Selective Dorsal and Ventral Processing Evidence fora Common Attentional Mechanism in Reaching and

Perception

Heiner Deubel Werner X Schneider and Ingo PaprottaInstitut fuumlr Psychologie Allgemeine und Experimentelle Psychologie

Ludwig-Maximilians-Universitaumlt Muumlnchen Germany

The primate visual system can be divided into a ventral stream for perceptionand recognition and a dorsal stream for computing spatial information for motoraction How are selection mechanisms in both processing streams coordinatedWe recently demonstrated that selection-for-perception in the ventral stream(usually termed ldquovisual attentionrdquo) and saccade target selection in the dorsalstream are tightly coupled (Deubel amp Schneider 1996) Here we investigatewhether such coupling also holds for the preparation of manual reaching move-ments A dual-task paradigm required the preparation of a reaching movementto a cued item in a letter string Simultaneously the ability to discriminatebetween the symbols ldquoE rdquo and ldquo$ rdquo presented tachistoscopically within the sur-rounding distractors was taken as a measure of perceptual performance The datademonstrate that discrimination performance is superior when thediscriminationstimulus is also the target for manual aiming when the discrimination stimulusandpointing targetrefertodifferentobjects performancedeteriorates Thereforeit is not possible to maintain attention on a stimulus for the purpose of discrimi-nationwhiledirecting amovement toaspatially separateobject Theresults arguefor an obligatory coupling of (ventral) selection-for-perception and (dorsal)selection-for-action

INTRODUCTION

Our knowledge about the architecture of the visual system of primates hasincreased enormously during the last two decades There is growing consensusthat visual processing occurs in parallel and interaction streams at different

Requests for reprints should be addressed to Heiner Duebel Institut fuumlr PsychologieLudwig-Maximilians-Universitaumlt Leopoldstr 13 D-80802 Muumlnchen Germany E-maildeubel psyuni-muenchende

This research was supported by the Deutsche Forschungsgemeinschaft SFB 462 (ldquoSensomo-torikrdquo)

Oacute 1998 Psychology Press Ltd

VISUAL COGNITION 1998 5 (12) 81ndash107

quasi-hierarchical levels (eg DeYoeampvanEssen 1988 LivingstoneampHubel1988 Milneramp Goodale 1995 Zeki 1993 Anumberof suggestions have beenmade how this parallel and distributed processing of visual information mightbe functionally organized Based on lesion work in monkeys Mishkin Unger-leider and Macko (1983) claimed that the visual system consists of two mainpathways namely the dorsal ldquowhererdquo-pathway and the ventral ldquowhatrdquo-path-way The suggested function of the ldquowhatrdquo-pathway is to recognize objectsbased on their visual appearance The ldquowhererdquo-pathway on the other handcomputes spatial information about objects At the cortical level the segrega-tion of both pathways can be tracked back to the primary visual cortex areaV1 From there the ldquowhererdquo-pathway runs dorsally into the posterior parietallobe whereas theldquowhatrdquo-pathway leads ventrally to the inferior temporal lobeSince this proposal a large body of research has supported this distinction oftwo main pathways (but see Zeki 1993) For instance patients with brainlesions restricted to the inferior temporal cortex have problems recognizingobjects by sight a symptom called ldquovisual agnosiardquo (eg Farah 1990 Kolb ampWhishaw 1990) At the same time spatial abilities such as pointing to anobject are left intact When agnosia is purely visual recognition by othersenses such as touch is still intact Lesions restricted to the superior parietalareas of the dorsal ldquowhererdquo-pathway on the other hand can cause a symptomcalled ldquooptic ataxiardquo (eg Milner amp Goodale 1995) These patients are able toidentify objects due to their visual appearance but they exhibit misreaching(mislocation) towards the same objects

The labelling of theventral and dorsal pathways as a ldquowhatrdquo- and a ldquowhererdquo-pathway respectively was recently criticized by Goodale and Milner (1992Milner amp Goodale 1995) Though they ascribe the computation of ldquowhatrdquo-as-pectsmdashthat is the identification of objectsmdashto the ventral pathway theydisagree about the function of the dorsal pathway They do not see perceptionof the spatial layout of the external world as its main task but insteadcomputation of spatial information for motor actions such as a saccade or areach towards an object In other words Goodale and Milner (1992) suggest ashift inemphasis fromspatial perceptiontospatial information foraction Theirview of dorsal processing is supported by human neuropsychological studiesand neurophysiological work in macaques especially by single-cell recordings(see Milner amp Goodale 1995) The literature reviewed indicates that the ideaof a single representation of external space is probably wrong and that insteadseveral spatial-motor representationsmdashsometimes called ldquoprocessingstreamsrdquomdashexist in parallel for different kinds of motor actions (see eg Graz-iano amp Gross 1994 Milner amp Goodale 1995 Rizzolatti Riggio amp Sheliga1994 Stein 1992) For instance information about saccade landing points isprobably computed and coded in the lateral intraparietal area (LIP) whileendpoints for grasping movements are computed in area 7b (both are part ofthe parietal lobe) Therefore the brain seems to code spatial information for

82 DEUBEL ET AL

different effectorsmdashthat is for different action classesmdashin different parts ofthe brain In summary Goodale and Milner (1992) suggest that the ventralstream is involved in visual perception and identification whereas the dorsalstream computes information for spatial-motor actions A related distinctionwas recently suggested by Jeannerod (1994) who differentiated between aldquosemantic moderdquo of processing located in the temporal lobe (ventral stream)and a ldquopragmatic moderdquo located in the parietal cortex (dorsal stream)

Visual processing in both streams does not occur in a purely automaticldquobottom-uprdquo driven manner Rather control of processing is task-dependentthis type of selectivity of visual processing has been called ldquoendogenous visualattentionrdquo (eg Posner 1980) Much research in experimental psychology andthe neurosciences has investigated the properties of these selection processesin vision (for overviews see Bundesen 1990 Desimone amp Duncan 1995Posner amp Petersen 1990 Schneider 1993 Treisman 1988 van der Heijden1992) Traditional experimental psychology has focused on the function ofvisual attention in the ventral stream that is on ldquoselection-for-visual-percep-tionrdquo For instance experiments on visual search (for overviews see Treismanamp Gormican 1988 Wolfe 1994) have attempted to determine how fast andhow accuratecertain visual attributes andtheirconjunctions canbe ldquoperceivedrdquoto be signalled In most of these investigations ldquoventralrdquo attributes such ascolour orientation and so on served as the properties that defined the searchtarget Therefore selection-for-visual-perception (in contrast to selection-for-spatial-motor-controlmdashthe dorsal processing domain) has been the main topicof searchtasks Anotherresearchtool fortheeffects of visual attentioninventralprocessing is the spatial pre-cueing paradigm (eg Eriksen amp Hoffman 1973Posner 1980 vanderHeijden 1992) Experiments haveshown thatpre-knowl-edge about the possible location of a target leads to faster and more accurateresponses to visual aspects such as alphanumeric identity or simple shapefeatures such as curved versus straight (for overviews see Posner amp Raichle1994 van der Heijden 1992)

This bias in measuring theeffect of visual attention mainly forventral visualprocessing can be traced back tothe suggested functions of attention Attentionis assumed to facilitate detection (Posner 1980) to allow ldquofeature integrationrdquo(Treisman amp Gelade 1980) ldquoobject recognitionrdquo (LaBerge amp Brown 1989Schneider 1995) and ldquoentry to visual short-term memoryrdquo (Bundesen 1990Duncan amp Humphreys 1989) However these assumptions do not imply thatthe selection mechanism itself is located in the ventral stream only Insteadseveral theories have suggested a central role of thedorsal stream incontrollingtheattentional mechanism sometimes calledthe ldquospatial attentionmechanismrdquo(eg La Berge amp Brown 1989 Posner amp Petersen 1990 Schneider 1995 vander Heijden 1992)

Compared to the large body of theoretical work on the relationship betweenattention and (ventral) perceptual processing there are scant data on the role of

REACHING AND ATTENTION 83

visual attention in dorsal processing more precisely the role of attention inspatial-motor control Allport (1987) and Neumann (1987) suggested thatspatial motor actions such as grasping one object from among other objectsmay also be a selection process what Allport (1987) called ldquoselection-for-actionrdquo Natural environments usually contain several objects and only one ofthemshouldbe usedas thetarget foranindividual action Forinstance graspinga pen among other pens requires the motor system to receive spatial informa-tion probably in arm-centred coordinates (Graziano amp Gross 1994) of theintended pen only Information from other pens has to be excluded fromcontrolling the grasping action In other words an attentional mechanism isneeded that selects the spatial information of the movement target Becausespatial information is provided by the visual system (the dorsal pathway)Allport (1987 1989) and Neumann (1987 1990) have suggested that visualattention is involved in this selection process Another example of selection-for-spatial-motor-action refers to the control of saccadic eye movementsBefore each saccade the next fixation point has to be selected among manypotential candidates in the environment

Unfortunately there has not been much experimental workon selection-for-spatial-motor-action Tipper Lortie and Baylis (1992) investigated the role ofvisual attention for manual reaching in an interference paradigm They wishedto determine if the interference effects found for ventral visual processing (egEriksen amp Eriksen 1974) can also be obtained for spatial-motor actions Thedegree of interference is usually considered as a measure of the efficiency ofattentional processes In these experiments subjects had to reach as fast andas precisely as possible from a starting position to one of nine locationsindicated by a red light (thetarget) In some trials a yellow light (thedistractor)appeared simultaneously with the red target light at a different locationSubstantial interference effects were obtained response latencies were pro-longed compared totrials where nodistractor appeared This interference effectwas only observedwhenthedistractorwas locatedbetweenthestarting positionand the target Tipper et al (1992) argued that their results reflect ldquoaction-cen-tred attentionrdquo emphasizing that the location of the movement target is mostrelevant to the amount of interference In summary these results show thatinterference effects by nearby objects can also be obtained for spatial-motoraction such as reaching suggesting that visual attention processes are alsoinvolved in selection-for-spatial-motor-action A similar conclusion wasreached by Castiello (1996) In one of his experiments subjects had to grasp atarget as their primary task A secondary non-spatial task was required for adifferent object located close to the target Castiello observed interferenceeffects of the secondary task on the kinematics of the primary graspingmovement given the subject performed a subsidiary task which involved thedistractor

84 DEUBEL ET AL

Another line of research dealing with dorsal selection concerns the relation-ship between eye movement control and visual attention The question iswhether visual attention for perceptual processing on the one hand and selec-tion of a target for a saccade on the other are independent or not The resultsof early experiments on this issue were controversial (eg Klein 1980 Posner1980) partly due to methodological problems (see Shepherd Findlay ampHockey 1986) More recent studies (Deubel amp Schneider 1996 Hoffman ampSubramaniam 1995 Kowler Anderson Dosher amp Blaser 1995 Schneider ampDeubel 1995) have clearly demonstrated a strict link between ventral selec-tion-for-perception and dorsal selection-for-a-saccade

In the experiments of Deubel and Schneider (1996) subjects had to saccadeto locations within horizontal letter strings left or right of a central fixationcross The performance in discriminating between the ldquoE rdquo and ldquo$ rdquo presentedtachistoscopically before the saccade within the surrounding distractors wastaken as a measure of visual attention in perception The results showed thatdiscrimination performance is best when discrimination target and saccadetargetrefer tothesameobject Thefindings argueforanobligatory andselectivecoupling of dorsal processing for saccade programming and ventral processingfor perception and discrimination this coupling is restricted to one commontarget object at a time

Based on these results and other computational considerations Schneider(1995) postulated a Visual Attention Model (VAM) that suggests a commonselectionmechanism forbothprocessing streams Inlinewithtwo-stagemodelsof perception and attention (Neisser 1967) a first stage of low-level visualprocessing computes in parallel in early visual areas of thebrain (eg V1 V2)elementary visual information in the form of ldquoprimitiverdquo object structures(visual units) Higher-level visual processing in the dorsal and ventral streamis assumed to be capacity-limited that is it occurs only for one visual unit (oneldquoobjectrdquo) at a time In the model visual attention is the mechanism thatdetermines the unit carries out the selection and gates the information flowfrom low- to high-level vision in a way that only information from one objectis furtherprocessed TheVAM claims thatvisual attentionselects one low-levelvisual objectata time leading toprioritizedperceptual processing intheventralstream (ie the object is recognized) Simultaneously possible spatial-motoractions (saccade pointing reaching grasping etc) towards this object areprogrammed in the dorsal stream Only the (effector-specific) ldquogordquo signal isnecessary to convert the programs into overt action

