The Effect of Pictorial Illusion on Prehension and Perceptiontres/neuroII/yanagi.pdf · The Effect of Pictorial Illusion on Prehension and Perception ... haus Illusion. Aglioti,

The Effect of Pictorial Illusion onPrehension and Perception

Angela M Haffenden and Melvyn A GoodaleUniversity of Western Ontario

Abstract

The present study examined the effect of a size-contrastillusion (Ebbinghaus or Titchener Circles Illusion) on visualperception and the visual control of grasping movements Sev-enteen right-handed participants picked up and on other trialsestimated the size of ldquopoker-chiprdquo disks which functioned asthe target circles in a three-dimensional version of the illusionIn the estimation condition subjects indicated how big theythought the target was by separating their thumb andforenger to match the targetrsquos size After initial viewing no

visual feedback from the hand or the target was availableScaling of grip aperture was found to be strongly correlatedwith the physical size of the disks while manual estimationsof disk size were biased in the direction of the illusion Evi-dently grip aperture is calibrated to the true size of an objecteven when perception of object size is distorted by a pictorialillusion a result that is consistent with recent suggestions thatvisually guided prehension and visual perception are mediatedby separate visual pathways

INTRODUCTION

Vision provides us with a vast array of information aboutthe world around usmdashinformation that can direct ourthoughts and guide our actions Yet our perception ofthe world can sometimes be misleading Objects canappear to be larger or smaller and sometimes closer orfurther away than they really are Nevertheless we rarelynotice these discrepancies between our perception andthe real world until they are pointed out to us Who hasnot been taken in by visual illusions at one time oranother And even when we know full well that objectsand their spatial relations cannot be as they appearmdashwhen we watch events unfold on the television or moviescreen for examplemdashthe changes in the real size anddistance of the distal stimulus have little effect on whatwe see (Gregory 1995 Hochberg 1987) In short ourperception of the world is remarkably insensitive toarbitrary changes in perspective and scale that com-monly occur in movies and television It is the relativesize and distance of objects in the array not their abso-lute size or location that appear to be critical for per-ception

Quite the opposite is true for the visual control ofskilled actions To pick up an object such as coffee cupit is not enough to know its relative size and location inthe visual array for the grasp to be successful our visuo-motor system must calculate the cuprsquos absolute size anddistance (for a discussion of these issues see BridgemanKirch amp Sperling 1981 Goodale amp Haffenden in pressMilner amp Goodale 1995) In short visuomotor computa-tions must be metrically accurate Moreover those com-

putations must be carried out in the appropriate ego-centric frames of reference for the intended action Anaccurate grasping movement for example requirescomputation of the parameters of the goal object inldquoarm-centeredrdquo coordinates (Graziano amp Gross 1994Soechting amp Flanders 1992)

Dissociations Between Perception and Action inNeurological Patients

These differences in the requirements of visual percep-tion and the visual control of action suggest that a singlegeneral-purpose representation of the world could notserve both functions Instead the different transforma-tions required for perception and action would appearto require separate visual mechanisms each adapted tothe requirements of the output system it serves This ideais supported by a number of studies in neurologicalpatients who show dissociations between visual percep-tion and the visual guidance of skilled actions followingdamage to different neural pathways Patients with dam-age in the superior regions of the posterior parietalcortex for example are often unable to use informationabout an objectrsquos size shape orientation or location tocontrol their visually guided reaching movements (Good-ale et al 1994 Jakobson Archibald Carey amp Goodale1991 Jeannerod 1988 Perenin amp Vighetto 1988) Yettheir visual perception of objects remains relatively in-tact and in many cases they are able to identify ordiscriminate between the very objects they cannot grasp(for discussion of this issue see Milner amp Goodale 1995)

copy 1998 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 101 pp 122ndash136

Damage to occipito-temporal cortex may lead to thereverse disorder in which the patient has impaired visualperception with spared visually guided movements Thisis the case with the patient DF whose grasping move-ments are well formed and accurate even though shecannot visually discriminate between or identify thesame objects to which she can direct well-formed grasps(Goodale et al 1994 Goodale Milner Jakobson amp Carey1991 Milner amp Goodale 1995) On the basis of theseneuropsychological studies and additional electrophysi-ological and behavioral studies in the monkey Goodaleand Milner (1992) have proposed that the distinctionbetween vision for perception and vision for action canbe mapped onto the two prominent pathways orldquostreamsrdquo of visual projections that have been identiedin the primate cerebral cortex a ventral stream whicharises from the primary visual cortex and projects to theinferotemporal cortex and a dorsal stream which alsoarises from primary visual cortex but projects instead tothe posterior parietal cortex (Ungerleider amp Mishkin1982)

Ungerleider and Mishkin (1982) originally proposedthat the ventral stream plays a special role in the iden-tication of objects whereas the dorsal stream is respon-sible for localizing objects in visual space Goodale andMilnerrsquos reinterpretation of this story places less empha-sis on the differences in the visual information that isreceived by the two streams (object features versusspatial location) than it does on the differences in thetransformations that the streams perform upon that in-formation (Goodale 1993 Goodale amp Milner 1992 Mil-ner amp Goodale 1993 1995) According to their accountboth streams process information about object featuresand about their spatial locations but each stream usesthis visual information in different ways In the ventralstream the transformations deliver the enduring charac-teristics of objects and their relations permitting theformation of long-term perceptual representations Suchrepresentations play an essential role in the identica-tion of objects and enable us to classify objects andevents attach meaning and signicance to them andestablish their causal relations Such operations are es-sential for accumulating a knowledge-base about theworld In contrast the transformations carried out by thedorsal stream provide the current location and disposi-tion of a goal object in egocentric coordinates andthereby mediate the visual control of skilled actionssuch as manual prehension directed at that object Ofcourse the two systems work together in controlling therich stream of behavior that we produce as we live ourcomplex lives Their respective roles in this control differhowever The perceptual representations constructed bythe ventral stream are part of a high-level cognitivenetwork that enables an organism to select a particularcourse of action with respect to objects in the worldthe visuomotor networks in the dorsal stream (and asso-ciated cortical and subcortical pathways) are responsible

for the programming and on-line control of the particu-lar movements that the selected action entails

Dissociations Between Perception and Action inNormal Observers

If visual perception and the visual control of actiondepend on different neural mechanisms in the humancerebral cortex it should be possible to demonstrate adissociation between these two kinds of visual process-ing in neurologically intact individuals In other wordseven in normal observers the visual information under-lying the calibration and control of a skilled motor actdirected at an object might not always match the per-ceptual judgments made about that object The trick isto nd the right task to demonstrate this dissociationThe present study used one particular paradigm a size-contrast illusion to do this

