantly thepresent study provides new ERP evidence concerningthe presence of similar tactile- auditory interactions inspatial attention Directing attention to the location ofrelevant tactile events affected subsequent auditory aswell as tactile ERPs on the cued versus uncued side (withmodulation for auditory stimuli apparent within 160- 200msec of their onset) This observation provides the firstdemonstration of tactile- auditory interactions in endog-enous spatial attention The cross-modal influence onauditory ERPs in the touch-relevant condition was sim-ilar to that previously observed by Eimer and Schroger(1998) in their vision-relevant condition in terms of thetiming and distribution of the auditory modulation

The present ERP evidence for an influence of cuedtactile side on auditory responses may initially appearinconsistent with recent unpublished work by LloydSpence Merat and McGlone (submitted) Using behav-ioral rather than ERP measures they found no evidencefor tactile- auditory interactions in endogenous spatialattention However it should be noted that auditory andtactile stimuli appeared at very different locations intheir study (in terms of stimulus elevation) unlike theclosely matched stimulus locations that were usedacross all modalities here which revealed tactile- audi-tory interactions for the first time

In marked contrast to the effects of spatial attentionon auditory ERPs in the touch-relevant condition direct-ing attention in the audition-relevant condition did notinfluence tactile ERPs While reliable attentional modu-lations of somatosensory ERPs were observed whentouch was relevant (similar to previous ERP studies oftactile spatial attention see Eimer amp Driver 2000Garc a-Larrea Lukaszewicz amp Maugiere 1995 MichieBearpark Crawford amp Glue 1987) no such effects werepresent in the audition-relevant condition even thoughclear cross-modal effects on visual ERPs were present inthe same auditory task This corroborates and extendsEimer and Driverrsquos (2000) proposal that the tactilemodality may be unique in that touch alone can belsquolsquodecoupledrsquorsquo from an influence of which side is cued foranother modality (provided that tactile events are en-tirely irrelevant throughout a block of trials) Touch maybe different in this respect because it inherently relatesto stimulation at the body surface rather than fromexternal space whereas audition and vision are bothmore distal senses Whatever the reason for touchbehaving differently the present results indicate thattouch can indeed be lsquolsquodecoupledrsquorsquo when task-irrelevantnot only from the spatial direction of attention for avisual task (as in Eimer amp Driver 2000) but also fromspatial attention for an auditory task as here

One striking feature of the multimodal interactionsobserved for ERPs to peripheral stimuli is that theytypically affected early components that are traditionallyconsidered to reflect lsquolsquounimodalrsquorsquo sensory- perceptual

processing Thus the visual N1 was affected by spatialattention when either audition or touch was relevantand likewise for the descending flank of the auditoryN1 when touch was relevant By contrast some latermodality-unspecific ERP components showed attention-al modulations only when a particular modality was task-relevant Thus the present study is consistent withprevious suggestions (eg McDonald amp Ward 2000Eimer amp Schroger 1998 Eimer amp Driver 2000) thatmultimodal influences of spatial attention may affectmodality-specific perceptual processing stages In con-trast attentional effects on later postperceptual stagesseem to be largely restricted to currently task-relevantmodalities (see Eimer 2001 for review)

Is Spatial Selection Supramodal

As mentioned in the Introduction two different ways ofthinking about cross-modal interactions in spatial atten-tion have emerged in the recent literature One ap-proach (eg Farah et al 1989) suggests thatattentional control mechanisms may be entirely supra-modal with cross-modal interactions in spatial attentionreflecting this directly Another approach (eg Spenceamp Driver 1996) argues that spatial selection may ariseprimarily within the task-relevant modality but thenspread (in attenuated fashion) to affect other modalities(a lsquolsquoseparable-but-linkedrsquorsquo view) The results from thepresent study suggest that some hybrid account may beneeded to provide a full explanation

The ERP evidence obtained here in the cue- targetinterval is strikingly consistent with the notion of supra-modal spatial selection of a cued location regardless ofwhether attention is directed to the relevant location oftactile auditory or visual events (eg see Figure 7) Asnoted above such selection could in principle operateprimarily within the medium of visual space yet stillremain lsquolsquosupramodalrsquorsquo in the sense that the same pro-cesses are involved in selecting a particular locationregardless of whether the task involves auditory tactileor visual judgements at that location