Such attention-mediated and object-specific coupling of dorsal and ventralprocessing has already been demonstrated for eye movement control andperceptual selection (Deubel amp Schneider 1996) More than just for saccadeshowever VAM predicts that the same coupling should also hold for aimingreaching and grasping (Schneider 1995 p 363) In the present study we

REACHING AND ATTENTION 85

analysed the coupling of reaching movements and visual discrimination Forthis purpose a dual-task paradigm similar to that used in our previous studieswas developed The primary task was to make a goal-directed reaching move-ment to a cued object measuring selection-for-spatial-motor-action in thedorsal stream Prior to the movement a secondary task required subjects todiscriminate between the characters ldquoE rdquo and ldquo$ rdquo measuring selection-for-per-ception (ldquotraditionalrdquo visual attention) in the ventral stream It is hypothesizedthat the programming of the reaching movement yokes the visual attentionmechanism so that during this selection process no other object can beprocessed in high-level ventral vision Consequently discrimination perform-ance should be best when discrimination target and reaching target refer to thesame object Fornon-corresponding reaching and discrimination targets betterthan chance performance should be possible only when visual attention shiftsfirst to the discrimination target and then to the reaching target In this caselonger initiation latencies for the movement should be expected

METHODS

Subjects

Five subjects participated in the experiments their age ranged from 22 to 28years They had normal vision and normal motor behaviour All subjects wereexperienced in a variety of experiments in oculomotor research One subjectwas one of the authors of the study the others were naive with respect to theaim of the experiments

Experimental Set-up

Figure 1 shows a sketch of the experimental set-up The subject was seated ina dimly lit room The visual stimuli were presented on a fast 21 inch colourmonitor (CONRAC 7550 C21) visible through a one-way mirror The monitorprovided a frame frequency of 100 Hz at a spatial resolution of 64 pixels perinch The active screen size was 40 times 30 cm theviewing distance was 577 cmThe video signals were generated by a freely programmable graphics board(Kontron KONTRAST 8000) controlled by a PC via the TIGA (Texas Instru-ments Graphics Adapter) interface The stimuli appeared on a grey backgroundadjusted to a mean luminance of 22 cdm2 The luminance of the stimuli was23 cdm2 The relatively high background brightness is essential to avoid theeffects of phosphor persistence (Wolf and Deubel 1997)

The use of a one-way mirror allowed free hand movements to the stimuliwithout visual feedback about hand position Reaching movements were re-corded with a Fastrak electromagnetic position and orientation measuring

86 DEUBEL ET AL

system (Polhemus Inc 1993) and sampled at 400 Hz The sender device wasfixed 60 cm in front of the subject The sender emits time-multiplexedorthogonal electromagnetic fields of 10 kHz frequency From induction in thereceiver which was mounted on the fingertip of the subjectrsquos right hand theorientation relative to the sender device is calculated by a central processingunit From the intensity of the electromagnetic field the distance betweensender and receiver is determined The position in space is calculated fromdistance and orientation by use of a specific digital signal processor(TI320C30) The device allows for a maximum translation range of 10 feetwith an accuracy of 003 inches RMS The frequency response is 120 Hzwithout further filtering the phase lag response is 4 msec Connected on thereceiver was a red LED (5 mm diameter) controlled by the PC The LEDallowed us to provide controlled visual feedback about the spatial position ofthe fingertip

Eye fixationwas monitoredby aninfraredeyetracker(IRIS SkalarMedical)with a temporal bandwidth of 240 Hz This device measures the reflectiondifference between the sclera and iris by infrared LEDs and phototransistorsthat are situated next to the subjectrsquos eyes Head movements were restricted byan adjustable chin rest The experiments were controlled by a 486 PC The PCalso served for the automatic off-line analysis of the pointing movement datafor which movement latencies and start and end positions of the manualresponses were determined

FIG 1 Experimental apparatus

REACHING AND ATTENTION 87

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

quasi-hierarchical levels (eg DeYoeampvanEssen 1988 LivingstoneampHubel1988 Milneramp Goodale 1995 Zeki 1993 Anumberof suggestions have beenmade how this parallel and distributed processing of visual information mightbe functionally organized Based on lesion work in monkeys Mishkin Unger-leider and Macko (1983) claimed that the visual system consists of two mainpathways namely the dorsal ldquowhererdquo-pathway and the ventral ldquowhatrdquo-path-way The suggested function of the ldquowhatrdquo-pathway is to recognize objectsbased on their visual appearance The ldquowhererdquo-pathway on the other handcomputes spatial information about objects At the cortical level the segrega-tion of both pathways can be tracked back to the primary visual cortex areaV1 From there the ldquowhererdquo-pathway runs dorsally into the posterior parietallobe whereas theldquowhatrdquo-pathway leads ventrally to the inferior temporal lobeSince this proposal a large body of research has supported this distinction oftwo main pathways (but see Zeki 1993) For instance patients with brainlesions restricted to the inferior temporal cortex have problems recognizingobjects by sight a symptom called ldquovisual agnosiardquo (eg Farah 1990 Kolb ampWhishaw 1990) At the same time spatial abilities such as pointing to anobject are left intact When agnosia is purely visual recognition by othersenses such as touch is still intact Lesions restricted to the superior parietalareas of the dorsal ldquowhererdquo-pathway on the other hand can cause a symptomcalled ldquooptic ataxiardquo (eg Milner amp Goodale 1995) These patients are able toidentify objects due to their visual appearance but they exhibit misreaching(mislocation) towards the same objects

The labelling of theventral and dorsal pathways as a ldquowhatrdquo- and a ldquowhererdquo-pathway respectively was recently criticized by Goodale and Milner (1992Milner amp Goodale 1995) Though they ascribe the computation of ldquowhatrdquo-as-pectsmdashthat is the identification of objectsmdashto the ventral pathway theydisagree about the function of the dorsal pathway They do not see perceptionof the spatial layout of the external world as its main task but insteadcomputation of spatial information for motor actions such as a saccade or areach towards an object In other words Goodale and Milner (1992) suggest ashift inemphasis fromspatial perceptiontospatial information foraction Theirview of dorsal processing is supported by human neuropsychological studiesand neurophysiological work in macaques especially by single-cell recordings(see Milner amp Goodale 1995) The literature reviewed indicates that the ideaof a single representation of external space is probably wrong and that insteadseveral spatial-motor representationsmdashsometimes called ldquoprocessingstreamsrdquomdashexist in parallel for different kinds of motor actions (see eg Graz-iano amp Gross 1994 Milner amp Goodale 1995 Rizzolatti Riggio amp Sheliga1994 Stein 1992) For instance information about saccade landing points isprobably computed and coded in the lateral intraparietal area (LIP) whileendpoints for grasping movements are computed in area 7b (both are part ofthe parietal lobe) Therefore the brain seems to code spatial information for

82 DEUBEL ET AL

different effectorsmdashthat is for different action classesmdashin different parts ofthe brain In summary Goodale and Milner (1992) suggest that the ventralstream is involved in visual perception and identification whereas the dorsalstream computes information for spatial-motor actions A related distinctionwas recently suggested by Jeannerod (1994) who differentiated between aldquosemantic moderdquo of processing located in the temporal lobe (ventral stream)and a ldquopragmatic moderdquo located in the parietal cortex (dorsal stream)

Visual processing in both streams does not occur in a purely automaticldquobottom-uprdquo driven manner Rather control of processing is task-dependentthis type of selectivity of visual processing has been called ldquoendogenous visualattentionrdquo (eg Posner 1980) Much research in experimental psychology andthe neurosciences has investigated the properties of these selection processesin vision (for overviews see Bundesen 1990 Desimone amp Duncan 1995Posner amp Petersen 1990 Schneider 1993 Treisman 1988 van der Heijden1992) Traditional experimental psychology has focused on the function ofvisual attention in the ventral stream that is on ldquoselection-for-visual-percep-tionrdquo For instance experiments on visual search (for overviews see Treismanamp Gormican 1988 Wolfe 1994) have attempted to determine how fast andhow accuratecertain visual attributes andtheirconjunctions canbe ldquoperceivedrdquoto be signalled In most of these investigations ldquoventralrdquo attributes such ascolour orientation and so on served as the properties that defined the searchtarget Therefore selection-for-visual-perception (in contrast to selection-for-spatial-motor-controlmdashthe dorsal processing domain) has been the main topicof searchtasks Anotherresearchtool fortheeffects of visual attentioninventralprocessing is the spatial pre-cueing paradigm (eg Eriksen amp Hoffman 1973Posner 1980 vanderHeijden 1992) Experiments haveshown thatpre-knowl-edge about the possible location of a target leads to faster and more accurateresponses to visual aspects such as alphanumeric identity or simple shapefeatures such as curved versus straight (for overviews see Posner amp Raichle1994 van der Heijden 1992)

This bias in measuring theeffect of visual attention mainly forventral visualprocessing can be traced back tothe suggested functions of attention Attentionis assumed to facilitate detection (Posner 1980) to allow ldquofeature integrationrdquo(Treisman amp Gelade 1980) ldquoobject recognitionrdquo (LaBerge amp Brown 1989Schneider 1995) and ldquoentry to visual short-term memoryrdquo (Bundesen 1990Duncan amp Humphreys 1989) However these assumptions do not imply thatthe selection mechanism itself is located in the ventral stream only Insteadseveral theories have suggested a central role of thedorsal stream incontrollingtheattentional mechanism sometimes calledthe ldquospatial attentionmechanismrdquo(eg La Berge amp Brown 1989 Posner amp Petersen 1990 Schneider 1995 vander Heijden 1992)

Compared to the large body of theoretical work on the relationship betweenattention and (ventral) perceptual processing there are scant data on the role of

REACHING AND ATTENTION 83

visual attention in dorsal processing more precisely the role of attention inspatial-motor control Allport (1987) and Neumann (1987) suggested thatspatial motor actions such as grasping one object from among other objectsmay also be a selection process what Allport (1987) called ldquoselection-for-actionrdquo Natural environments usually contain several objects and only one ofthemshouldbe usedas thetarget foranindividual action Forinstance graspinga pen among other pens requires the motor system to receive spatial informa-tion probably in arm-centred coordinates (Graziano amp Gross 1994) of theintended pen only Information from other pens has to be excluded fromcontrolling the grasping action In other words an attentional mechanism isneeded that selects the spatial information of the movement target Becausespatial information is provided by the visual system (the dorsal pathway)Allport (1987 1989) and Neumann (1987 1990) have suggested that visualattention is involved in this selection process Another example of selection-for-spatial-motor-action refers to the control of saccadic eye movementsBefore each saccade the next fixation point has to be selected among manypotential candidates in the environment

Unfortunately there has not been much experimental workon selection-for-spatial-motor-action Tipper Lortie and Baylis (1992) investigated the role ofvisual attention for manual reaching in an interference paradigm They wishedto determine if the interference effects found for ventral visual processing (egEriksen amp Eriksen 1974) can also be obtained for spatial-motor actions Thedegree of interference is usually considered as a measure of the efficiency ofattentional processes In these experiments subjects had to reach as fast andas precisely as possible from a starting position to one of nine locationsindicated by a red light (thetarget) In some trials a yellow light (thedistractor)appeared simultaneously with the red target light at a different locationSubstantial interference effects were obtained response latencies were pro-longed compared totrials where nodistractor appeared This interference effectwas only observedwhenthedistractorwas locatedbetweenthestarting positionand the target Tipper et al (1992) argued that their results reflect ldquoaction-cen-tred attentionrdquo emphasizing that the location of the movement target is mostrelevant to the amount of interference In summary these results show thatinterference effects by nearby objects can also be obtained for spatial-motoraction such as reaching suggesting that visual attention processes are alsoinvolved in selection-for-spatial-motor-action A similar conclusion wasreached by Castiello (1996) In one of his experiments subjects had to grasp atarget as their primary task A secondary non-spatial task was required for adifferent object located close to the target Castiello observed interferenceeffects of the secondary task on the kinematics of the primary graspingmovement given the subject performed a subsidiary task which involved thedistractor