Most of the studies examining such dissociations innormal observers have focused on the spatial frames ofreference used by visual perception and the visual con-trol of action and have not looked at the effects offrames of reference on object size (for review see Goo-dale amp Haffenden in press Milner amp Goodale 1995) Thebulk of this work on spatial location suggests that themechanisms mediating the perception of object locationoperate largely in allocentric coordinates whereas thosemediating the control of object-directed actions (egsaccadic eye movements and aiming movements withthe limb) operate in egocentric coordinates In otherwords perception uses a coordinate system that isworld-based in which objects are seen as changing loca-tion relative to a stable or constant world the systemscontrolling action systems however cannot afford thesekinds of constancies and must compute the location ofthe object with respect to the effector that is directedat that target In short visuomotor control demandsdifferent kinds of visual computations than visual per-ception As we suggested earlier it is this difference inthe computational requirements of the visual control ofaction and visual perception that explains the divisionof labor between the two streams of visual processingin the primate cerebral cortex

Just as the perception of object location appears tooperate within relative or allocentric frames of refer-ence so does the perception of object size Although weoften make subtle judgments about the relative sizes ofobjects we rarely make judgments of their absolute sizeIndeed our judgments of size appear to be so inherentlyrelative that we can sometimes be fooled by visual dis-plays in which visual stimuli of the same size are posi-tioned next to comparison stimuli that are either muchsmaller or much larger than the target stimuli Suchsize-contrast illusions are a popular demonstration inmany introductory textbooks in psychology and percep-tion One such illusion is the so-called Ebbinghaus Illu-sion (or Titchener Circles Illusion) in which two target

Haffenden and Goodale 123

circles of equal size each surrounded by a circular arrayof either smaller or larger circles are presented side byside (see Figure 1a) Subjects typically report that thetarget circle surrounded by the array of smaller circlesappears larger than the one surrounded by the array oflarger circles presumably because of the difference inthe contrast in size between the target circles and thesurrounding circles Although the illusion is usually de-picted in the way just described it is also possible tomake the two target circles appear identical in size byincreasing the actual size of the target circle surroundedby the array of larger circles (see Figure 1b)

Although perception is clearly affected by these ma-nipulations of the stimulus array there is good reason tobelieve that the calibration of size-dependent motor out-puts such as grip aperture during grasping would notbe After all when we reach out to pick up an objectparticularly one we have not seen before our visuomo-tor system must compute the objectrsquos size accurately if

we are to pick it up efciently It is not enough to knowthat the target object is larger or smaller than surround-ing objects the visuomotor systems controlling handaperture must compute its real size

Moreover the visual control of motor output cannotafford to be fooled by an accidental conjunction ofcontours in the visual array that might lead to illusionsof size or location There are situations in which the lifeor death of the organism will depend on the accuracyof a motor output sensitivity to a visual illusion in thescene could be an enormous liability For these reasonstherefore one might expect grip scaling to be insensitiveto the kind of size-contrast illusion seen in the Ebbing-haus Illusion

Aglioti DeSouza and Goodale (1995) tested this ideaby using a three-dimensional version of the EbbinghausIllusion Thin plastic disks resembling poker chips wereused as the center circles these disks were placed withina conventional Ebbinghaus array drawn on a horizontaldisplay card Two kinds of trials were used On one thetwo disks appeared to be different in size even thoughthey were actually physically identical in the other trialthe two disks appeared to be equal in size even thoughthe disk in the annulus of large circles was actuallyslightly larger than the disk in the small circle annulus(see Figure 1) These two kinds of trials were randomlypresented to the subjects and the left-right positions ofthe large and small annuli were also randomly inter-leaved Subjects were asked to pick up the target disk onthe left or right side of the display based on whetherthey thought the two disks looked the same or differentin size The maximum aperture of the subjectsrsquo graspswere measured in ight using optoelectronic recordingEven though the subjectsrsquo choices clearly indicated thatthey were subject to the illusion throughout testing theynevertheless scaled their grip appropriately to the physi-cal size of the disks In short the calibration of grip sizewas largely impervious to the effects of the size-contrastillusion

The Present Study

Although the Aglioti et al (1995) results appear tosuggest that the perception of object size depends ondifferent computations from those mediating the calibra-tion of the grasp there were some problems with thestudy that undercut this conclusion First there is theissue of visual feedback during the execution of thegrasping movement It is possible that subjects may havebeen adjusting their grip aperture in ight by comparingthe size of the disk and the opening between their ngerand thumb as they got closer and closer to the diskAglioti and his colleagues argued against this possibilityand cited experiments showing that at least 500 msecwere needed for processing visual information aboutchanges in size and shape of the goal object before anew motor command could be produced Maximum grip

Figure 1 The Ebbinghaus Illusion (a) The standard version of the il-lusion The target circles in the center of the two annuli appear tobe different in size even though they are physically identical Peopletypically report that the circle surrounded by the annulus of smallercircles appears to be larger than the circle surrounded by the annu-lus of larger circles (b) A version of the illusion in which the targetcircle in an array of larger circles is physically larger than the othertarget circle The two target circles should now appear to be identi-cal in size

124 Journal of Cognitive Neuroscience Volume 10 Number 1

aperture which was the kinematic marker used byAglioti et al is typically reached around 420 msec aftera grasping movement has begun or about 70 of theway through the reach (Jeannerod 1984) This suggeststhat maximum grip aperture reects early programmingand not on-line adjustments to the grasp More recentwork challenges this conclusion however In an experi-ment examining rapid adjustments of the grasp to sud-den and unexpected changes in object size Castiello andJeannerod (1991) found that a much shorter time wasrequired for updating the motor output In their experi-ment subjects required an average of 320 msec to adjustthe posture of their hand in order to accurately grasp arod that changed in size when the reach was initiatedThe fact that hand posture can be adjusted in such ashort time at least under certain conditions makes itnecessary to eliminate the possibility that subjects in anEbbinghaus-type task were adjusting their grasp on thebasis of the visual feedback Clearly it would be moreconvincing to run the Aglioti et al study in visual openloop in other words under conditions in which thesubjects have absolutely no opportunity to view theirhand or the target during the execution of the graspingmovement Therefore in the present experiment wetested all our subjects under open-loop conditions toeliminate the possibility of on-line visual control of gripaperture

There is no doubt that subjects will continue to scalefor object size under open-loop conditions Previousresearch by Jakobson and Goodale (1991) for examplehas demonstrated that although there is an overall in-crease in maximum grip aperture during reaching move-ments when visual feedback is unavailable maximumgrip aperture is still scaled for object size in the absenceof visual feedback and the variance in grip apertureunder these conditions does not differ from the variancewhen visual feedback is available From this we ex-pected that when we presented subjects with the three-dimensional version of the Ebbinghaus Illusion just asAglioti et al (1995) had done but under visual open loopthe maximum grip aperture achieved by subjects wouldstill be scaled to the physical size of the disks