On the other hand further aspects of our data (and ofprevious findings) may seem more consistent with alsquolsquoseparable-but-linkedrsquorsquo view than with an entirely supra-modal account First effects of spatial attention onsensory processing are typically larger in amplitude forthe primary task-relevant modality than for irrelevantsecondary modalities (eg see Eimer amp Schroger 1998Hillyard et al 1984) Second as discussed above laterERP effects of spatial attention subsequent to the sen-sory components are typically found only for task-rele-vant modalities (Eimer 2001) Third touch canapparently be lsquolsquodecoupledrsquorsquo from spatial attention inother modalities (audition here vision in Eimer ampDriver 2000) when entirely task-irrelevant

The present ERP data from the cue- target intervalsuggest that the same supramodal attentional control

Eimer van Velzen and Driver 267

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

processes were activated during attentional shifts re-gardless of which modality (audition or touch) was task-relevant On the other hand the subsequent effects ofspatial attention on the ERPs elicited by peripheralauditory tactile or visual stimuli depended to someextent on which modality was task-relevant This appa-rent discrepancy implies that some aspect of preparatorystate must have differed between the audition- andtouch-relevant conditions even though the ERP dataobtained during attentional orienting were indistin-guishable for these two conditions The fact that therelevant modality remained constant for a seriesof blocks (while spatial attention was cued on a trial-by-trial basis) may have resulted in systematic tonicdifferences in the overall activation level between thetactile- and audition-relevant conditions For example alsquolsquodecouplingrsquorsquo of touch from multimodal spatial attentionmight in principle be achieved simply by reducing overallactivity levels for touch during blocks where tactilestimuli are irrelevant Such lsquolsquobaseline shiftsrsquorsquo in overallactivity have recently been uncovered in several func-tional imaging studies of unimodal visual attention (egKastner et al 1999 see Driver and Frith 2000 forreview) Applying functional imaging measures such asfMRI to the present design could shed further light onthis issue as some shifts in baseline activity for onemodality versus another might be found between blocks

Such considerations lead us to propose a lsquolsquohybridrsquorsquoaccount of cross-modal interactions in spatial attentionwhich combines supramodal attentional control pro-cesses with others that depend on the current taskrelevance of each modality The phasic selection ofrelevant locations may typically operate in a supramodalmanner (as the present ERP data from the cue- targetinterval strongly suggest) so that selecting a particularside for a task in one modality can influence processing inother modalities But the influence from such supra-modal spatial selection on stimulus processing within aparticular modality will also depend on the tonic state ofactivity in that modality which can vary with task rele-vance Note that this proposal combines aspects of twocontrasting approaches to multimodal interactions inspatial attention which are usually considered as mutu-ally exclusive (ie a strictly supramodal attentional con-trol system vs some degree of modality-independence inthe control of spatially selective processing) On ourpresent account spatial selection would operate supra-modally but its consequences for stimulus processingmay also depend on baseline shifts related to the taskrelevance of one modality versus another

In conclusion the present ERP data from the cue-target interval strongly suggest that directing attentionto the location of task-relevant stimuli is based on similarcontrol processes regardless of the modality of thesestimuli This implies a multimodal process of spatialselection which would in turn explain why selectinga particular side for one modality could affect modality-

specific ERPs for stimuli in other currently irrelevantmodalities On the other hand while such multimodalinfluences on sensory responses are indeed observedERP effects (and behavioral effects) of spatial attentionare usually maximal for the task-relevant modality Ahybrid account of cross-modal interactions in spatialattention which includes supramodal control of phasicattentional shifts plus tonic and modality-specific lsquolsquobase-line shiftsrsquorsquo may accommodate all of these findings

METHODS

Participants

Nineteen paid volunteers participated in the experi-ment Two had to be excluded because of poor eyefixation control in the cue- target interval one furtherparticipant was excluded because of a large number ofeye blinks during trials Thus 16 participants (9 women)aged 19- 30 years (mean of 24 years) remained in thesample Thirteen participants were right-handed threewere left-handed and all had normal or corrected visionby self-report