84 DEUBEL ET AL

Another line of research dealing with dorsal selection concerns the relation-ship between eye movement control and visual attention The question iswhether visual attention for perceptual processing on the one hand and selec-tion of a target for a saccade on the other are independent or not The resultsof early experiments on this issue were controversial (eg Klein 1980 Posner1980) partly due to methodological problems (see Shepherd Findlay ampHockey 1986) More recent studies (Deubel amp Schneider 1996 Hoffman ampSubramaniam 1995 Kowler Anderson Dosher amp Blaser 1995 Schneider ampDeubel 1995) have clearly demonstrated a strict link between ventral selec-tion-for-perception and dorsal selection-for-a-saccade

In the experiments of Deubel and Schneider (1996) subjects had to saccadeto locations within horizontal letter strings left or right of a central fixationcross The performance in discriminating between the ldquoE rdquo and ldquo$ rdquo presentedtachistoscopically before the saccade within the surrounding distractors wastaken as a measure of visual attention in perception The results showed thatdiscrimination performance is best when discrimination target and saccadetargetrefer tothesameobject Thefindings argueforanobligatory andselectivecoupling of dorsal processing for saccade programming and ventral processingfor perception and discrimination this coupling is restricted to one commontarget object at a time

Based on these results and other computational considerations Schneider(1995) postulated a Visual Attention Model (VAM) that suggests a commonselectionmechanism forbothprocessing streams Inlinewithtwo-stagemodelsof perception and attention (Neisser 1967) a first stage of low-level visualprocessing computes in parallel in early visual areas of thebrain (eg V1 V2)elementary visual information in the form of ldquoprimitiverdquo object structures(visual units) Higher-level visual processing in the dorsal and ventral streamis assumed to be capacity-limited that is it occurs only for one visual unit (oneldquoobjectrdquo) at a time In the model visual attention is the mechanism thatdetermines the unit carries out the selection and gates the information flowfrom low- to high-level vision in a way that only information from one objectis furtherprocessed TheVAM claims thatvisual attentionselects one low-levelvisual objectata time leading toprioritizedperceptual processing intheventralstream (ie the object is recognized) Simultaneously possible spatial-motoractions (saccade pointing reaching grasping etc) towards this object areprogrammed in the dorsal stream Only the (effector-specific) ldquogordquo signal isnecessary to convert the programs into overt action

Such attention-mediated and object-specific coupling of dorsal and ventralprocessing has already been demonstrated for eye movement control andperceptual selection (Deubel amp Schneider 1996) More than just for saccadeshowever VAM predicts that the same coupling should also hold for aimingreaching and grasping (Schneider 1995 p 363) In the present study we

REACHING AND ATTENTION 85

analysed the coupling of reaching movements and visual discrimination Forthis purpose a dual-task paradigm similar to that used in our previous studieswas developed The primary task was to make a goal-directed reaching move-ment to a cued object measuring selection-for-spatial-motor-action in thedorsal stream Prior to the movement a secondary task required subjects todiscriminate between the characters ldquoE rdquo and ldquo$ rdquo measuring selection-for-per-ception (ldquotraditionalrdquo visual attention) in the ventral stream It is hypothesizedthat the programming of the reaching movement yokes the visual attentionmechanism so that during this selection process no other object can beprocessed in high-level ventral vision Consequently discrimination perform-ance should be best when discrimination target and reaching target refer to thesame object Fornon-corresponding reaching and discrimination targets betterthan chance performance should be possible only when visual attention shiftsfirst to the discrimination target and then to the reaching target In this caselonger initiation latencies for the movement should be expected

METHODS

Subjects

Five subjects participated in the experiments their age ranged from 22 to 28years They had normal vision and normal motor behaviour All subjects wereexperienced in a variety of experiments in oculomotor research One subjectwas one of the authors of the study the others were naive with respect to theaim of the experiments

Experimental Set-up

Figure 1 shows a sketch of the experimental set-up The subject was seated ina dimly lit room The visual stimuli were presented on a fast 21 inch colourmonitor (CONRAC 7550 C21) visible through a one-way mirror The monitorprovided a frame frequency of 100 Hz at a spatial resolution of 64 pixels perinch The active screen size was 40 times 30 cm theviewing distance was 577 cmThe video signals were generated by a freely programmable graphics board(Kontron KONTRAST 8000) controlled by a PC via the TIGA (Texas Instru-ments Graphics Adapter) interface The stimuli appeared on a grey backgroundadjusted to a mean luminance of 22 cdm2 The luminance of the stimuli was23 cdm2 The relatively high background brightness is essential to avoid theeffects of phosphor persistence (Wolf and Deubel 1997)

The use of a one-way mirror allowed free hand movements to the stimuliwithout visual feedback about hand position Reaching movements were re-corded with a Fastrak electromagnetic position and orientation measuring

86 DEUBEL ET AL

system (Polhemus Inc 1993) and sampled at 400 Hz The sender device wasfixed 60 cm in front of the subject The sender emits time-multiplexedorthogonal electromagnetic fields of 10 kHz frequency From induction in thereceiver which was mounted on the fingertip of the subjectrsquos right hand theorientation relative to the sender device is calculated by a central processingunit From the intensity of the electromagnetic field the distance betweensender and receiver is determined The position in space is calculated fromdistance and orientation by use of a specific digital signal processor(TI320C30) The device allows for a maximum translation range of 10 feetwith an accuracy of 003 inches RMS The frequency response is 120 Hzwithout further filtering the phase lag response is 4 msec Connected on thereceiver was a red LED (5 mm diameter) controlled by the PC The LEDallowed us to provide controlled visual feedback about the spatial position ofthe fingertip

Eye fixationwas monitoredby aninfraredeyetracker(IRIS SkalarMedical)with a temporal bandwidth of 240 Hz This device measures the reflectiondifference between the sclera and iris by infrared LEDs and phototransistorsthat are situated next to the subjectrsquos eyes Head movements were restricted byan adjustable chin rest The experiments were controlled by a 486 PC The PCalso served for the automatic off-line analysis of the pointing movement datafor which movement latencies and start and end positions of the manualresponses were determined

FIG 1 Experimental apparatus

REACHING AND ATTENTION 87

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

different effectorsmdashthat is for different action classesmdashin different parts ofthe brain In summary Goodale and Milner (1992) suggest that the ventralstream is involved in visual perception and identification whereas the dorsalstream computes information for spatial-motor actions A related distinctionwas recently suggested by Jeannerod (1994) who differentiated between aldquosemantic moderdquo of processing located in the temporal lobe (ventral stream)and a ldquopragmatic moderdquo located in the parietal cortex (dorsal stream)

Visual processing in both streams does not occur in a purely automaticldquobottom-uprdquo driven manner Rather control of processing is task-dependentthis type of selectivity of visual processing has been called ldquoendogenous visualattentionrdquo (eg Posner 1980) Much research in experimental psychology andthe neurosciences has investigated the properties of these selection processesin vision (for overviews see Bundesen 1990 Desimone amp Duncan 1995Posner amp Petersen 1990 Schneider 1993 Treisman 1988 van der Heijden1992) Traditional experimental psychology has focused on the function ofvisual attention in the ventral stream that is on ldquoselection-for-visual-percep-tionrdquo For instance experiments on visual search (for overviews see Treismanamp Gormican 1988 Wolfe 1994) have attempted to determine how fast andhow accuratecertain visual attributes andtheirconjunctions canbe ldquoperceivedrdquoto be signalled In most of these investigations ldquoventralrdquo attributes such ascolour orientation and so on served as the properties that defined the searchtarget Therefore selection-for-visual-perception (in contrast to selection-for-spatial-motor-controlmdashthe dorsal processing domain) has been the main topicof searchtasks Anotherresearchtool fortheeffects of visual attentioninventralprocessing is the spatial pre-cueing paradigm (eg Eriksen amp Hoffman 1973Posner 1980 vanderHeijden 1992) Experiments haveshown thatpre-knowl-edge about the possible location of a target leads to faster and more accurateresponses to visual aspects such as alphanumeric identity or simple shapefeatures such as curved versus straight (for overviews see Posner amp Raichle1994 van der Heijden 1992)

This bias in measuring theeffect of visual attention mainly forventral visualprocessing can be traced back tothe suggested functions of attention Attentionis assumed to facilitate detection (Posner 1980) to allow ldquofeature integrationrdquo(Treisman amp Gelade 1980) ldquoobject recognitionrdquo (LaBerge amp Brown 1989Schneider 1995) and ldquoentry to visual short-term memoryrdquo (Bundesen 1990Duncan amp Humphreys 1989) However these assumptions do not imply thatthe selection mechanism itself is located in the ventral stream only Insteadseveral theories have suggested a central role of thedorsal stream incontrollingtheattentional mechanism sometimes calledthe ldquospatial attentionmechanismrdquo(eg La Berge amp Brown 1989 Posner amp Petersen 1990 Schneider 1995 vander Heijden 1992)

Compared to the large body of theoretical work on the relationship betweenattention and (ventral) perceptual processing there are scant data on the role of

REACHING AND ATTENTION 83

visual attention in dorsal processing more precisely the role of attention inspatial-motor control Allport (1987) and Neumann (1987) suggested thatspatial motor actions such as grasping one object from among other objectsmay also be a selection process what Allport (1987) called ldquoselection-for-actionrdquo Natural environments usually contain several objects and only one ofthemshouldbe usedas thetarget foranindividual action Forinstance graspinga pen among other pens requires the motor system to receive spatial informa-tion probably in arm-centred coordinates (Graziano amp Gross 1994) of theintended pen only Information from other pens has to be excluded fromcontrolling the grasping action In other words an attentional mechanism isneeded that selects the spatial information of the movement target Becausespatial information is provided by the visual system (the dorsal pathway)Allport (1987 1989) and Neumann (1987 1990) have suggested that visualattention is involved in this selection process Another example of selection-for-spatial-motor-action refers to the control of saccadic eye movementsBefore each saccade the next fixation point has to be selected among manypotential candidates in the environment

Unfortunately there has not been much experimental workon selection-for-spatial-motor-action Tipper Lortie and Baylis (1992) investigated the role ofvisual attention for manual reaching in an interference paradigm They wishedto determine if the interference effects found for ventral visual processing (egEriksen amp Eriksen 1974) can also be obtained for spatial-motor actions Thedegree of interference is usually considered as a measure of the efficiency ofattentional processes In these experiments subjects had to reach as fast andas precisely as possible from a starting position to one of nine locationsindicated by a red light (thetarget) In some trials a yellow light (thedistractor)appeared simultaneously with the red target light at a different locationSubstantial interference effects were obtained response latencies were pro-longed compared totrials where nodistractor appeared This interference effectwas only observedwhenthedistractorwas locatedbetweenthestarting positionand the target Tipper et al (1992) argued that their results reflect ldquoaction-cen-tred attentionrdquo emphasizing that the location of the movement target is mostrelevant to the amount of interference In summary these results show thatinterference effects by nearby objects can also be obtained for spatial-motoraction such as reaching suggesting that visual attention processes are alsoinvolved in selection-for-spatial-motor-action A similar conclusion wasreached by Castiello (1996) In one of his experiments subjects had to grasp atarget as their primary task A secondary non-spatial task was required for adifferent object located close to the target Castiello observed interferenceeffects of the secondary task on the kinematics of the primary graspingmovement given the subject performed a subsidiary task which involved thedistractor

84 DEUBEL ET AL

Another line of research dealing with dorsal selection concerns the relation-ship between eye movement control and visual attention The question iswhether visual attention for perceptual processing on the one hand and selec-tion of a target for a saccade on the other are independent or not The resultsof early experiments on this issue were controversial (eg Klein 1980 Posner1980) partly due to methodological problems (see Shepherd Findlay ampHockey 1986) More recent studies (Deubel amp Schneider 1996 Hoffman ampSubramaniam 1995 Kowler Anderson Dosher amp Blaser 1995 Schneider ampDeubel 1995) have clearly demonstrated a strict link between ventral selec-tion-for-perception and dorsal selection-for-a-saccade