A second issue that the present study addressed wasthe method of measuring the strength of the illusion Inthe Aglioti et al (1995) study subjectsrsquo perceptions ofdisk size were measured by a samedifferent choice Thismeant that a dichotomous measure of the perceptualeffect of the illusion was being compared with a con-tinuous measure of grip calibration Clearly it would bebetter if a continuous measure were used for both out-puts An ideal measure of perceived size of course wouldbe one that was directly comparable to the grip calibra-tion measure For this reason we asked subjects to esti-mate the size of a disk by matching the distancebetween their thumb and index nger to the width ofthe disk It is important to emphasize that although handand nger movements were required in this manual

estimation task the programming and execution of thosemovements does not involve the same control systemsused in grasping In the manual estimation task subjectsare simply asked to provide a kind of manual ldquoread-outrdquoof what they perceive In the grasping task they areengaging the visuomotor networks that mediate theskilled movements involved in human prehension Inshort the manual estimation task provided a measure ofvisual perception that could be more directly comparedto the scaling of prehension The manual estimation taskwas also carried out in visual open loop after subjectsviewed the object estimations were made in the darkSubjects were also allowed to reach out and pick up thetarget disk after the estimation ensuring that they re-ceived the same amount of haptic feedback about thephysical size of the disks as they did in the grasping task

One nal issue that was addressed in the presentexperiment was the question of the general effect onprehension of having an annulus surrounding a goalobject The inuence of this factor on the grasp needsto be separated from any inuence the illusion back-ground might have on grip scaling To investigate thisquestion two additional background conditions wereused in the present experiment The rst consisted oftwo annuli made up of equal-sized circles midway in sizebetween the circles in the large annulus and the smallannulus from the illusion background (Figure 2) Thesecond control background involved no surrounding an-nuli the target disks were presented on their own butin the same positions as when they were surrounded byannuli By including these two conditions it was possibleto see if the scaling of the grasp was affected by reachingfor a target surrounded by other forms (and whether ornot this effect was independent of any effect the simplepresence of annuli might have on perception) For ex-ample reaching for an object surrounded by an annulus

Figure 2 The equal annuli background The circles composing theannuli are midway in size between the circles in the large annulusand the circles in the small annulus from the illusion background


may be like reaching into a hole the opening of the handmust be reduced to avoid hitting the sides of the holebut not so much as to miss the edges of the target

To recap the purpose of this study was threefold rstto replicate the ndings of Aglioti et al (1995) but withan open loop design to eliminate the use of visual feed-back second to use a continuous measure of perceiveddisk size that can be more directly compared to gripcalibration and third to examine the effect of the sur-rounding annulus on grasping and manual estimates ofdisk size independent of the illusion It was expectedthat grasp would not succumb to the illusion but thatmanual size estimations would correspond to the sub-jectsrsquo changing perceptions of the same disks Such anoutcome would be consistent with the idea that thevisual mechanisms that mediate prehension are quiteseparate from those that mediate perception

RESULTS

Psychophysical testing carried out during practice trialsallowed us to determine the difference in size betweenthe two target disks that produced judgments of percep-tual equivalence for each subject (Figure 1b) This differ-ence turned out to be 244 mm averaged across the 18subjects we tested in other words for a pair of disks tobe judged as equivalent on the illusion background thedisk centered in the annulus of large circles had to be244 mm larger than the disk centered in the annulus ofsmall circles For 11 subjects the required difference was2 mm for 6 subjects it was 3 mm and for 1 subject itwas 4 mm The most common situation was a 28-mmdisk placed in the small circle annulus and a 30-mm diskplaced in the large circle annulus to achieve equivalencein perceived size Because different combinations wererequired to create perceptual equivalence across thesubjects the disks are referred to in relative terms (ieas the large disk and the small disk) When subjects werepresented with physically identical disk pairs they werepresented with two small disks on half of the trials andtwo large disks on the other half of the trials

Subjects were tested on both a grasping task (Figure3a) and a manual estimation task (Figure 3b) the orderof testing was counterbalanced across subjects Post hoccomparisons were performed as paired t tests usingBonferroni corrections to adjust the error rate for thenumber of tests performed On one half of the 70 trialsof each task the two disks in the array were physicallyidentical in size on the other half of the trials whichwere randomly interleaved the two disks were physi-cally different in size The disks were surrounded eitherby the illusion display or the equal annuli display or theywere presented on a blank background (see ldquoMethodsrdquofor details of counterbalancing) Subjects were in-structed to look carefully at the disks when they werepresented and to decide whether they were identical ordifferent in size If they appeared different they were to

pick up (or manually estimate) the disk on the left usingthe index nger and thumb of their right hand if thetwo disks appeared identical they were to pick up (ormanually estimate) the disk on the right (instructionswere reversed for half the subjects) As soon as theirhand left the table the overhead light went out elimi-nating the subjectrsquos view of their hand and the targetdisplay The amplitude of the opening between theirindex nger and thumb was tracked using optoelec-tronic recording of small infrared light-emitting diodesattached to the tips of both digits (see Figure 3)

All the subjects remained sensitive to the size-contrastillusion throughout testing When an illusion backgroundwas present and the two disks were identical in size thechoice made by the subjects indicated that they thoughtthe disks were different on illusion trials in which thedisks were physically different they behaved as thoughthe two disks were the same size Quite the oppositepattern of results was observed when the surroundingannuli were the same size or the background was blankOn these trials the subjectsrsquo choices reected the realdifference in size between the two disks

A clear dissociation was observed on illusion trialsbetween the calibration of the grip aperture during thegrasping task and the separation between the indexnger and thumb during the manual estimation task AsFigure 4a illustrates on trials in which subjects reportedthat the disks were identical in size even though theywere physically different the calibration of grip aperturewas scaled to the actual not the apparent size of thedisks In fact the difference of 211 mm between themean maximum grip apertures for the large and smalldisks was only slightly less than the average differencein disk size of 244 mm needed to produce perceptualequivalence on the illusion background In contrast asFigure 4b illustrates when subjects manually estimatedthe size of the disks on trials in which they believed thedisks were identical in size (even though they weredifferent) their estimates corresponded to the apparentnot the real size of the disks In other words theirmanual estimates of the size of the small and large diskwere virtually identical

When the illusion was presented in the traditionalfashion with two physically identical disks that appeardifferent in size (Figure 1a) the illusory perceptionsagain failed to fool the calibration of the grasp but werereected in the manual estimations of disk size Thus asFigure 4c illustrates grip scaling for the physically iden-tical pairs of small or large disks was unaffected by thesize of the circles in the surrounding annuli In otherwords subjects again scaled their grip to the actual sizeof the target disks rather than to their apparent size Thereverse effect was seen in the manual estimation taskHere the estimates of target disk size reected the ex-pected effect of the illusion the disks surrounded by thesmall circle annulus were estimated as being larger thanthe disks surrounded by the large circle annulus (Figure


4d) This was true for both the small disk pair and thelarge disk pair

In summary when disks were presented on the illu-sion background maximum grip aperture correspondedto the physical size of the disks whether the disksappeared to be the same or different in size Manualestimations of disk size showed exactly the oppositepattern they were clearly inuenced by the illusionImportantly this dissociation was seen despite the factthat the same effector the hand was used to producethe response for both tasks in one case to reach out andgrasp the disk and in the other case to ldquomatchrdquo the size

of the disk to the distance between the thumb and indexnger

Of course when no illusion background was presentas in the case where the disks were presented either ona blank background or on an equal annuli backgroundwe expected the grasping and manual estimation tasksto yield similar results On these backgrounds the per-ceived size of the disks and their actual size shouldcoincide The results for physically different pairs ofdisks are presented in Figure 5 and for physically identi-cal pairs of disks in Figure 6