Stimuli and Apparatus

Participants sat in a dimly lit experimental chamber witha head-mounted microphone positioned about 2 cm infront of the mouth A computer monitor was placed infront of the participant at a viewing distance of 55 cmEnsembles comprising two adjacent triangles were em-ployed as central cue stimuli Together these twotriangles covered a visual angle of about 358 pound 258and were presented at the center of the computerscreen at an angle of about 308 below eye level Oneof the triangles was red the other blue and they alwayspointed in opposite directions (lsquolsquorsquorsquo or lsquolsquorsquorsquo) Thesedifferent cue arrangements were equiprobable and ran-domly distributed in each block A central fixation crosslocated in the space between the two triangles wascontinuously present on the computer screen through-out the experimental blocks Tactile stimuli were pre-sented via one of two solenoids on left and right sidewhich drove a metal rod with a blunt conical tip througha hole to contact the outside of either index finger Eachrod thus made contact with the outer side of the middlesegment of an index finger whenever a current waspassed through the solenoid The rods and the indexfingers were occluded so that participants could not seeany rod movements Continuous white noise was pre-sented from a central loudspeaker at 65 dB SPL asmeasured from the participantrsquos head throughout theexperimental blocks in order to mask any sounds madeby the operation of the tactile stimulators Peripheralauditory stimuli were louder bursts of white noise (200-msec duration including 5-msec rise and 5-msec falltimes amplitude 80 dB SPL) presented from one of two

268 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

loudspeakers located on the left and right side neareither hand Visual stimuli were 200-msec illuminationsof one of two ensembles of green LEDs on the left andright side consisting of six segments arranged in a circleplus one central segment with each ensemble locatednear one hand The angular size of each LED was 0658the diameter of the circle was 248 The two tactilestimulators the two peripheral loudspeakers and thetwo LED ensembles were placed on a table 258 to the leftor right of the central fixation cross in close spatialregister and at a viewing distance of about 45 cm

Tactile nontarget stimuli consisted of one rod tipcontacting the participantsrsquo index finger for 200 msecTactile target stimuli had a gap where this continuouscontact was interrupted for 10 msec after a duration of95 msec Auditory nontarget stimuli consisted of acontinuous white noise burst from either of the periph-eral loudspeakers For auditory target stimuli the noisewas interrupted after 90 msec by a 20-msec silentinterval after which the noise was turned on again for90 msec The gap was twice as long for auditory targetsthan for tactile targets because results from pilot studiesindicated that otherwise it was much easier to detecttactile gap targets than auditory gap targets All visualstimuli consisted of a continuous 200-msec illuminationof one LED ensemble (no visual targets were requiredas there was no visual task) Vocal response onset timeswere measured with a voice key

Procedure

The experiment consisted of 16 experimental blocks with72 trials per block Each trial started with the presentationof a central cue (100-msec duration) which was followedafter an interval of 600 msec by an auditory tactile orvisual peripheral stimulus (200-msec duration) Intertrialinterval was 1000 msec Two task conditions were deliv-ered each consisting of eight successive blocks In theaudition-relevant condition the task was to respondvocally (by saying lsquolsquoyesrsquorsquo) whenever an auditory gap targetwas presented at the location (left or right) indicated bythe central cue on that trial Auditory events on theuncued side (including those with gaps) were to beignored as were any tactile or visual events on eitherside In the touch-relevant condition the task was torespond vocally whenever a tactile gap target was pre-sented at a cued location with all the tactile events on theother side to be ignored as for all visual and auditoryevents on either side The relevant location for auditionor touch was cued by the direction of either the red or theblue central triangle For half of the participants bluetriangles indicated the attended location while red tri-angles were relevant for the other half of the participantsRelevant left- and right-pointing triangles were presentedwith equal probability to the left or right of fixation