In the experiments of Deubel and Schneider (1996) subjects had to saccadeto locations within horizontal letter strings left or right of a central fixationcross The performance in discriminating between the ldquoE rdquo and ldquo$ rdquo presentedtachistoscopically before the saccade within the surrounding distractors wastaken as a measure of visual attention in perception The results showed thatdiscrimination performance is best when discrimination target and saccadetargetrefer tothesameobject Thefindings argueforanobligatory andselectivecoupling of dorsal processing for saccade programming and ventral processingfor perception and discrimination this coupling is restricted to one commontarget object at a time

Based on these results and other computational considerations Schneider(1995) postulated a Visual Attention Model (VAM) that suggests a commonselectionmechanism forbothprocessing streams Inlinewithtwo-stagemodelsof perception and attention (Neisser 1967) a first stage of low-level visualprocessing computes in parallel in early visual areas of thebrain (eg V1 V2)elementary visual information in the form of ldquoprimitiverdquo object structures(visual units) Higher-level visual processing in the dorsal and ventral streamis assumed to be capacity-limited that is it occurs only for one visual unit (oneldquoobjectrdquo) at a time In the model visual attention is the mechanism thatdetermines the unit carries out the selection and gates the information flowfrom low- to high-level vision in a way that only information from one objectis furtherprocessed TheVAM claims thatvisual attentionselects one low-levelvisual objectata time leading toprioritizedperceptual processing intheventralstream (ie the object is recognized) Simultaneously possible spatial-motoractions (saccade pointing reaching grasping etc) towards this object areprogrammed in the dorsal stream Only the (effector-specific) ldquogordquo signal isnecessary to convert the programs into overt action

Such attention-mediated and object-specific coupling of dorsal and ventralprocessing has already been demonstrated for eye movement control andperceptual selection (Deubel amp Schneider 1996) More than just for saccadeshowever VAM predicts that the same coupling should also hold for aimingreaching and grasping (Schneider 1995 p 363) In the present study we

REACHING AND ATTENTION 85

analysed the coupling of reaching movements and visual discrimination Forthis purpose a dual-task paradigm similar to that used in our previous studieswas developed The primary task was to make a goal-directed reaching move-ment to a cued object measuring selection-for-spatial-motor-action in thedorsal stream Prior to the movement a secondary task required subjects todiscriminate between the characters ldquoE rdquo and ldquo$ rdquo measuring selection-for-per-ception (ldquotraditionalrdquo visual attention) in the ventral stream It is hypothesizedthat the programming of the reaching movement yokes the visual attentionmechanism so that during this selection process no other object can beprocessed in high-level ventral vision Consequently discrimination perform-ance should be best when discrimination target and reaching target refer to thesame object Fornon-corresponding reaching and discrimination targets betterthan chance performance should be possible only when visual attention shiftsfirst to the discrimination target and then to the reaching target In this caselonger initiation latencies for the movement should be expected

METHODS

Subjects

Five subjects participated in the experiments their age ranged from 22 to 28years They had normal vision and normal motor behaviour All subjects wereexperienced in a variety of experiments in oculomotor research One subjectwas one of the authors of the study the others were naive with respect to theaim of the experiments

Experimental Set-up

Figure 1 shows a sketch of the experimental set-up The subject was seated ina dimly lit room The visual stimuli were presented on a fast 21 inch colourmonitor (CONRAC 7550 C21) visible through a one-way mirror The monitorprovided a frame frequency of 100 Hz at a spatial resolution of 64 pixels perinch The active screen size was 40 times 30 cm theviewing distance was 577 cmThe video signals were generated by a freely programmable graphics board(Kontron KONTRAST 8000) controlled by a PC via the TIGA (Texas Instru-ments Graphics Adapter) interface The stimuli appeared on a grey backgroundadjusted to a mean luminance of 22 cdm2 The luminance of the stimuli was23 cdm2 The relatively high background brightness is essential to avoid theeffects of phosphor persistence (Wolf and Deubel 1997)

The use of a one-way mirror allowed free hand movements to the stimuliwithout visual feedback about hand position Reaching movements were re-corded with a Fastrak electromagnetic position and orientation measuring

86 DEUBEL ET AL

system (Polhemus Inc 1993) and sampled at 400 Hz The sender device wasfixed 60 cm in front of the subject The sender emits time-multiplexedorthogonal electromagnetic fields of 10 kHz frequency From induction in thereceiver which was mounted on the fingertip of the subjectrsquos right hand theorientation relative to the sender device is calculated by a central processingunit From the intensity of the electromagnetic field the distance betweensender and receiver is determined The position in space is calculated fromdistance and orientation by use of a specific digital signal processor(TI320C30) The device allows for a maximum translation range of 10 feetwith an accuracy of 003 inches RMS The frequency response is 120 Hzwithout further filtering the phase lag response is 4 msec Connected on thereceiver was a red LED (5 mm diameter) controlled by the PC The LEDallowed us to provide controlled visual feedback about the spatial position ofthe fingertip

Eye fixationwas monitoredby aninfraredeyetracker(IRIS SkalarMedical)with a temporal bandwidth of 240 Hz This device measures the reflectiondifference between the sclera and iris by infrared LEDs and phototransistorsthat are situated next to the subjectrsquos eyes Head movements were restricted byan adjustable chin rest The experiments were controlled by a 486 PC The PCalso served for the automatic off-line analysis of the pointing movement datafor which movement latencies and start and end positions of the manualresponses were determined

FIG 1 Experimental apparatus

REACHING AND ATTENTION 87

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

visual attention in dorsal processing more precisely the role of attention inspatial-motor control Allport (1987) and Neumann (1987) suggested thatspatial motor actions such as grasping one object from among other objectsmay also be a selection process what Allport (1987) called ldquoselection-for-actionrdquo Natural environments usually contain several objects and only one ofthemshouldbe usedas thetarget foranindividual action Forinstance graspinga pen among other pens requires the motor system to receive spatial informa-tion probably in arm-centred coordinates (Graziano amp Gross 1994) of theintended pen only Information from other pens has to be excluded fromcontrolling the grasping action In other words an attentional mechanism isneeded that selects the spatial information of the movement target Becausespatial information is provided by the visual system (the dorsal pathway)Allport (1987 1989) and Neumann (1987 1990) have suggested that visualattention is involved in this selection process Another example of selection-for-spatial-motor-action refers to the control of saccadic eye movementsBefore each saccade the next fixation point has to be selected among manypotential candidates in the environment

Unfortunately there has not been much experimental workon selection-for-spatial-motor-action Tipper Lortie and Baylis (1992) investigated the role ofvisual attention for manual reaching in an interference paradigm They wishedto determine if the interference effects found for ventral visual processing (egEriksen amp Eriksen 1974) can also be obtained for spatial-motor actions Thedegree of interference is usually considered as a measure of the efficiency ofattentional processes In these experiments subjects had to reach as fast andas precisely as possible from a starting position to one of nine locationsindicated by a red light (thetarget) In some trials a yellow light (thedistractor)appeared simultaneously with the red target light at a different locationSubstantial interference effects were obtained response latencies were pro-longed compared totrials where nodistractor appeared This interference effectwas only observedwhenthedistractorwas locatedbetweenthestarting positionand the target Tipper et al (1992) argued that their results reflect ldquoaction-cen-tred attentionrdquo emphasizing that the location of the movement target is mostrelevant to the amount of interference In summary these results show thatinterference effects by nearby objects can also be obtained for spatial-motoraction such as reaching suggesting that visual attention processes are alsoinvolved in selection-for-spatial-motor-action A similar conclusion wasreached by Castiello (1996) In one of his experiments subjects had to grasp atarget as their primary task A secondary non-spatial task was required for adifferent object located close to the target Castiello observed interferenceeffects of the secondary task on the kinematics of the primary graspingmovement given the subject performed a subsidiary task which involved thedistractor

84 DEUBEL ET AL

Another line of research dealing with dorsal selection concerns the relation-ship between eye movement control and visual attention The question iswhether visual attention for perceptual processing on the one hand and selec-tion of a target for a saccade on the other are independent or not The resultsof early experiments on this issue were controversial (eg Klein 1980 Posner1980) partly due to methodological problems (see Shepherd Findlay ampHockey 1986) More recent studies (Deubel amp Schneider 1996 Hoffman ampSubramaniam 1995 Kowler Anderson Dosher amp Blaser 1995 Schneider ampDeubel 1995) have clearly demonstrated a strict link between ventral selec-tion-for-perception and dorsal selection-for-a-saccade

In the experiments of Deubel and Schneider (1996) subjects had to saccadeto locations within horizontal letter strings left or right of a central fixationcross The performance in discriminating between the ldquoE rdquo and ldquo$ rdquo presentedtachistoscopically before the saccade within the surrounding distractors wastaken as a measure of visual attention in perception The results showed thatdiscrimination performance is best when discrimination target and saccadetargetrefer tothesameobject Thefindings argueforanobligatory andselectivecoupling of dorsal processing for saccade programming and ventral processingfor perception and discrimination this coupling is restricted to one commontarget object at a time

Based on these results and other computational considerations Schneider(1995) postulated a Visual Attention Model (VAM) that suggests a commonselectionmechanism forbothprocessing streams Inlinewithtwo-stagemodelsof perception and attention (Neisser 1967) a first stage of low-level visualprocessing computes in parallel in early visual areas of thebrain (eg V1 V2)elementary visual information in the form of ldquoprimitiverdquo object structures(visual units) Higher-level visual processing in the dorsal and ventral streamis assumed to be capacity-limited that is it occurs only for one visual unit (oneldquoobjectrdquo) at a time In the model visual attention is the mechanism thatdetermines the unit carries out the selection and gates the information flowfrom low- to high-level vision in a way that only information from one objectis furtherprocessed TheVAM claims thatvisual attentionselects one low-levelvisual objectata time leading toprioritizedperceptual processing intheventralstream (ie the object is recognized) Simultaneously possible spatial-motoractions (saccade pointing reaching grasping etc) towards this object areprogrammed in the dorsal stream Only the (effector-specific) ldquogordquo signal isnecessary to convert the programs into overt action

Such attention-mediated and object-specific coupling of dorsal and ventralprocessing has already been demonstrated for eye movement control andperceptual selection (Deubel amp Schneider 1996) More than just for saccadeshowever VAM predicts that the same coupling should also hold for aimingreaching and grasping (Schneider 1995 p 363) In the present study we

REACHING AND ATTENTION 85

analysed the coupling of reaching movements and visual discrimination Forthis purpose a dual-task paradigm similar to that used in our previous studieswas developed The primary task was to make a goal-directed reaching move-ment to a cued object measuring selection-for-spatial-motor-action in thedorsal stream Prior to the movement a secondary task required subjects todiscriminate between the characters ldquoE rdquo and ldquo$ rdquo measuring selection-for-per-ception (ldquotraditionalrdquo visual attention) in the ventral stream It is hypothesizedthat the programming of the reaching movement yokes the visual attentionmechanism so that during this selection process no other object can beprocessed in high-level ventral vision Consequently discrimination perform-ance should be best when discrimination target and reaching target refer to thesame object Fornon-corresponding reaching and discrimination targets betterthan chance performance should be possible only when visual attention shiftsfirst to the discrimination target and then to the reaching target In this caselonger initiation latencies for the movement should be expected

METHODS

Subjects

Five subjects participated in the experiments their age ranged from 22 to 28years They had normal vision and normal motor behaviour All subjects wereexperienced in a variety of experiments in oculomotor research One subjectwas one of the authors of the study the others were naive with respect to theaim of the experiments

Experimental Set-up

Figure 1 shows a sketch of the experimental set-up The subject was seated ina dimly lit room The visual stimuli were presented on a fast 21 inch colourmonitor (CONRAC 7550 C21) visible through a one-way mirror The monitorprovided a frame frequency of 100 Hz at a spatial resolution of 64 pixels perinch The active screen size was 40 times 30 cm theviewing distance was 577 cmThe video signals were generated by a freely programmable graphics board(Kontron KONTRAST 8000) controlled by a PC via the TIGA (Texas Instru-ments Graphics Adapter) interface The stimuli appeared on a grey backgroundadjusted to a mean luminance of 22 cdm2 The luminance of the stimuli was23 cdm2 The relatively high background brightness is essential to avoid theeffects of phosphor persistence (Wolf and Deubel 1997)