When a physically different disk pair (one large one

Figure 3 Photographs of thetwo tasks performed by eachsubject using a three-dimen-sional version of the Ebbing-haus Illusion Note theinfrared light-emitting diodes(IREDs) attached to the ngerthumb and wrist that allowedfor the optoelectronic track-ing (a) The grasping task Thesubjectrsquos hand is pictured in-ight on the way to the targetdisk (b) The manual estima-tion task The heel of the sub-jectrsquos hand is resting on thetable and the thumb and in-dex nger are opened amatching amount to the per-ceived size of target disk Inthe actual tasks used in thepresent experiment of coursethe overhead light wouldhave been turned off duringthe performance of the move-ments


small) was placed on a background consisting of twoannuli of equal mid-sized circles or no surrounding an-nuli both grip scaling (Figure 5a) and manual estimations(Figure 5b) reliably reected the difference in size be-tween the disks A similar pattern of results was observedwhen the disks were placed on a plain white back-ground Physically different disks generally produced sig-nicantly greater grip apertures (Figure 5c) andsignicantly greater manual estimates (Figure 5d) for thelarge disk than for the small disk

When identical disks were used (both large or bothsmall) no signicant differences were seen betweenidentical disks in either the grasping task or the manualestimation task for either the equal annuli backgroundor the blank background (Figure 6) Comparisons of

average response to both large disks and both small disksrevealed appropriate scaling in both the grasping andthe manual estimation task Moreover grip scaling andmanual estimations showed comparable levels of accu-racy

In short when no illusion was present manual estima-tions and grip scaling both followed the true size of thedisks Nevertheless the presence of the equal annuli didhave an effect on the magnitude of the responses andthis effect differed between the visuomotor and theperceptual tasks These differences are illustrated in Fig-ure 7 In the grasping task disks on the backgroundconsisting of equal-sized annuli produced smaller maxi-mum grip apertures than the same disks presented ontheir own with no surrounding annuli Conversely in the

Figure 4 Graphs illustratinggrip aperture and manual esti-mations for trials where targetdisks were placed on the illu-sion background (a) Meanmaximum grip aperture on tri-als in which subjects reportedthat the disks were identicalin size even though theywere physically different Thedifference between the maxi-mum grip aperture achievedfor large disks was sig-nicantly greater than themaximum grip apertureachieved for the small diskst(17) = 392 p lt 005 (b)Mean manual estimations ofdisk size on perceptually iden-tical physically different trialsThe manual estimations oflarge disk size were not sig-nicantly greater than thoseof small disk size t(16) = 014p gt 005 (c) Mean maximumgrip aperture on trials inwhich two physically identicaldisks appeared different insize There was no signicantdifference between the maxi-mum grip apertures achievedfor small disks in the small an-nuli and the small disks in thelarge annuli t(17) = 62 p gt005 Similarly there was not asignicant difference in maxi-mum grip apertures for thelarge disk pair t(17) = 236p gt 005 (d) Mean manual es-timations on trials in whichtwo physically identical disksappeared different in size Thedisks surrounded by the smallcircle annulus were estimatedas being signicantly largerthan the disks surrounded bythe large circle annulus for both the small disk pair t(16) = 326 p lt 005 and the large disk pair t(16) = 437 p lt 001Error bars depict standard errors of the means


manual estimation task the disks surrounded by themedium circle annuli were estimated as being largerthan the disks that were presented without surroundingannuli

DISCUSSION

The results of the present experiment provide strongsupport for the idea that the visual mechanisms under-lying perception are distinct from those underlying thecontrol of skilled actions Despite the presence of thesize-contrast illusion created by the Ebbinghaus displaythe scaling of the grasp corresponded to the actual notthe apparent size of the target disks Although theseresults parallel those of Aglioti et al (1995) the presentstudy was run in visual open loop In the present experi-

ment there was no opportunity for subjects to use visualfeedback from their hand or the target to adjust theirgrip aperture in ight The fact that maximum grip aper-ture under these conditions continued to correspond tothe physical rather than the perceived size of an objectsuggests that this real-world metrical calibration couldnot have been based on adjustments made during theexecution of the grasp Instead subjects must have pro-grammed the aperture of their grasp on the basis of theirinitial glimpse of the target disk presumably in relationto its real size Moreover the accurate scaling of thegrasp occurred while subjects were in the act of indicat-ing their illusory perceptions by reaching out and pick-ing up the appropriate target disk This strikingdissociation in normal human observers is consistentwith the proposal put forward by Goodale and Milner

Figure 5 Results for the con-ditions in which a physicallydifferent disk pair (one largeone small) was placed on aeither a background consist-ing of two annuli of equalmid-sized circles (a and b) ora blank background with nosurrounding annuli (c and d)(a) Mean maximum grip aper-tures achieved for large diskson the equal annuli back-ground were larger thanthose achieved for small disksThis difference was not sig-nicant t(17) = 306 p gt005 (b) Manual estimationsof large disks on the equal an-nuli background were sig-nicantly larger than those forsmall disks t(16) 442 p lt001 (c) Maximum grip aper-tures achieved for large diskswere signicantly greater thanthose for small disks t(17)425 p lt 001 (d) Manual esti-mations were signicantlylarger for the large disk thanfor the small disk t(16) 557p lt 001 Error bars depictstandard errors of the means


(1992) on the basis of neuropsychological evidence thatthere are separate neural substrates for visual perceptionand the visual control of action

The relative insensitivity of reaching and grasping topictorial illusions has been demonstrated in two otherrecent experiments employing classical pictorial il-lusions Vishton and Cutting (1995) have shown thatgrasping movements are relatively insensitive to the hori-zontal-vertical illusion and Ian Whishaw (personal com-munication 1996) has obtained similar ndings using thePonzo illusion In neither of these studies however wasan explicit comparison made between the calibration ofgrasp and perceptual judgments using the same read-outmdashthe separation between the nger and thumb Inour experiment however we compared grip scalingwith manual estimations of the size of the same target

stimuli When subjects estimated the size of a disk byopening their nger and thumb a matching amount theirestimations corresponded to the apparent rather thanthe real size of the disk In other words their manualestimation of disk size unlike their visuomotor scalingwas completely driven by the illusory condition inwhich the disk was placedmdasheven though subjects hadthe same amount of haptic feedback in both conditions(They always picked up the disk after estimating its size)These results are particularly compelling because as wehave already emphasized in both the manual estimationtask and the visuomotor task movements of the handand ngers were required Yet in one case the perceivedsize of the object dominated the response whereas inthe other the real size of the object dominated Some of the other ndings in the present study also