In 60 trials per block auditory tactile or visual non-targets were presented with equal probability and in

random order on the left or right side at cued or uncuedlocations In the remaining randomly intermingled 12trials gap targets within the relevant modality werepresented on the left or right side Eight of these targetswere presented on the cued side (and thus required aresponse) while the remaining four were presented onthe uncued side (requiring no response) The order inwhich the audition- and touch-relevant conditions weredelivered was balanced across participants Instructionsspecifying the relevant modality were displayed on acomputer screen prior to the start of each blockParticipants were instructed to direct their attention tothe cued location in just the relevant modality in orderto respond as quickly and accurately as possible toauditory (or tactile depending on block) target stimulipresented at this location while withholding responsesto all other stimuli They were explicitly encouraged tomaintain central eye fixation throughout the trials Sev-eral training blocks were run prior to the beginning eachof the two task conditions Eye movements were closelymonitored during these training blocks Whenever thehorizontal EOG revealed that participants did not main-tain central eye fixation they were reminded again ofthe necessity of continuously fixating the central crossthroughout an experimental block Additional trainingblocks were run until fixation control was regarded assatisfactory

Recording and Data Analysis

EEG was recorded with Ag- AgCl electrodes and linked-earlobe reference from F7 F3 Fz F4 F8 FC5 FC6 T7C3 Cz C4 T8 CP5 CP6 P7 P3 Pz P4 and P8 (accord-ing to the 10- 20 system) and from OL and OR (locatedhalfway between O1 and P7 and O2 and P8 respec-tively) Horizontal EOG was recorded bipolarly from theouter canthi of both eyes The impedance for the EOGelectrodes was kept below 10 klaquo and for all otherelectrodes below 5 klaquo The amplifier bandpass was 01to 40 Hz EEG and EOG were sampled with a digitizationrate of 200 Hz and stored on disk Voice onset timeswere measured for each vocal response

EEG and EOG were epoched off-line into 1400-msecperiods starting 100 msec prior to cue onset and ending600 msec after the onset of the peripheral stimulusSeparate averages were computed for ERPs recorded inthe cue- target interval (relative to a 100-msec baselinepreceding cue onset) and for ERPs in response tosubsequent peripheral stimuli (relative to a 100-msecbaseline preceding the onset of these stimuli) Trialswith eyeblinks (Fpz exceeding plusmn60 mV relative to base-line) horizontal eye movements (HEOG exceeding plusmn30mV relative to baseline) or other artifacts (a voltageexceeding plusmn60 mV at any electrode location relative tobaseline) were excluded from analysis Averaged HEOGwaveforms in response to cues directing attention to theleft versus right side obtained in both task conditions

Eimer van Velzen and Driver 269

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

were scored for systematic deviations of eye positionindicating residual tendencies to move the eyes towardsthe cued location A residual HEOG deviation exceedingplusmn2 mV led to the disqualification of two participants

The EEG obtained in the cue- target interval wasaveraged for all combinations of task condition (audi-tion- vs touch-relevant) and cued direction (left vsright) Mean amplitude values were computed at later-al anterior sites (F78 F34 FC56) lateral central sites(T78 C34 CP56) and lateral posterior sites (P78P34 OLR) within different latency windows relative tocue onset and these values were analyzed separatelyfor anterior central and posterior electrodes by re-peated measures ANOVAs for the factors of task con-dition cued direction and recording side (left vs righthemisphere) The EEG obtained in response to pe-ripheral stimuli was averaged for nontarget (ie singlepulse) trials only to avoid contamination by vocalresponses to double-pulse targets Trials where false-positive vocal responses were recorded on nontargettrials were excluded from analysis

Separate averages were computed for auditory tac-tile and visual nontarget stimuli for all combinationsof task condition cued direction and stimulus side(left vs right) Mean amplitude values were computedfor auditory somatosensory and visual ERPs withindifferent latency windows measured relative to onsetof the peripheral stimulus Mean amplitude valuesobtained at lateral anterior sites lateral central siteslateral posterior sites and at midline electrodes (FzCz Pz) were submitted to separate ANOVAs For theanalyses of visual ERPs electrode Oz was included inthe set of midline electrodes Analyses of auditory andsomatosensory ERPs in response to peripheral stimuliincluded the factors of electrode site recording side(for lateral electrodes only) task condition spatialattention (stimulus at cued location vs uncued loca-tion) and stimulus side Visual ERPs were analyzedseparately for the audition- and touch-relevant condi-tions omitting the factor of task condition Whenappropriate Greenhouse- Geisser adjustments to thedegrees of freedom were performed and the adjustedp values are reported Nonsignificant terms and resultsdue trivially to stimulus and anatomical laterality arenot reported Whenever interactions between attentionand electrode site were found (for brevity these arenot all reported in full) additional analyses were con-ducted for single electrode sites For vocal responsesrepeated measures ANOVAs were performed on re-sponse latencies and on arc sin-transformed error ratesfor the factors of task condition and target location(left vs right)