The use of a one-way mirror allowed free hand movements to the stimuliwithout visual feedback about hand position Reaching movements were re-corded with a Fastrak electromagnetic position and orientation measuring

86 DEUBEL ET AL

system (Polhemus Inc 1993) and sampled at 400 Hz The sender device wasfixed 60 cm in front of the subject The sender emits time-multiplexedorthogonal electromagnetic fields of 10 kHz frequency From induction in thereceiver which was mounted on the fingertip of the subjectrsquos right hand theorientation relative to the sender device is calculated by a central processingunit From the intensity of the electromagnetic field the distance betweensender and receiver is determined The position in space is calculated fromdistance and orientation by use of a specific digital signal processor(TI320C30) The device allows for a maximum translation range of 10 feetwith an accuracy of 003 inches RMS The frequency response is 120 Hzwithout further filtering the phase lag response is 4 msec Connected on thereceiver was a red LED (5 mm diameter) controlled by the PC The LEDallowed us to provide controlled visual feedback about the spatial position ofthe fingertip

Eye fixationwas monitoredby aninfraredeyetracker(IRIS SkalarMedical)with a temporal bandwidth of 240 Hz This device measures the reflectiondifference between the sclera and iris by infrared LEDs and phototransistorsthat are situated next to the subjectrsquos eyes Head movements were restricted byan adjustable chin rest The experiments were controlled by a 486 PC The PCalso served for the automatic off-line analysis of the pointing movement datafor which movement latencies and start and end positions of the manualresponses were determined

FIG 1 Experimental apparatus

REACHING AND ATTENTION 87

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

Another line of research dealing with dorsal selection concerns the relation-ship between eye movement control and visual attention The question iswhether visual attention for perceptual processing on the one hand and selec-tion of a target for a saccade on the other are independent or not The resultsof early experiments on this issue were controversial (eg Klein 1980 Posner1980) partly due to methodological problems (see Shepherd Findlay ampHockey 1986) More recent studies (Deubel amp Schneider 1996 Hoffman ampSubramaniam 1995 Kowler Anderson Dosher amp Blaser 1995 Schneider ampDeubel 1995) have clearly demonstrated a strict link between ventral selec-tion-for-perception and dorsal selection-for-a-saccade

In the experiments of Deubel and Schneider (1996) subjects had to saccadeto locations within horizontal letter strings left or right of a central fixationcross The performance in discriminating between the ldquoE rdquo and ldquo$ rdquo presentedtachistoscopically before the saccade within the surrounding distractors wastaken as a measure of visual attention in perception The results showed thatdiscrimination performance is best when discrimination target and saccadetargetrefer tothesameobject Thefindings argueforanobligatory andselectivecoupling of dorsal processing for saccade programming and ventral processingfor perception and discrimination this coupling is restricted to one commontarget object at a time

Based on these results and other computational considerations Schneider(1995) postulated a Visual Attention Model (VAM) that suggests a commonselectionmechanism forbothprocessing streams Inlinewithtwo-stagemodelsof perception and attention (Neisser 1967) a first stage of low-level visualprocessing computes in parallel in early visual areas of thebrain (eg V1 V2)elementary visual information in the form of ldquoprimitiverdquo object structures(visual units) Higher-level visual processing in the dorsal and ventral streamis assumed to be capacity-limited that is it occurs only for one visual unit (oneldquoobjectrdquo) at a time In the model visual attention is the mechanism thatdetermines the unit carries out the selection and gates the information flowfrom low- to high-level vision in a way that only information from one objectis furtherprocessed TheVAM claims thatvisual attentionselects one low-levelvisual objectata time leading toprioritizedperceptual processing intheventralstream (ie the object is recognized) Simultaneously possible spatial-motoractions (saccade pointing reaching grasping etc) towards this object areprogrammed in the dorsal stream Only the (effector-specific) ldquogordquo signal isnecessary to convert the programs into overt action

Such attention-mediated and object-specific coupling of dorsal and ventralprocessing has already been demonstrated for eye movement control andperceptual selection (Deubel amp Schneider 1996) More than just for saccadeshowever VAM predicts that the same coupling should also hold for aimingreaching and grasping (Schneider 1995 p 363) In the present study we

REACHING AND ATTENTION 85

analysed the coupling of reaching movements and visual discrimination Forthis purpose a dual-task paradigm similar to that used in our previous studieswas developed The primary task was to make a goal-directed reaching move-ment to a cued object measuring selection-for-spatial-motor-action in thedorsal stream Prior to the movement a secondary task required subjects todiscriminate between the characters ldquoE rdquo and ldquo$ rdquo measuring selection-for-per-ception (ldquotraditionalrdquo visual attention) in the ventral stream It is hypothesizedthat the programming of the reaching movement yokes the visual attentionmechanism so that during this selection process no other object can beprocessed in high-level ventral vision Consequently discrimination perform-ance should be best when discrimination target and reaching target refer to thesame object Fornon-corresponding reaching and discrimination targets betterthan chance performance should be possible only when visual attention shiftsfirst to the discrimination target and then to the reaching target In this caselonger initiation latencies for the movement should be expected

METHODS

Subjects

Five subjects participated in the experiments their age ranged from 22 to 28years They had normal vision and normal motor behaviour All subjects wereexperienced in a variety of experiments in oculomotor research One subjectwas one of the authors of the study the others were naive with respect to theaim of the experiments

Experimental Set-up

Figure 1 shows a sketch of the experimental set-up The subject was seated ina dimly lit room The visual stimuli were presented on a fast 21 inch colourmonitor (CONRAC 7550 C21) visible through a one-way mirror The monitorprovided a frame frequency of 100 Hz at a spatial resolution of 64 pixels perinch The active screen size was 40 times 30 cm theviewing distance was 577 cmThe video signals were generated by a freely programmable graphics board(Kontron KONTRAST 8000) controlled by a PC via the TIGA (Texas Instru-ments Graphics Adapter) interface The stimuli appeared on a grey backgroundadjusted to a mean luminance of 22 cdm2 The luminance of the stimuli was23 cdm2 The relatively high background brightness is essential to avoid theeffects of phosphor persistence (Wolf and Deubel 1997)

The use of a one-way mirror allowed free hand movements to the stimuliwithout visual feedback about hand position Reaching movements were re-corded with a Fastrak electromagnetic position and orientation measuring

86 DEUBEL ET AL

system (Polhemus Inc 1993) and sampled at 400 Hz The sender device wasfixed 60 cm in front of the subject The sender emits time-multiplexedorthogonal electromagnetic fields of 10 kHz frequency From induction in thereceiver which was mounted on the fingertip of the subjectrsquos right hand theorientation relative to the sender device is calculated by a central processingunit From the intensity of the electromagnetic field the distance betweensender and receiver is determined The position in space is calculated fromdistance and orientation by use of a specific digital signal processor(TI320C30) The device allows for a maximum translation range of 10 feetwith an accuracy of 003 inches RMS The frequency response is 120 Hzwithout further filtering the phase lag response is 4 msec Connected on thereceiver was a red LED (5 mm diameter) controlled by the PC The LEDallowed us to provide controlled visual feedback about the spatial position ofthe fingertip

Eye fixationwas monitoredby aninfraredeyetracker(IRIS SkalarMedical)with a temporal bandwidth of 240 Hz This device measures the reflectiondifference between the sclera and iris by infrared LEDs and phototransistorsthat are situated next to the subjectrsquos eyes Head movements were restricted byan adjustable chin rest The experiments were controlled by a 486 PC The PCalso served for the automatic off-line analysis of the pointing movement datafor which movement latencies and start and end positions of the manualresponses were determined

FIG 1 Experimental apparatus

REACHING AND ATTENTION 87

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

analysed the coupling of reaching movements and visual discrimination Forthis purpose a dual-task paradigm similar to that used in our previous studieswas developed The primary task was to make a goal-directed reaching move-ment to a cued object measuring selection-for-spatial-motor-action in thedorsal stream Prior to the movement a secondary task required subjects todiscriminate between the characters ldquoE rdquo and ldquo$ rdquo measuring selection-for-per-ception (ldquotraditionalrdquo visual attention) in the ventral stream It is hypothesizedthat the programming of the reaching movement yokes the visual attentionmechanism so that during this selection process no other object can beprocessed in high-level ventral vision Consequently discrimination perform-ance should be best when discrimination target and reaching target refer to thesame object Fornon-corresponding reaching and discrimination targets betterthan chance performance should be possible only when visual attention shiftsfirst to the discrimination target and then to the reaching target In this caselonger initiation latencies for the movement should be expected

METHODS

Subjects

Five subjects participated in the experiments their age ranged from 22 to 28years They had normal vision and normal motor behaviour All subjects wereexperienced in a variety of experiments in oculomotor research One subjectwas one of the authors of the study the others were naive with respect to theaim of the experiments

Experimental Set-up

Figure 1 shows a sketch of the experimental set-up The subject was seated ina dimly lit room The visual stimuli were presented on a fast 21 inch colourmonitor (CONRAC 7550 C21) visible through a one-way mirror The monitorprovided a frame frequency of 100 Hz at a spatial resolution of 64 pixels perinch The active screen size was 40 times 30 cm theviewing distance was 577 cmThe video signals were generated by a freely programmable graphics board(Kontron KONTRAST 8000) controlled by a PC via the TIGA (Texas Instru-ments Graphics Adapter) interface The stimuli appeared on a grey backgroundadjusted to a mean luminance of 22 cdm2 The luminance of the stimuli was23 cdm2 The relatively high background brightness is essential to avoid theeffects of phosphor persistence (Wolf and Deubel 1997)

The use of a one-way mirror allowed free hand movements to the stimuliwithout visual feedback about hand position Reaching movements were re-corded with a Fastrak electromagnetic position and orientation measuring

86 DEUBEL ET AL

system (Polhemus Inc 1993) and sampled at 400 Hz The sender device wasfixed 60 cm in front of the subject The sender emits time-multiplexedorthogonal electromagnetic fields of 10 kHz frequency From induction in thereceiver which was mounted on the fingertip of the subjectrsquos right hand theorientation relative to the sender device is calculated by a central processingunit From the intensity of the electromagnetic field the distance betweensender and receiver is determined The position in space is calculated fromdistance and orientation by use of a specific digital signal processor(TI320C30) The device allows for a maximum translation range of 10 feetwith an accuracy of 003 inches RMS The frequency response is 120 Hzwithout further filtering the phase lag response is 4 msec Connected on thereceiver was a red LED (5 mm diameter) controlled by the PC The LEDallowed us to provide controlled visual feedback about the spatial position ofthe fingertip

Eye fixationwas monitoredby aninfraredeyetracker(IRIS SkalarMedical)with a temporal bandwidth of 240 Hz This device measures the reflectiondifference between the sclera and iris by infrared LEDs and phototransistorsthat are situated next to the subjectrsquos eyes Head movements were restricted byan adjustable chin rest The experiments were controlled by a 486 PC The PCalso served for the automatic off-line analysis of the pointing movement datafor which movement latencies and start and end positions of the manualresponses were determined

FIG 1 Experimental apparatus

REACHING AND ATTENTION 87

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

system (Polhemus Inc 1993) and sampled at 400 Hz The sender device wasfixed 60 cm in front of the subject The sender emits time-multiplexedorthogonal electromagnetic fields of 10 kHz frequency From induction in thereceiver which was mounted on the fingertip of the subjectrsquos right hand theorientation relative to the sender device is calculated by a central processingunit From the intensity of the electromagnetic field the distance betweensender and receiver is determined The position in space is calculated fromdistance and orientation by use of a specific digital signal processor(TI320C30) The device allows for a maximum translation range of 10 feetwith an accuracy of 003 inches RMS The frequency response is 120 Hzwithout further filtering the phase lag response is 4 msec Connected on thereceiver was a red LED (5 mm diameter) controlled by the PC The LEDallowed us to provide controlled visual feedback about the spatial position ofthe fingertip