Figure 6 Results for physi-cally identical disk pairs (bothlarge or both small) placed oneither the equal annuli back-ground (a and b) or the blankbackground (c and d) All com-parisons between meansachieved for physically identi-cal disks were not signicantin either the grasping task orthe manual estimation taskp gt 005 (a) Small t(17) =082 large t(17) = 198(b) Small t(16) = 084 larget(16) = 182 (c) Small t(17) =058 large t(17) = 077(d) Small t(16) = 103 larget(16) = 042 Error bars depictstandard errors of the means


point to separate mechanisms for perception and visuo-motor control For one thing not only did the illusionhave different effects on manual estimation and maxi-mum grip aperture but the amplitude of the openingbetween the nger and thumb was quite different inthese two response conditions (even though the sameeffector was being used) (It is worth noting again thatthe opening between the thumb and forenger wasmeasured using the same method for both manual esti-mates and grip scaling) Manual estimations producednger-thumb apertures of about 40 mm a distance thatwas only slightly larger than the physical size of thedisks In short subjects were showing us what they sawGrasping however produced maximum apertures thatwere around 60 mm wide a value that was often twiceas large as the actual size of the disks Such large gripapertures are quite typical Many investigators havedemonstrated that even though grip aperture is highlycorrelated with object size subjects always achievemaximum apertures that are quite a bit larger than thegoal object (eg Jakobson amp Goodale 1991 Jeannerod1984) In other words subjects are not trying to matchtheir grip in ight to the objectrsquos actual size Instead theyopen their hand in a remarkably stereotyped fashionreaching maximum aperture approximately 70 of theway through the reach and then clamping down on theobject as the hand gets closer (Jeannerod 1984) In ourexperiment subjects behaved in exactly the same wayIt is difcult to argue therefore that subjects were some-how consciously and deliberately scaling their grip ap-

erture to match the size of the target disk If this werethe case why would they use an estimate that was oftentwice as large as the object they were trying to pick upInstead the large mismatch between the amplitude ofthe grip aperture and the size of the target disk probablyreects the normal processes underlying visuomotorcontrol of object-directed actionsmdashjust as the relativelycloser match between manual estimates and the size ofthe disk reects the operation of the normal processesunderlying the visual perception of those same objects

Another aspect of the results that lends support to theidea of a dissociation between perception and visuomo-tor control is the observation that subjects remainedsensitive to the illusion over the 2-hr testing periodmdashde-spite the fact that they were receiving haptic feedbackabout the real size of the disks throughout testing Manysubjects commented that they were surprised by thesize of the disk when they picked it up it was smalleror larger than they had expected Yet this feedback fromprehension of the disks did not diminish the effect ofthe illusion In addition the continued exposure to thesame pairs of disks on the illusion background did notcause the illusory effect to diminish

Although it is clear that the calibration of graspingmust use the real size of the target object it is notimmediately apparent why perceptual judgments of ob-ject size are taken in by the illusion Again the explana-tion turns on the fact that perception typically involvesrelative not absolute judgments of object size Indeedthe Ebbinghaus Illusion like a number of size-contrast

Figure 7 Results showingthe effect of a surrounding an-nulus on grasp and on manualestimations (a) In the grasp-ing task small disks on thebackground consisting ofequal-sized annuli producedsignicantly smaller maximumgrip apertures than smalldisks presented on the blankbackground t(35) = 253 p lt005 (b) In the manual estima-tion task large disks sur-rounded by the equal annuliproduced signicantly largerestimations than the largedisks that were presented onthe blank background t(34) =305 p lt 001 Error bars de-pict standard errors of themeans


illusions appears to depend on rather high-level percep-tual processes in which relative size judgments of similarobjects are being made (Coren amp Enns 1993 Coren ampMiller 1974 Weintraub amp Schneck 1986) For examplethe strength of the illusion depends more on whetheror not the surrounding context stimuli are the same classof object as the test stimulus rather than on whether ornot they have the same visual geometry (Coren amp Enns1993) This implies that some sort of comparison is beingmade between the surrounding annuli and the test stim-uli Indeed it has been suggested that the illusion de-pends on a size-constancy computation in which thearray of smaller circles is assumed to be more distantthan the array of larger circles as a consequence thetarget circle within the array of smaller circles is per-ceived as more distant (and therefore larger) than thetarget circle of equivalent retinal image size within thearray of larger circles (Coren 1971 Gregory 1963)

Mechanisms such as these in which the relationsbetween objects in the visual array play a crucial role inscene interpretation are clearly central to perceptionMoreover such comparisons are quite obligatory Wecannot but fail to see some things as smaller or larger(or closer or further away) than others But if such anobligatory analysis is being carried out on the visualarray inappropriate comparisons might sometimes beexpected to take place In other words illusions couldarise because of accidental conjunctions and alignmentsof contours objects or surfaces in the visual array Psy-chologists have capitalized on this fact by creating pic-torial illusions By studying the errors in judgment thatsubjects make when they look at such illusions psy-chologists have learned a good deal about normal per-ception Such illusions are probably quite rare in ourdaily lives But even when they do arise they will havelittle consequence for our behavior since we are rarelyasked to make explicit judgments about object size anddistance

Although small errors in size and distance estimationmight have little consequence for our perception of theworld such miscalculations in the visuomotor domaincould be devastating To be successful a goal-directed actlike prehension must use computations that deal withthe metrical properties of the object independent of thecontext of the visual array in which that object is em-bedded The real distance of the object must be com-puted presumably on the basis of reliable cues such asstereopsis and retinal motion rather than on the basis ofpictorial cues which as we have seen can be quiteunreliable Once distance has been computed this esti-mate combined with the retinal image size of the objectwill deliver the true size of the object for calibrating thegrasp By relying on these sorts of cues the computationsmediating grasping movements are quite insensitive tothe cues that drive the size-contrast illusions in whichmore relative scene-based calculations are at work

Even though relative judgments of size and distance

appear to be central to our perception of the world whyis it that we do not routinely incorporate the accuratemetrical information that is clearly available to the visuo-motor system There could be a number of reasons forthis First many perceptual judgments involve distal stim-uli in the visual array that are beyond the range ofmechanisms like stereopsis that are designed to deliveraccurate metrical computations Second as we have al-ready suggested there is little consequence anyway forany errors in perceptual judgments that we might makeFinally because perceptual representations unlike a goal-directed movement involve an analysis of the entirevisual array a metrically accurate representation wouldbe computationally expensive

The proposed dissociation between the mechanismsunderlying visually guided movements and those under-lying perception does not preclude the possibility ofperception inuencing visually guided movements In-deed under normal circumstances perception acts inconcert with prehension and our hand posture variesdepending on the perceived identity of the target objectFor example our hand posture when we reach for a forkthat we are going to put away is very different from theposture we adopt when we are going to use the samefork to eat with (for a discussion of this issue see Milneramp Goodale 1995) In the present study scaling of thegrasp was largely dependent on the physical size of thedisks yet a small inuence of perception on grasp couldbe seen in reaches to the physically identical disk pairson the illusion background In other words subjectssometimes opened their hand slightly wider when reach-ing for a disk surrounded by the small circle annulusthan when reaching for a physically identical disk sur-rounded by the large circle annulus Although this effectwas not signicant a similar perceptual effect on gripcalibration was seen in the Aglioti et al (1995) study andin their case was signicant This suggests that the per-ception of size can exert some inuence on the pro-gramming of hand posture for prehension Indeedalthough the visual mechanisms underlying perceptionare largely separate from the visual mechanisms under-lying prehension this separation does not mandate thatthere be no communication between the mechanismsIn fact as was just discussed such communication wouldappear to be necessary for generating hand postures thatare appropriate to the function of the goal object