Acknowledgments

This research was supported by a program grant from theMedical Research Council (UK) to JD and ME JD is a Royal

Society- Wolfson Research Merit Award Holder The authorsthank Heijo van de Werf Kate Parmenter Steffan Kennett andFriederike Schlaghecken for technical assistance

Reprint requests should be sent to Martin Eimer Departmentof Psychology Birkbeck College University of London MaletStreet London WC1E 7HX England UK or via e-mailmeimerbbkacuk

Notes

1 The only difference between the present results andprevious ERP patterns observed in the cue- target interval forvisual- spatial attention was that no posterior EDAN was foundat electrodes contralateral to the direction of an attentionalshift The absence of an EDAN in the present study mayreflect functional differences between attentional orientingprocesses in vision versus audition and touch Alternativelythe EDAN observed in some previous unimodal visual studiesmay at least partially result from physical dissimilaritiesbetween cues directing attention to the left versus rightWith one notable exception (Nobre et al 2000) all previousERP studies reporting EDAN-like effects in the cue- targetinterval compared left- and right-pointing arrow cues thatdiffered physically2 It is unlikely that the ADAN is an artifact of undetected eyemovements in the direction indicated by the cue Trials witheye movements were excluded from analysis and the residualHEOG deviation in the cue- target interval in response to leftand right cues was below plusmn1 mV Inspection of Figure 2 showsthat ADAN amplitudes were of comparable size at frontalelectrode pairs (F78 F34) and at C34 and tended to belargest at FC56 This distribution is inconsistent with the ideathat the ADAN is caused by eye movement contaminationwhich should be most pronounced at frontal sites

REFERENCES

Andersen R A Snyder L H Bradley D C amp Xing J (1997)Multimodal representations of space in the posterior parietalcortex and its use in planning movements Annual Review ofNeuroscience 20 303- 330

Corbetta M Miezin F M Shulman G L amp Petersen S E(1993) A PET study of visuospatial attention Journal ofNeuroscience 13 1202- 1226

Driver J amp Frith C (2000) Shifting baselines in attentionresearch Nature Reviews Neuroscience 1 231- 232

Driver J amp Spence C (1998) Attention and the crossmodalconstruction of space Trends in Cognitive Sciences 2254- 262

Eimer M (2001) Crossmodal links in spatial attention be-tween vision audition and touch Evidence from event-related brain potentials Neuropsychologia 391292- 1303

Eimer M amp Driver J (2000) An event-related brain potentialstudy of cross-modal links in spatial attention between visionand touch Psychophysiology 37 697- 705

Eimer M amp Schroger E (1998) ERP effects of intermodalattention and cross-modal links in spatial attention Psycho-physiology 35 313- 327

Farah M J Wong A B Monheit M A amp Morrow L A(1989) Parietal lobe mechanisms of spatial attentionModality-specific or supramodal Neuropsychologia 27461- 470

Garc a-Larrea L Lukaszewicz A C amp Mauguiere F (1995)Somatosensory responses during selective spatial attentionThe N120-to-N140 transition Psychophysiology 32 526- 537

Graziano M S A Yap G S amp Gross C G (1994) Coding

270 Journal of Cognitive Neuroscience Volume 14 Number 2

of visual space by premotor neurons Science 2661054- 1057

Harter M R Miller S L Price N J LaLonde M E amp KeyesA L (1989) Neural processes involved in directing attentionJournal of Cognitive Neuroscience 1 223- 237