Eye fixationwas monitoredby aninfraredeyetracker(IRIS SkalarMedical)with a temporal bandwidth of 240 Hz This device measures the reflectiondifference between the sclera and iris by infrared LEDs and phototransistorsthat are situated next to the subjectrsquos eyes Head movements were restricted byan adjustable chin rest The experiments were controlled by a 486 PC The PCalso served for the automatic off-line analysis of the pointing movement datafor which movement latencies and start and end positions of the manualresponses were determined

FIG 1 Experimental apparatus

REACHING AND ATTENTION 87

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

Calibration and Data Analysis

Each session started with calibration of the eyetracker the subject having tosequentially fixate three positions arranged on a horizontal line at distances of85deg Also the origin and coordinate alignment frame of the position sensorwere set relative to the projected position of the monitorrsquos centre The positionsensor behaved linearly within 30 cm around the central position The overallaccuracy was better than 2 mm To determine latency amplitude and durationof the reaching movements an off-line program for evaluation of movementtrajectory parameters searched the movement record for the transgression andsubgression of a vectorial velocity threshold of 10 mms (which is equivalentto about 1degsec) The beginning and the end of the reaching movement werecalculated as linear regressions in a 200 msec time window around thesepoints

Experimental Paradigm

After an initial training block that was not included in the data analysis eachsubject underwent six blocks (three blocks per day) of each of theexperimentseach block consisted of 120 single trials The subject performed a dual taskinvolving both manual reaching and visual discrimination In each experimen-tal trial the reaching movement was guided by a central symbolic cue thatindicated the movement target (MT) within a string of letters Moreover thesubject had to report the identity of a discrimination target (DT) presentedtachistoscopically in the string Two experiments were performed In Experi-ment 1 the DT appeared before the hand movement For each experimentalblock the position of the DT was held constant either on the right or on theleft and on the central position of the string Experiment 2 was similar toExperiment 1 except that the DT was presented at the onset of the reachingmovement

Figure 2 shows an example for the sequence of stimuli in a single trial ofExperiment 1 Each trial started with the presentation of a small fixation crossin the centre of the screen with a size of 025deg Simultaneously two strings ofpre-mask characters appeared to the left and right of the central fixation eachconsisting of five pre-mask items resembling the number ldquoI$ rdquo The width ofeach item was 09deg of visual angle its height was 14deg The distance betweenthe items was 24deg with the central item of the five letters being presented atan eccentricity of 765deg The three central items of each letter string appearedon ellipses coloured red (r) green (g) andblue (b) as indicated inFig 2 Colourintensities of the ellipses were adjusted by flicker-photometry to make themequally salient

The subject was asked to maintain strict fixation at the centre of the screeninitially indicated by a central fixation cross throughout the trial Maintenance

88 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

of fixation was monitored by the IRIS oculometer At the beginning of thetrialthe subject had to position his or her fingertip on the location of the centralcross The position of the fingertip is indicated by the arrowhead in Fig 2 Inthis phase the LED was switched on aiding precise positioning After a delayof 1000ndash1600 msec a symbolic cue in the form of a red green or blue triangleappeared in the centre of the screen pointing either to the right or to the leftside Colour and pointing direction of the triangle thus unequivocally indicateda specific item the movement target (MT) within the string The primary taskwas to ldquopoint to this target as fast and precisely as possiblerdquo Simultaneouslywith cue onset the LEDwas switched off todisable any furthervisual feedbackof hand or pointing position Then 150 msec after the appearance of the cuewell before the onset of the pointing movement the pre-mask characterschanged into nine distractors and one discrimination target The distractors

FIG 2 Stimulus sequence in Experiment 1 The trial starts with the presentation of a small fixationcross and two strings of characters to the left and right of the central fixation The three central itemsof each letter string appear on ellipses coloured red (r) green (g) and blue (b) Initially the subjectpositions his or her fingertip on the location of the central cross (fingertip position is indicated by thearrowhead) Aftera delayof 1ndash16 sec a symbolic cue intheformof aredgreenorbluetriangleappearsin the centreof the screen pointing eitherto the rightor to the leftside this cue specifies the movementtarget within the string Then 150 msec later the pre-maskcharacters change intonine distractors andonediscriminationtarget(ldquoE rdquo orldquo$ rdquo) The targetand distractors remainvisible for 150 msec Then thecharacters and the central cue are removedand only the coloured ellipses remain

REACHING AND ATTENTION 89

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

were randomly selected among the characters ldquo rdquo and ldquo rdquo The centralcharacter on one of both sides was replaced by the discrimination target (DT)which consisted either of the letter ldquoE rdquo or its mirror image (ldquo$ rdquo) The positionof theDT was constant during each block and known to the subject (eg centralposition of the DT was constant during each block and known to the subject(eg central position in the string on the right side) The movement targetpositions however were varied independently within the central three itemsof the strings resulting in 12 combinations of movement target and discrimi-nation target positions All experimental conditions occurred with equal prob-ability The target and distractors remained visible for 150 msec Then theitems and the central cue were removed and only the coloured ellipses re-mained

Due to the timing of the stimulus presentation the discrimination target wasno longer present 300 msec after the appearance of the coloured triangle As aresult most reaching movements were initiated well after the disappearance oftarget and distractors (see Figure 5) To eliminate occasional responses thatoccurred too early the off-line data analysis discarded movements withlatencies shorter than 200 msec Also trials with movement velocities smallerthan 11 mms2 and durations shorter than 50 msec and longer than 600 msecwere not considered in the analysis This accounted for less than 2 of alltrials

One secondafter theonset of the reaching movement theLEDwas switchedon again to enable control of visual feedback of the finger position reachedFinally thesubject indicated withouttimepressure theidentity of thediscrimi-nation target (ldquoE rdquo or ldquo$ rdquo) by pressing one of two buttons (2AFC task) Thecentral fixation cross reappeared after the subjectrsquos decision and the next trialwas initiated by the computer

In separate sessions two types of ldquosingle-taskrdquo controls were run A firstcontrol task (ldquono discriminationndashreaching onlyrdquo single-task condition) servedto assess pointing reaction times in a single-task situation For this purpose thesubject was asked to point to the indicated position but was not required todiscriminate Asecond control task (ldquono reachingndashdiscrimination onlyrdquo single-task condition) served to test discrimination performance without pointingHere the subject was only asked to indicate the identity of the discriminationtarget no reach was required Each subject performed two blocks of eachcontrol task

Experiment 2 was very similar to Experiment 1 except that the presentationof the discrimination stimulus occurred at the onset of the reaching movementFor this purpose the computer performed an on-line calculation of movementvelocity Stimulus presentation was triggered when the velocity exceeded athreshold of 1degsec

90 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

RESULTS

Experiment 1

Movement Performance After the initial training block all five subjectswere able to produce reaching movements with surprisingly consistent accu-racy and latency Figure 3 gives examples of several manual responses fromone of the subjects The graph displays the registered finger position as afunction of time for the different movement target eccentricities It can be seenfrom the raw data that the end positions of the movements correlate well withtheMT positions Some of the responses showeda small overshootwithrespectto the movement end position The amplitude data reported in the followingrefer only to the final movement position Moreover the movements were ingeneral very consistent with respect to their velocity profiles only a fewmovements with multiple velocity peaks were observed

The impression of the homogeneity of movement responses is confirmed byanalysis of the movement data Figure 4a shows mean movement amplitudesand Figure 4b mean movement durations as a function of the movement targetlocation The vertical bars denote the standard error they are only visible forthe cases where the error exceeds symbol size The data are plotted separatelyfor the cases where the discrimination stimulus was present at the centralposition on the right (open circles) and on the left (solid circles) It is easy tosee that the amplitudes are independent of the position of the discriminationtarget One rationale of the experimental approach was that the discriminationtask should not interfere with the reaching task this analysis of amplitudessuggests that this was indeed the case Moreover the mean movement ampli-tudes demonstrate that the reaching movements were very precise meanamplitudes were highly correlated with the given MT positions (r = 099) Afurther data analysis in the form of a two-way ANOVA (repeated measures)confirmed a highly significant main effect of MT position F(520) = 1078 anon-significanteffectof DTposition F(14) = 09 p gt 1 anda non-significantinteraction F(520) = 089

Asimilarconclusion holds forthe movement durations (Figure 4b) Averagemovement durations were 202 260 and 315 msec for the small medium andlarge target eccentricities respectively Again the data are independent of DTlocation suggesting that the execution of the movement itself is not affectedby the presentation of the test item Accordingly ANOVA showed a highlysignificantmaineffectof MTposition F(520) = 2637 anon-significanteffectof DTposition F(14) = 044 and anon-significant interaction F(520) = 080

Figure 5a displays mean movement onset latencies and standard errors as afunction of MT location Again the data are given separately for the blockswhere the discrimination target was on the right (open circles) and where DT

REACHING AND ATTENTION 91

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

FIG 3 Timecourses of manual reachingresponses are measuredwith the PolhemusFastracksystemThe graph shows examples of reaching movements from one subject and for the various movementtarget eccentricities

92

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

FIG 4 (a) Mean movementamplitudes as a function of the movementtarget location in Experiment1 Vertical bars denotestandarderrors Dataareplottedseparately for thecases wherethediscriminationstimulus was present at the central position on the right (open circles) and on the left (solid circles) (b)Movement durations

(a)

(b)

93

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

FIG 5 (a) Mean movementonset latencies and standard errors as a function of MT location Dataare given separately for the blocks where the discrimination target was on the right (open circles) andon the left (solid circles) Opentrianglesdisplay the latency datafrom the ldquono discriminationndash reachingonlyrdquo single-task control condition (b) Histograms of the latency distribution presented individuallyfor the five subjects

(a)

(b)

94

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

was on the left (solid circles) Mean movement onset latency averaged over allconditions was 4378 msec A two-way ANOVA revealed that the latenciesdepended neither on MT location F(520) = 074 nor on DT location F(14)= 0 Also the interaction was not significant F(520) = 21 p gt 05 The opentriangles in the graph display the latency data from the ldquono discrimina-tionndashreaching onlyrdquo single-task control condition For this type of experimentmean latency was 4369 msec Again the response latency was independent ofMT location F(520) = 134 p gt 1

Figure 5b shows histograms of the distribution of the movement onsetlatencies individually for the five subjects who participated in the experimentIt can be seen that while mean latency varies the distributions for all subjectsare unimodal and are skewed with the long tail towards longer latencies

Perceptual Performance The subjects reported that they had no difficul-ties pointing quickly totheindicatedtarget iteminthestring However initiallythey were very uncertain about their ability to discriminate between the DTitems Performance improved considerably after some practice Therefore thefirst session served for training and was not included in the data analysis Aftertheexperiment the subjects were askedfor their subjective impression andhowthey solved the task They reported that theperipheral items that were indicatedas movement targets seemed to ldquolight uprdquo in a row in an almost unstructuredvisual field They also had the impression that they could identify the distractor(ldquo rdquo or ldquo rdquo) exactly when it appeared at the movement target position

Our indicator for the momentary allocation of attention in the ventral streamis theaccuracy withwhich thediscriminationtarget can be identified Discrimi-nation performance can be expressed as the percentage of correct decisions oftarget identity chance level is 50 correct Figure 6 presents discriminationperformance as a function of movement target location Since performance wasnot significantly different for DT on the left or on the right data from the twoconditions were pooled in Figure 6 such that the position of the discriminationtarget always refers to the position indicated in the graph (at + 765deg) In otherwords negative MT locations refer to the cases where MT and DT were inopposite hemifields

Figure 6a shows discrimination performance as a function of relative MTpositionforall response latencies (solid squares) The horizontal line representsthe discrimination performance from the ldquono reachingndashonly discriminationrdquocontrol task The data suggest that performance depends on the relationshipbetween the position of the discrimination stimulus and the location of theindicated movement target position performance is best when the MT and DTpositions coincide (DT = MT) When the movement is not directed to thecritical item performance decreases sharply Performance is worst when thesubject points to a direction opposite to the DT position The performanceadvantage for the coincidence of MT and DT positions was confirmed by

REACHING AND ATTENTION 95

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

FIG 6 (a) Discrimination performance as a function of movement target location Data for DT onthe left and on the right are pooled such that the position of the discrimination target always refers tothe position indicated in the graph at + 765deg Vertical bars indicate standard errors Horizontal linerepresents discriminationperformance from the ldquoNo reaching ndash only discriminationrdquocontrol trials (b)Discriminationperformance dataafter mediansplit Solidcircles are for the fasthalf of responses opencircles are for the slow half of responses