An inuence of perception on action has also beendemonstrated in certain neurological cases JeannerodDecety and Michel (1994) for example have describeda patient with damage to the posterior parietal cortexwho was unable to scale her grasp for ldquoneutralrdquo objectseven though she could scale her grasp to familiar objectsof the same size In another patient an interesting dis-connection between visual recognition and functionalcontrol of the grasp has been demonstrated (Sirigu et al1995) This patient who shows good metrical scaling ofher grasps to objects cannot incorporate functional ele-


ments into the grasp even though the patient recognizesthe identity of the object Taken together these resultssuggest that perception makes an important but perhapssomewhat independent contribution to the organizationof manual prehension movements The neural pathwayssupporting this perceptual or cognitive mediation ofgrasping remain unknown

Nevertheless it is important to emphasize that theprincipal determinant of grip calibration is the real sizeof the object and that in the present experiment theillusion reduced the correlation between object size andmaximum grip aperture only very slightly Grip aperturewas still scaled to the real size of the objects Moreoverthis scaling was sensitive to rather small differences inthe size of the goal objects the actual difference in sizebetween the large and the small disks was only 244 mmon average Evidently the visual processes underlyingprehension are nely tuned to the real size of an objectand these processes are quite resilient to the inuenceof perceptions that are at odds with the actual objectsize

There was another feature of the illusion design thatmay have played a role in the calibration of the graspthe fact that the target disk was surrounded by othervisual forms The equal and no annuli backgrounds in thepresent study were employed to examine the effect thatthe presence of surrounding annuli may have had on thecalibration of the grasp It is possible that independentof any possible illusory effects having an annulus sur-rounding the target disk may have created a situationsimilar to reaching into a hole with the end result thatgrip aperture was programmed to be tapered so that itwould t into that hole In fact the difference seenbetween reaches to disks on the equal annuli back-ground and disks on the no annuli background may havereected such an effect grasps made to the target diskwhen it was surrounded by the equal-sized annuli hadsmaller apertures than grasps to the same disk when itwas presented on a blank background This was particu-larly true for the small disk This difference may haveresulted from an attempt to taper the grasp appropri-ately This effect was not observed in the manual estima-tion condition In fact the effects were quite theopposite In the manual estimation condition the pres-ence of a surrounding annulus increased rather thandecreased the magnitude of the opening between thenger and thumb The fact that the disks were apparentlyperceived as larger when surrounded by annuli suggeststhat some sort of illusory effect was at work Althoughthe equal-sized annuli were designed to measure thedirect effect of having a surrounding annulus on bothresponse modes the annuli were still composed of cir-cles inevitably evoking a size-contrast effect based on arelative size comparison Because the two annuli wereidentical however any perceptual inuence of the an-nuli would be the same for both disks Nevertheless theperceptual effect could be seen when the manual esti-

mates with this background were compared to thoseseen when no annuli were present in the background

In conclusion a clear dissociation was shown be-tween the visual mechanisms underlying prehension andthe visual mechanisms underlying perception in neurol-ogically intact individuals Scaling of the grasp corre-sponded to the actual not the perceived size of thetarget disks manual estimations of the size of the diskscorresponded to the perceived not the actual size Thesendings are consistent with recent neuropsychologicalevidence suggesting that there are separate visual sys-tems for perception and action in the cerebral cortex(Goodale amp Milner 1992 Milner amp Goodale 1995) Eachsystem has evolved to transform visual inputs for quitedifferent functional outputs and as a consequence ischaracterized by quite different transformational proper-ties The parallel operation of these two systems lies atthe heart of the paradox that what we think we ldquoseerdquo isnot always what guides our actions In everyday life ofcourse the two systems work as an integrated wholemdashjust as all systems in the brain do Nevertheless the twosystems are complementary to each other and theirseparate evolution reects the different information re-quirements of perception and action

METHOD

Subjects

The 18 subjects (9 males and 9 females) in this studyranged in age from 19 to 28 years (mean 2267 years)and were all strongly right-handed as assessed by amodied version of the Edinburgh Handedness Inven-tory (Oldeld 1971) Their vision was normal or cor-rected-to-normal and their stereoacuity was within thenormal range as assessed by the Randot Stereotest (Ste-reo Optical Chicago) All subjects were paid for theirparticipation

Apparatus and Procedure

Hand position was recorded using the Optotrak (manu-factured by Northern Digital Inc Waterloo Ontario)which creates a 3-D representation of the trajectory ofthe hand and ngers by recording infrared light signalsThree high-resolution infrared-sensitive cameras moni-tored the position of infrared light-emitting diodes(IREDs) the positions of which were digitized at 100 HzOff-line the data were run through a low-pass second-order Butterworth lter with a 7-Hz cutoff

Subjects had IREDs placed on their index ngerthumb and wrist The IREDs were held in place withsmall pieces of cloth adhesive tape which allowed free-dom of movement of the hand and ngers The IREDson the thumb and index nger were placed at thecorners of the opposing nails (If one imagines the righthand placed at on a table the IRED on the index nger


would be on the left side of the base of the nail and theIRED on the thumb would be on the right side of thebase of the nail) The distance between these IREDs wasthe measure for maximum grip aperture and the manualsize estimations The wrist IRED was placed opposite thestyloid process on the ulna and provided measurementsof the transport component of the reach to the displayplatform (eg time to initiate the reach and velocity)

The display platform sat in a cardboard frame that wasattached to the table to maintain a constant position Thetable was covered with a black cloth which provided auniform viewing background for the display A circularoverhead uorescent light positioned approximately 75cm above the display provided illumination during theviewing period Onset and offset of the light were con-trolled by a computer with this system the light couldbe turned on by the experimenter for a specic lengthof time or in other conditions could be turned on andthen turned off when the subject released the startbutton The latter arrangement was used to create theopen-loop conditions in the visuomotor task thus pre-venting the possible inuence of visual feedback on thescaling of the opening between the nger and thumb

The target disks were constructed of 3-mm-thickwhite plastic with a thin black line drawn around thecircumference on the top surface and ranged in diameterfrom 27 to 32 mm (in 1-mm steps) During presentationtwo disks were placed on the display platform with thecenter of the disks spaced 120 mm apart The display satparallel to the surface of the table on a platform incor-porating a rotating ball-bearing mechanism which oper-ated in much the same fashion as a turntable Thisallowed the experimenter to easily alternate the leftright position of the display when necessary