Hillyard S A Simpson G V Woods D L Van Voorhis S ampMunte T F (1984) Event-related brain potentials andselective attention to different modalities In F Reinoso-Suarez amp C Ajmone-Marsan (Eds) Cortical integration(pp 395- 414) New York Raven Press

Hopfinger J B Buonocore M H amp Mangun G R (2000)The neural mechanisms of top-down attentional controlNature Neuroscience 3 284- 291

Hopfinger J B Jha A P Hopf J M Girelli M amp MangunG R (2000) Electrophysiological and neuroimaging studiesof voluntary and reflexive attention In S Monsell amp J Driver(Eds) Attention and performance XVII (pp 125- 153)Cambridge MIT Press

Kastner S Pinsk M A De Weerd P Desimone R ampUngerleider L G (1999) Increased activity in human visualcortex during directed attention in the absence of visualstimulation Neuron 22 751- 761

Kennett S Eimer M Spence C amp Driver J (2001) Tactile-visual links in exogenous spatial attention under differentpostures Convergent evidence from Psychophysics andERPs Journal of Cognitive Neuroscience 13 1- 16

LaBerge D (1995) Attentional processing CambridgeHarvard University Press

Lloyd D M Spence C Merat N amp McGlone F (submitted)Crossmodal links in covert endogenous spatial attentionbetween audition and touch

Macaluso E Frith C amp Driver J (2000a) Selective spatialattention in vision and touch Unimodal and crossmodalmechanisms revealed with PET Journal of Neurophysiology83 3062- 3075

Macaluso E Frith C D amp Driver J (2000b) Modulation ofhuman visual cortex by crossmodal spatial attentionScience 289 1206- 1208

Mangun G R (1994) Orienting attention in the visual fieldsAn electrophysiological analysis In H-J Heinze T F Munteamp G R Mangun (Eds) Cognitive electrophysiology (pp 81-101) Boston Birkhauser

McDonald J J amp Ward L M (2000) Involuntary listening aidsseeing Evidence from human electrophysiologyPsychological Science 11 167- 171

Michie P T Bearpark H M Crawford J M amp Glue L C T(1987) The effects of spatial selective attention on thesomatosensory event-related potential Psychophysiology24 449- 463

Nobre A C Sebestyen G N amp Miniussi C (2000) Thedynamics of shifting visuospatial attention revealed by event-related brain potentials Neuropsychologia 38 964- 974

Posner M I amp Petersen S E (1990) The attention systemof the human brain Annual Review of Neurosciences 1325- 42

Rosen A C Rao S M Caffara P Scaglioni A Bobholz JWoodley S C Hammerke T A Cunningham J M PrietoT E amp Binder J R(1999) Neural basis of endogenous andexogenous spatial orienting A functional MRI study Journalof Cognitive Neuroscience 11 135- 152

Spence C amp Driver J (1996) Audiovisual links in endogenouscovert spatial attention Journal of Experimental Psychol-ogy Human Perception and Performance 22 1005- 1030

Spence C Pavani F amp Driver J (2000) Crossmodal linksbetween vision and touch in covert endogenous spatialattention Journal of Experimental Psychology HumanPerception and Performance 26 1298- 1319

Teder-Salejarvi W A Munte T F Sperlich F-J amp HillyardS A (1999) Intra-modal and cross-modal spatial attention toauditory and visual stimuli An event-related brain potential(ERP) study Cognitive Brain Research 8 327- 343

Ward L M (1994) Supramodal and modality-specificmechanisms for stimulus-driven shifts of auditory and visualattention Canadian Journal of Experimental Psychology48 242- 259

Ward L M McDonald J J amp Lin D (2000) On asymmetriesin cross-modal spatial attention orienting Perception andPsychophysics 62 1258- 1264

Yamaguchi S Tsuchiya H amp Kobayashi S (1994) Electro-encephalographic activity associated with shifts of visuospa-tial attention Brain 117 53- 562

Yamaguchi S Tsuchiya H amp Kobayashi S (1995)Electrophysiologic correlates of age effects on visuospatialattention shift Cognitive Brain Research 3 41- 49

Eimer van Velzen and Driver 271