(a)

(b)

96

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

further statistical analysis ANOVA showed a highly significant effect ofrelative MT position F(520) = 1512 p lt 0001 In a post-hoc Student-New-manndashKeuls test the performance at DT = MT proved to be superior to all othercases which did not differ significantly (p lt 01)

Upon questioning after the experiments subjects occasionally reported thatthey had the feeling that they performed better in the discrimination task whenthey delayed the manual response An interpretation of this observation is thatin these cases DT is discriminated first and only later is movement program-ming initiated This should result in longer movement latencies In other wordsone should expect an interaction between movement latency and perceptualperformance Therefore we analysed performance for each subject separatelyfor the fast half of responses (ie faster than the median latency of the subject)and for the slow half of responses The averaged data are shown in Figure 6bFor the fast responses (solid circles) performance superiority at DT = MT wasstill more pronounced For these fast responses directed to the discriminationstimulus performance was even superior to discrimination performance in theldquono movementrdquo control condition (891 vs 783correct) Forthe slow portionof responses (open circles) the spatial selectivity all but disappeared Com-pared to the fast reactions there was also a general tendency for discriminationto improve in those cases where MT and DT were presented in oppositedirections A two-factor ANOVA showed a significant main effect of relativeMT position F(520) = 1473 p lt 0001 and a non-significant main effect oflatency F(14) = 0 05 As expected the interaction between response latencyand MT position was significant F(520) = 414 p lt 01 Post-hoc Newman-Keuls tests showed that for the fast half of responses performance at MT =DT was significantly better than for the other relative MT positions (p lt 01)For the slow responses the superiority of MT = DT with respect to the otherrelative movement positions disappeared (p gt 05) In summary the data showthat the ability to discriminate between objects in a multi-object scene duringthe preparation of a reaching movement is spatially selective and superior atthe movement goal This is most pronounced for fast manual reactions

Experiment 2

Movement Performance InExperiment2 thepresentationof the discrimi-nation target occurred at the onset of the manual response The mean (plusmn SE)movement onset latency was 4412 plusmn 45 msec Since the characteristics of thelatency data in this experiment were identical tothose of Experiment 1 the dataare not presented in more detail here

In this experiment the discrimination stimulus appeared at movement onsetand was present during most of the movement Therefore the question ariseswhether presence of the DT affected the precision of the reaching movementandor its dynamic properties For this reason we again analysed the depend-

REACHING AND ATTENTION 97

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

ence of movement amplitude and duration on DT location The results areshown in Figure 7 Figure 7a displays movement amplitude as a function ofMT position It can be seen that as in Experiment 1 the overall movement wasrather precise and there was no effect of DT position Accordingly a two-wayANOVA yielded a highly significant main effect of MT position F(520) =4108 a non-significant effect of DT position F(14) = 3 41 p gt 1 and nointeraction F(520) = 141 p gt 1

Figure 7b displays mean movement durations Although there seemed to bea general tendency for movements to be shorter for DT appearing in the righthemifield this effect did not reach statistical significance ANOVA yielded asignificant main effect of MT position F(520) = 2048 p lt 0001 but anon-significant effect of DT position F(14) = 009 and a non-significantinteraction F(520) = 073 In summary as in the previous experiment therewas no indication that the movement itself was affected by the presentation ofthe DT

Perceptual Performance Figure 8 gives discrimination performance inExperiment 2 as a function of the relative position of the movement targetpooled over five subjects In this case also discrimination was superior whenDT and MT referred to the same object Accordingly ANOVA yielded asignificant effect of relative MT position F(45) = 442 p lt 01 A post-hocNewman-Keuls test confirmed a significant difference in the DT = MT condi-tion with respect to the other conditions (p lt 05) All other data points did notdiffer significantly

DISCUSSION

The main aim of this study was to determine if and how selection in the ventralstream (ldquoselection-for-perceptionrdquo) and selection of visual targets for reachingmovements in the dorsal stream (ldquoselection-for-spatial-motor-actionrdquo) are cou-pled This study developed from the theoretical perspective provided by VAM(Schneider 1995) arecently developedmodel of thecontrol of visual attentionand from empirical evidence confirming such coupling in the preparation ofsaccadic eye movements (Deubel amp Schneider 1996 Hoffman amp Subrama-niam 1995 Kowler et al 1995)

VAMstates thatacommon selectionmechanism exists fordorsal andventralprocessing This mechanism is suggested to select one object at a time in theldquoearlyrdquo stages of the visual system resulting in an increased activation of thevisual representations of this object in higher-level ventral and dorsal visualareas This increased activation allows the selective perceptual analysis of theselected object to the level of recognition and the selective computation of itsspatial parameters such that saccading reaching and grasping movements areprepared Therefore VAM suggests a strict one-object-at-a-time rule When-

98 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

FIG 7 (a) Mean movementamplitudes as a function of the movementtarget locationin Experiment2 Vertical bars denotestandarderrors Dataare plottedseparatelyfor thecaseswherethe discriminationstimulus was present at the central position on the right (open circles) andon the left (solid circles) (b)Movementdurations

(a)

(b)

99

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

ever a goal-directed action towards an object is prepared only this movementtarget can be perceptually processed in higher-level ventral areas On the otherhand whenever visual attention focuses on an item for the purpose of objectrecognition no other objects can be selected for goal-directed actions Accord-ing to VAM dissociations can only occur by a serial process implying that thevisual recognition of an object should considerably delay a motor responsetowards a different spatially separate target It should also be emphasized thattheselection is object-specific this is incontrast toothers whoassumeaspatialorganization of attentional selection (eg Hughes amp Zimba 1987 RizzolattiRiggio Dascola amp Umiltagrave 1987)

The results from our experiments are perfectly consistent with these theo-retical conjectures The discrimination data from Experiment 1 demonstratethat well before movement onset perceptual performance depends strongly onwhere in space the reaching movement is directed Discrimination is best whenthe reaching movement and perceptual task refer to the same object and isstrongly reduced prior toa reach when an objectother thanthe movement targethas to be perceptually analysed Our interpretation is that the (dorsally based)preparationof agoal-directedmotorresponse hereareaching movement bindsthe (perceptual) processing capacities of the ventral stream to the same objectDuring the preparation phase objects other than the movement target aretemporarily excluded from ventral high-level visual analysis Similar results

FIG 8 Discriminationperformance as a function of movementtarget locationin Experiment 2 Datafor DT on the left and on the right are pooledsuch that the position of the discriminationtarget alwaysrefer to the position indicated in the graph at + 765deg Vertical bars indicate standard errors

100 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

have been obtained by Irwin and Gordon (this issue) for the coupling ofsaccade programming and encoding of visual-perceptual information intotrans-saccadic memory

The amount of spatial selectivity reflected inour data is surprising It reflectsthe current spatial location of a common selection mechanism for dorsal andventral processing The fact that spatial selectivity was so clear in our experi-ments is probably due to the brief presentation time of the stimuli thuspreventing additional attentional shifts In contrast investigations using reac-tion time paradigms where attentional shifts cannot be excluded often reveala rather broad gradient of attentional effects as a result of cueing (eg Downingamp Pinker 1985)

The object specificity of the coupling is in line with the findings of Castiello(1996) whodeterminedif thekinematics of thetarget movement are influencedby non-target objects Castiellorsquos results indeed demonstrated interactionswhen the distractor object had to be used also for carrying out a simultaneoussecondary task However interference disappeared when thesecondary purelyperceptual task (counting the number of times an object was illuminated)referred to the same object which also served as the reaching target Thissuggests that preparing and executing a reaching movement cannot be donesimultaneously with attentional selection in the ventral stream when the twoselection processes refer to different objects When both tasks referred to thesame object parallel selection was possible

The fact that the coupling between perception and action in our experimentsoccurred inspite of the subjectrsquos complete knowledge of the location where thediscrimination target was presented argues for the assumption that thecoupling is obligatory Even with the incentive for separating visual perceptionand motor programming subjects do not succeed in decoupling both proc-esses On the other hand it is well known that visual attention can be shiftedwithout concomitant eye or hand movements (eg Posner 1980) Like Rizzo-latti et al (1987) we think that the strict coupling holds for the preparation andprogramming of the movement but does not necessarily require or entail itsovert initiation Therefore in cases where visual attention but not the handmoves we assume that the spatial parameters for the potential movement areavailable and provided by the attentional mechanism but that the movement isprevented from being converted into overt action due to the non-release of theldquogordquo signal

An interesting aspect of our data results from the median split analysis ofdiscrimination performance based on movement latencies (Figure 6) Theresults suggest that the coupling is restricted mainly to the fast responsesslower responses seem to allow better perception of the discrimination targetin the non-corresponding cases Again this is consistent with our theoreticalconsiderations In cases where the initiation of the reach is not done as fast aspossible (long latencies) it should be possible to undertake the discrimination

REACHING AND ATTENTION 101

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

task first then the reaching task Viewed from VAMrsquos perspective this meansthat the unitary visual attention mechanism first shifts to the known discrimi-nation target location allowing for discrimination and storage in short-termmemory Only then does attention shift to the movement target occur and theprogramming is continued

The instructions required the subjects to give priority to the reaching taskwhich had to be performed as fast and as precisely as possible visual discrimi-nation was the secondary task This is of some importance for the interpretationof the results since we wanted to avoid any crossover when measuringperceptual performance on the motor action Our results suggest that this aimwas indeed fulfilled Neither response latency and amplitude nor movementduration depended on the presentation of the discrimination target This con-trasts with the findings of Tipper et al (1992) and Pratt and Abrams (1994)who showed that distractors that appear on theway tothemovement target leadto delayed latencies of the reaching movement Two reasons may account forthis discrepancy First in the study of Tipper et al the distractors appearedsimultaneously with the movement target whereas in our study the discrimi-nation target was presented 150 msec after movement cue onset Thereforeone can assume that the programming of the movement might already havebeen completed before distractor onset Second the distractors used by Tipperet al were coloured objects appearing abruptly in the visual field such suddenonsets are generally assumed to attract attention automatically (Jonides 1981Yantis amp Jonides 1984) Similar reasoning holds for the results of Pratt andAbrams (1994) Inourparadigm ontheotherhand thetransients at themomentof DT presentation were equally distributed over all 10 items in the visual field(for each of the items two lines elements disappeared) In consequence it isunlikely that the presentation of the DT per se attracted attention Finally it isimportant to note that the targets did not ldquopop outrdquo from the distractors becauseof figural reasons whichwould again entail an automatic attraction of attentionto the discrimination target Similar approaches were used by Cheal and Lyon(1988) and Nakayama and Mackeben (1989)

The second experiment showed that coupling between dorsal and ventralprocessing is effective even during movement execution It appears that visualattention remained on the movement target even during execution of themovement We assume that this continuous coupling is necessary becausesubjects may evaluate movement success by means of the visual feedbackprovided by the LEDafter thereach Correspondingly theaverage movementswere amazingly precise and consistent as reflected in the high accuracy andlow variability of themovementdata However we donotclaimthatmovementexecution is necessarily accompanied with a binding of the attentionalmechanism at themovement target position Attention should only be allocatedto the future movement target when it is necessary to evaluate the success ofthe movement by comparing (proprioceptive or visual) information about the

102 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

actual movement end position with the intended target position Thiscomparison can probably not be done pre-attentively On the other hand whena movement is highly practisedmdashthis touches the issue of ldquoautomaticityrdquo (foroverviews see Neumann 1984 Shiffrin 1988)mdashand does not requirefeedbackcontrol thenattentiontotheresults of theactionmay notbenecessaryAn example of such an action might be shifting gears while driving a car

The results obtained here for reaching movements are to a significantdegree similar to our previous findings on the relation of saccades and objectrecognition (Deubel amp Schneider 1996 Schneider amp Deubel 1995) Theseexperiments revealed a similar amount of spatial restriction of perceptualcapabilities to the intended saccade target Also despite their knowledge of thelocation of the discrimination stimulus it was not possible for the subjects torecognize the object while preparing a saccade to a different target Finally asin the present experiments performance for non-target stimuli improved withlonger saccadic latencies (unpublished observations) These coincidencesprovide strong support for VAMrsquos assumption of a control mechanism thatis common for saccades and reaching and possibly for other types ofgoal-directed motor actions