The Ebbinghaus Illusion background was mounted onBristol board and attached to the rotating ball-bearingmechanism This formed the display platform The illu-sion background consisted of a small circle annulus(each of the 11 circles was 10 mm in diameter) and alarge circle annulus (each of the 5 circles was 54 mm indiameter) The inner diameter of the small annulus was38 mm the inner diameter of the large annulus was 55mm The centers of the two annuli were 120 mm apartIn order to examine the effect produced by having thetarget disk surrounded by an annulus independent ofillusion two additional backgrounds were employed Anequal annuli background was used consisting of sixcircles (each circle 22 mm in diameter) the inner diame-ter of each annulus was 42 mm The centers of the annuliwere also 120 mm apart A nal condition was used inwhich no annuli were present The target disks in thiscondition were always positioned 120 mm apart centerto center (the positions were indicated by small markson the background that were covered by the disks dur-ing presentation) The equal annuli background and noannuli background were presented in the same manneras the regular Ebbinghaus display The absolute position

of the target disks was identical for each backgroundcondition

In order that prehension and perception could bedirectly compared a visuomotor task and a perceptualtask were employed the order of which were counter-balanced across subjects The elements common to bothtasks will be described rst Following this the specicsof each task will be outlined separately

During testing subjects stood in front of the tablelooking down at the display surface This gave them abirdrsquos eye view of the display thus allowing them to seethe display straight on In both the visuomotor task andthe perceptual task subjects were asked to choose thetarget disk on the left or right side of the display basedon whether they thought the two disks looked the sameor different in size Half the subjects were instructed tochoose the disk on the left if they thought the two diskslooked the same the other half were asked to choosethe disk on the right The same instructions were main-tained for individual subjects across the two tasks Sub-jects were rst given some practice trials In addition toreceiving practice on the task they were systematicallytested with different sizes of disks to establish whichpair would be reliably judged as equivalent in size on theillusion background In other words the size of the diskin the large annulus of the illusory display was madelarger than that in the small annulus until the subjectsreported that the two disks appeared equivalent in sizeSubjects were not told that this psychophysical proce-dure was being carried out during the practice trials

The within-subjects variable for both the visuomotorand the perceptual task was the display condition inwhich the two disks were presented There were 14levels of this variable each of which consisted of acombination of three different components the back-ground display on which the disks were placed theleftright orientation of this display and the disk pairsthat were presented (ie physically identical or physi-cally different)

The background displays (Ebbinghaus Illusion theequal annuli and the no annuli displays) were presentedin randomly alternating trials As previously mentionedthe display was mounted on a rotating ball-bearingmechanism which allowed for easy alternation of theleftright position of the display Thus for the illusorydisplay sometimes the small annulus was on the left andsometimes the large annulus was on the left

For each of the three background displays two differ-ent disk pairs were presented In the rst type of trialthe disks were physically different The size of the largeand small disks used in these pairs had been determinedfor each subject during the practice trials On the illusionbackground the large disk was always placed in the largecircle annulus and the small disk was placed in the smallcircle annulus creating the illusion that the disks werethe same size With the equal annuli and no annulibackgrounds the large disk and the small disk were


presented an equal number of times on the left and rightsides of the display This ensured that the subjects wouldhave the opportunity to pick up a large disk and a smalldisk from the pair an equal number of times In thesecond type of trial the two disks were physically iden-tical on half the trials the two disks were both small andon the other half they were both large The small diskpair was composed of two versions of the same smalldisk from the physically different disk pair used toachieve perceptual equivalence with the illusory back-ground the large disk pair was composed of two ver-sions of the large disk from the perceptually equivalentpair On the illusion background the physically identicalpairs appeared to be different with the disk in the largecircle annulus appearing smaller than the disk in thesmall circle annulus To create a situation in which boththe large and small target disks would be surrounded byeach annulus on the illusion background an equal num-ber of times in each annulus position both pairs werepresented with the illusion background in both leftrightorientations In this way half the time the target diskwould (ideally) appear to be perceptually smaller andhalf the time it would appear to be perceptually largerFor the equal annuli and no annuli background condi-tions this counterbalancing was not required becausethe displays were symmetrical with respect to the dis-play background As a check for scaling constancy withthe equal annuli and no annuli backgrounds compari-sons were made between responses to each target diskfrom the physically different pair and those made to thetarget disk of matching size from each of the physicallyidentical pairs (This meant of course that comparisonswere being made between responses directed to targetdisks on the left of the display and those directed to thephysically identical disk on the right)

The 14 different disk presentation conditions wererandomly displayed ve times each for a total of 70 trialsin each task Subjects were given a break every 35 trialsTwo different versions of randomized trials were usedThe order in which these two randomized trial versionswere presented and the task (visuomotor or perceptual)assigned to each version were counterbalanced acrosssubjects In other words every subject performed boththe visuomotor task and the perceptual task and re-ceived both variations of the randomized trials Subjectswere asked to keep their eyes closed between trialswhile the experimenter set up the display

The Visuomotor Task

At the beginning of each trial an overhead light cameon allowing subjects to view the display and make theirdecision about whether the two disks looked the sameor different in size Then when subjects took theirthumb and index nger off of the start button to reachfor their chosen disk the overhead light turned off thuseliminating their view of both the target disk and their

hand during the reach Subjects were instructed to closetheir eyes between trials while the display and diskswere being set up The next trial began when the instruc-tions were given regarding the samedifferent decisionand subjects had verbally indicated they were in theready position with their thumb and index nger to-gether on the start button Half the subjects were giventhe instructions ldquopick up the disk on the right if the twodisks look the same and pick up the disk on the left ifthe two disks look differentrdquo The other half of thesubjects were given the opposite instructions concern-ing the target disk for the samedifferent choice

In the visuomotor task subjects were instructed topick up the disks with the index nger and thumb oftheir right hand at approximately the 1 and 7 orsquoclockpositions This position which is spontaneously adoptedby many people was used to maintain consistencythroughout the subjects Subjects were asked to placethe disk that they had picked up on the table to theright of the display platform This allowed the experi-menter to determine the subjectsrsquo samedifferent judg-ment of disk size (based on whether they had picked upthe disk on the left or right side of the display) withoutrequiring them to make a verbal report of their percep-tion which might have inuenced the way they tried toform their grip as they picked up the disk

The Perceptual Task

For the estimation task subjects were allowed to viewthe display for a xed period of 1 secmdasha period of timethat approximated the average length of time the displaywas in view during the visuomotor task in which thelight went off as soon as the subjects initiated theirreach Although this period of time was sufcient to viewthe display the manual estimate was always made whilethe light was off Thus in both the perceptual estimationtask and the visuomotor task the subjects were respond-ing in visual open-loop conditions

In the manual estimation task the subjects used thesame decision criteria as in the reaching task only thistime they estimated the size of the disk before reachingfor it Half the subjects were given the instructions ldquoes-timate the size of the disk on the right if the two diskslook the same and estimate the size of the disk on theleft if the two disks look differentrdquo The other half of thesubjects were given the opposite instructions A trial wasinitiated after the subjects were given these instructionsand had indicated that they were in the ready positionwith the heel of their hand resting on the start buttonand their index nger and thumb held together abovethe surface of the table At this point the light came onfor 1 sec allowing subjects to view the display Subjectsthen estimated the size of the chosen disk with theirthumb and index nger by matching the distance be-tween them to the diameter of the disk thus producinga perceptual measure of disk size The subjects indicated


when they had their size estimation ready and at thispoint their hand posture was recorded for 500 msecwithout the subjects being aware of when the recordingwas taking place Subjects were then cued with a tonewhich indicated that they could reach out and pick upthe disk they had just estimated In doing so subjectsreceived the same amount of haptic feedback about thephysical size of the disks size as they had during thevisuomotor task As in the visuomotor task the samedif-ferent perceptual decisions were noted based onwhether the right or left disk was placed beside thedisplay