Two further attentional theories explicitly include selection in the dorsalstream namely the ldquopremotor hypothesisrdquo of Rizzolatti et al (1987 1994) andthe ldquointegrated competition hypothesisrdquo of Duncan (1996) The central claimof the premotor theory is that the control of ldquospatial attentionrdquo originates in thedorsal spatial-motor areas In the original proposal only areas related to eyemovements were suggested to control spatial attention (Rizzolatti et al 1987)In contrast to VAM the premotor theory does not state whether separatemechanisms exist for dorsal and ventral visual processing nor how they arerelated Moreover in contrast to Posner and Petersen (1990) and VAMRizzolatti Gentilucci and Matelli (1985) claim that multiple attentional centresexist and that there is no need for a unitary mechanism for attentional control(see also Allport 1993) Our results argue for just the opposite namely for theexistence of a unitary visual attention mechanism that controls both ventral anddorsal processing

Duncan (1996) also proposed a framework for attentional processes in theprimate brain that incorporates dorsal spatial-motor processes According tohis ldquointegrated competition hypothesisrdquo ldquoattentionrdquo is considered to be anemerging state in which visual representations of one object win the competi-tion against representations of other objects Biasing the competition towardsone object is assumed to be controlled by the current task instruction and tooriginate in brain areas where the task-relevant attributes are computed There-fore analogous to VAM the integrated competition hypothesis predicts anobject-specific coupling between the ventral and dorsal stream (see also Dun-can 1984) Whenreaching orsaccading form theprimary task thetarget shouldwin the competition in both streams Other objects should be temporarily

REACHING AND ATTENTION 103

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

decoupled from action control and their perceptual representations properlyaccessed

We have previously noted the lack of behavioural investigations analysingtherelationship between selection-for-perception and selection-for-action Thesituation is similar with respect to neurophysiological studies on this issue Toour knowledge only one prominent single-cell study has directly addressed theeffects of (eye) movement programming on ventral processing ChelazziMiller Duncan and Desimone (1993) studied the activity of neurons in theinferior temporal cortex (IT) in tasks involving the preparation and executionof saccades in targetdistractor configurations These IT neurons are assumedto compute the identity of objects based on visual shape (see Oram amp Perrett1994) The results of Chelazzi et al (1993) demonstrated that the preparationof a goal-directed saccade to a target surrounded by distractors leads to adecrease in firing rate of the neurons that represent a distractor this decreaseoccurred shortly (90ndash120 msec) before saccade initiation Therefore selectionof an object as a movement target is coupled with ventral suppression ofdistractor information suggesting a neural mechanism for target selectionBased on our results we predict similar patterns of neural activity for othertypes of goal-directed movements such as reaching and grasping

In summary our study is the first to demonstrate directly an obligatoryspatially highly selective coupling of selection-for-object-recognition and se-lection-for-action in a task involving manual reaching In line with previoustheoretical considerations our findings argue for a unitary control mechanismof visual attention This mechanism selects objects for perceptual processing(object recognition) and at the same time provides the spatial parameters forgoal-directed actions such as reaching and grasping

REFERENCESAllport DA (1987) Selection for action Some behavioural andneurophysiological considera-

tions of attention and action In H Heuer amp AF Sanders (Eds) Perspectives on perceptionand action (pp 395ndash419) Hillsdale NJ Lawrence Erlbaum Associates Inc

Allport DA (1989) Visual attention In MI Posner (Ed) Foundations of cognitive science(pp 631ndash682) Cambridge MA MIT Press

Allport DA (1993) Attentionandcontrol Have webeen askingthewrong questionsAcriticalreview of twenty-five years In DE Meyer amp S Kornblum (Eds) Attention and performanceXIV Synergies in experimental psychology artificial intelligence an cognitive neuroscience(pp 183ndash218) Cambridge MA MIT Press

Bundesen C (1990) A theory of visual attention Psychological Review 97 523ndash547Castiello U (1996) Grasping a fruit selection for action Journal of Experimental Psychology

Human Perception and Performance 22 582ndash603Cheal M ampLyon DR (1988) Central andperipheral precuing of forced-choicediscrimination

Quarterly Journal of Experimental Psychology 43A 859ndash880Chelazzi L Miller EK Duncan J amp Desimone R (1993) A neural basis for visual search

in inferior temporal cortex Nature 363 345ndash347

104 DEUBEL ET AL

Desimone R amp Duncan J (1995) Neural mechanisms of selective visual attention AnnualReview of Neuroscience 18 193ndash222

Deubel H ampSchneider WX (1996) Saccade target selectionandobjectrecognition Evidencefor a common attentional mechanism Vision Research 36 1827ndash1837

DeYoe EA ampvanEssen DC (1988) Concurrentprocessing streams inmonkey visual cortexTrends in Neurosciences 11 219ndash226

Downing CJ amp Pinker S (1985) The spatial structure of visual attention In MI Posner ampOSM Martin (Eds) Attention and performance XI (pp 171ndash187) Hillsdale NJ LawrenceErlbaum Associates Inc

Duncan J (1984) Selective attention and the organization of visual information Journal ofExperimental Psychology General 113 501ndash517

Duncan J (1996) Coordinated brain systems in selective perception and action In T Inui ampJL McClelland (Eds) Attention and performance XVI (pp 549ndash578) Cambridge MA MITPress

Duncan J amp Humphreys GW (1989) Visual search and stimulus similarity PsychologicalReview 96 433ndash458

Eriksen BA amp Eriksen CW (1974) Effects of noise letters uponthe identification of a targetletter in a nonsearch task Perception and Psychophysics 16 143ndash149

Eriksen CW amp Hoffman JE (1973) The extent of processing of noise elements duringselective encoding from visual displays Perception and Psychophysics 1 155ndash160

Farah MJ (1990) Visual agnosia Disorders of object recognition and what they tell us aboutnormal vision Cambridge MA MIT Press

Goodale MA amp Milner AD (1992) Separate visual pathways for perception and actionTrends in Neurosciences 15 20ndash25

Graziano MSA amp Gross CG (1994) Mapping space with neurons Current Directions inPsychological Science 3 164ndash167

Hoffman JE amp Subramaniam B (1995) The role of visual attention in saccadic eye move-ments Perception and Psychophysics 57 787ndash795

Hughes HC amp Zimba LD (1987) Natural boundaries for thespatial spread of directed visualattention Neuropsychologia 25 5ndash18

Jeannerod M (1994) Therepresenting brain Neural correlates of motor intentionandimageryBehavioral and Brain Sciences 17 187ndash245

Jonides J (1981) Voluntary vs automatic control over the mindrsquos eyersquos movement In J Longamp A Baddeley (Eds) Attention and performance IX(pp 187ndash203) Hillsdale NJ LawrenceErlbaum Associates Inc

Klein R (1980) Does oculomotor readiness mediate cognitive control of visual attentionIn RNickerson (Ed) Attention and performance VIII (pp 259ndash276) Hillsdale NJ LawrenceErlbaum Associates Inc

Kolb B amp Whishaw IQ (1990) Fundamentals of human neuropsychology New York WHFreeman

Kowler E Anderson E Dosher B amp Blaser E (1995) The role of attention in the program-ming of saccades Vision Research 35 1897ndash1916

LaBerge D amp Brown V (1989) Theory of attentional operations in shape identificationPsychological Review 96 101ndash124

Livingstone MS amp Hubel D (1988) Segregation of form color movement and depthAnatomy physiology and perception Science 240 740ndash749

Milner AD ampGoodale MA (1995) The visual brain inaction New York OxfordUniversityPress

Mishkin M Ungerleider LG amp Macko KA (1983) Object vision and spatial vision Twocortical pathways Trends in Neurosciences 6 414ndash417

Neisser U (1967) Cognitive psychology New York Appleton-Century-Crofts

REACHING AND ATTENTION 105

Nakayama K amp Mackeben M (1989) Sustained and transient components of focal visualattention Vision Research 29 1631ndash1647

Neumann O (1984) Automatic processing A review of recent findings and a plea for an oldtheory In W Prinz amp AF Sanders (Eds) Cognition and motor processes (pp 227ndash267) (pp255ndash293) Heidelberg Springer-Verlag

Neumann O (1987) Beyond capacity A functional view of attention In H Heuer amp AFSanders (Eds) Perspectives on perception and action (pp 361ndash394) Hillsdale NJ LawrenceErlbaum Associates Inc

Neumann O (1990) Visual attention and action In O Neumann amp W Prinz (Eds) Relation-ships between perception and action Current approaches (pp 227ndash267) Berlin Springer-Verlag

Oram MW amp Perrett DI (1994) Modeling visual recognition from neurobiological con-straints Neural Networks 7 945ndash972

Posner MI (1980) Orienting of attention Quarterly Journal of Experimental Psychology 323ndash25

Posner MI amp Petersen SE (1990) The attention system of the human brain Annual Reviewof Neuroscience 13 25ndash42

Posner MI amp Raichle ME (1994) Images of Mind New York Scientific American LibraryPratt J amp Abram RA (1994) Action-centered inhibition Effects of distractors on movement

planning and execution Human Movement Science 13 245ndash254Rizzolatti G Gentilucci M amp Matelli M (1985) Selective spatial attention One center one

circuit or many circuits In MI Posner amp OSM Marin (Eds) Attention and performanceXI (pp 251ndash265) Hillsdale NJ Lawrence Erlbaum Associates Inc

Rizzolatti G Riggio L Dascola I amp Umiltagrave C (1987) Reorienting attention across thehorizontal and vertical meridians Evidence in favor of a premotor theory of attentionNeuoropsychologia 25 31ndash40

Rizzolatti G Riggio L amp Sheliga BM (1994) Space and selective attention In C Umiltagrave ampM Moscovitch (Eds) Attention and performance XV Conscious and nonconscious informa-tion processing (pp 231ndash265) Cambridge MA MIT Press

Schneider WX (1993) Space-based visual attention models and object selection Constraintsproblems and possible solutions Psychological Research 56 35ndash43

Schneider WX (1995) VAM Aneuro-cognitive model forvisual attention control of segmen-tation object recognition and space-based motor action Visual Cognition 2 331ndash375

Schneider WX amp Deubel H (1995) Visual attentionand saccadic eye movements Evidencefor obligatory and selective spatial coupling In JM Findlay R Walker amp RW Kentridge(Eds) Eye movement research (pp 317ndash324) Amsterdam Elsevier

Shepherd M Findlay JM amp Hockey RJ (1986) The relationship between eye movementsand spatial attention Quarterly Journal of Experimental Psychology 38A 475ndash491

Shiffrin RM (1988) Attention In RC Atkinson RJ Herrnstein G Lindsay amp RD Luce(Eds) Stevensrsquos handbookof experimental psychology (2nd edn Vol2 pp 739ndash811) NewYork Wiley

Stein JF (1992) The representation of egocentric space in the posterior parietal cortex Behav-ioral and Brain Sciences 15 691ndash700

Tipper SP Lortie C Baylis GC (1992) Selective reaching Evidence for action-centredattention Journal of Experimental Psychology Human Perception and Performance 18891ndash905

Treisman A (1988) Features and objects The fourteenth Bartlett memorial lecture QuarterlyJournal of Experimental Psychology 40 201ndash237

Treisman A amp Gelade G (1980) Afeature-integration theory of attention Cognitive Psychol-ogy 12 97ndash136

Treisman A amp Gormican S (1988) Feature analysis in early vision Evidence from searchasymmetries Psychological Review 95 15ndash48

106 DEUBEL ET AL

Van der Heijden AH (1992) Selective attention in vision London RoutledgeWolf W amp Deubel H (1997) P31 phosphor persistence at photopic luminance level Spatial

Vision 10 323ndash333Wolfe JM (1994) Guided search 20 A revised model of visual search Psychonomic Bulletin

and Review 1 202ndash238Yantis S amp Jonides J (1984) Abruptvisual onsets andselectiveattention Evidencefrom visual

search Journal of Experimental Psychology Human Perception and Psychophysics 10601ndash620

Zeki SM (1993) Avision of the brain Oxford Blackwell Scientific

REACHING AND ATTENTION 107