Acknowledgments

The authors wish to thank an anonymous reviewer for themany helpful suggestions made on the manuscript This re-search was supported by a grant from the Medical ResearchCouncil of Canada to M A Goodale

Reprint requests should be sent to M A Goodale Departmentof Psychology University of Western Ontario London OntarioN6A 5C2 Canada or via e-mail goodaleuwoca

REFERENCES

Aglioti S DeSouza J F X amp Goodale M A (1995) Size-con-trast illusions deceive the eye but not the hand CurrentBiology 5 679ndash685

Bridgeman B Kirch M amp Sperling A (1981) Segregation ofcognitive and motor aspects of visual function using in-duced motion Perception and Psychophysics 29 336ndash342

Castiello U amp Jeannerod M (1991) Measuring time toawareness Neuroreport 2 797ndash800

Coren S (1971) A size contrast illusion without physicalsize differences American Journal of Psychology 84565ndash566

Coren S amp Enns J (1993) Size contrast as a function of con-ceptual similarity between test and inducers Perceptionand Psychophysics 54 579ndash588

Coren S amp Miller J (1974) Size contrast as a function ofgural similarity Perception and Psychophysics 16 355ndash357

Goodale M A (1993) Visual pathways supporting percep-tion and action in the primate cerebral cortex CurrentOpinion in Neurobiology 3 578ndash585

Goodale M A amp Haffenden A M (in press) Frames of refer-ence for perception and action in the human visual sys-tem Neuroscience and Biobehavioral Reviews

Goodale M A Meenan J P Buumllthoff H H Nicolle D AMurphy K J amp Racicot C I (1994) Separate neural path-ways for the visual analysis of object shape in perceptionand prehension Current Biology 4 604ndash610

Goodale M A amp Milner A D (1992) Separate visual path-ways for perception and action Trends in Neuroscience15 20ndash25

Goodale M A Milner A D Jakobson L S amp Carey D P(1991) A neurological dissociation between perceiving ob-jects and grasping them Nature 349 154ndash156

Graziano M amp Gross C G (1994) Mapping space with neu-rons Current Directions in Psychological Science 3 164ndash167

Gregory R L (1963) Distortion of visual space as inappropri-ate constancy scaling Nature 199 678ndash680

Gregory R L (1995) Realities of virtual reality [Editorial]Perception 24 1369ndash1371

Hochberg J (1987) Perception of motion pictures In R LGregory amp O L Zangwill (Eds) The Oxford companionto the mind (pp 604ndash608) Oxford Oxford UniversityPress

Jakobson L S Archibald Y M Carey D P amp Goodale M A(1991) A kinematic analysis of reaching and graspingmovements in a patient recovering from optic ataxiaNeuropsychologia 29 803ndash809

Jakobson L S amp Goodale M A (1991) Factors affectinghigher-order movement planning A kinematic analysis ofhuman prehension Experimental Brain Research 86199ndash208

Jeannerod M (1984) The timing of natural prehension move-ments Journal of Motor Behavior 16 235ndash254

Jeannerod M (1988) The neural and behavioral organiza-tion of goal directed movements Oxford Oxford Univer-sity Press

Jeannerod M Decety J amp Michel F (1994) Impairment ofgrasping movements following a bilateral posterior parie-tal lesion Neuropsychologia 32 369ndash380

Milner A D amp Goodale M A (1993) Visual pathways to per-ception and action In T P Hicks S Molotchnikoff amp TOno (Eds) Progress in Brain Research 95 (pp 317ndash337) Amsterdam Elsevier

Milner A D amp Goodale M A (1995) The visual brain inaction Oxford Oxford University Press

Oldeld R C (1971) The assessment and analysis of handed-ness The Edinburgh inventory Neuropsychology 9 97ndash112

Perenin M-T amp Vighetto A (1988) Optic ataxia A specicdisruption in visuomotor mechanisms I Different aspectsof the decit in reaching for objects Brain 111 643ndash674

Sirigu A Cohen L Duhamel J-R Pillon B Dubois B ampAgid Y (1995) A selective impairment of hand posture forobject utilization in apraxia Cortex 31 41ndash55

Soechting J F amp Flanders M (1992) Moving in three-dimen-sional space Frames of references vectors and coordinatesystems Annual Review of Neuroscience 15 167ndash191

Ungerleider L G amp Mishkin M (1982) Two cortical visualsystems In D J Ingle M A Goodale amp R J W Manseld(Eds) Analysis of visual behavior (pp 549ndash586) Cam-bridge MA MIT Press

Vishton P M amp Cutting J E (1995) Veridical size percep-tion for action Reaching vs estimating Abstract for The As-sociation for Research in Vision and Ophthalmology 36S358

Weintraub D J amp Schneck M K (1986) Fragments of Del-boeuf and Ebbinghaus illusions Contourcontext explora-tions of misjudged circle size Perception andPsychophysics 40 147ndash158


Damage to occipito-temporal cortex may lead to thereverse disorder in which the patient has impaired visualperception with spared visually guided movements Thisis the case with the patient DF whose grasping move-ments are well formed and accurate even though shecannot visually discriminate between or identify thesame objects to which she can direct well-formed grasps(Goodale et al 1994 Goodale Milner Jakobson amp Carey1991 Milner amp Goodale 1995) On the basis of theseneuropsychological studies and additional electrophysi-ological and behavioral studies in the monkey Goodaleand Milner (1992) have proposed that the distinctionbetween vision for perception and vision for action canbe mapped onto the two prominent pathways orldquostreamsrdquo of visual projections that have been identiedin the primate cerebral cortex a ventral stream whicharises from the primary visual cortex and projects to theinferotemporal cortex and a dorsal stream which alsoarises from primary visual cortex but projects instead tothe posterior parietal cortex (Ungerleider amp Mishkin1982)

Ungerleider and Mishkin (1982) originally proposedthat the ventral stream plays a special role in the iden-tication of objects whereas the dorsal stream is respon-sible for localizing objects in visual space Goodale andMilnerrsquos reinterpretation of this story places less empha-sis on the differences in the visual information that isreceived by the two streams (object features versusspatial location) than it does on the differences in thetransformations that the streams perform upon that in-formation (Goodale 1993 Goodale amp Milner 1992 Mil-ner amp Goodale 1993 1995) According to their accountboth streams process information about object featuresand about their spatial locations but each stream usesthis visual information in different ways In the ventralstream the transformations deliver the enduring charac-teristics of objects and their relations permitting theformation of long-term perceptual representations Suchrepresentations play an essential role in the identica-tion of objects and enable us to classify objects andevents attach meaning and signicance to them andestablish their causal relations Such operations are es-sential for accumulating a knowledge-base about theworld In contrast the transformations carried out by thedorsal stream provide the current location and disposi-tion of a goal object in egocentric coordinates andthereby mediate the visual control of skilled actionssuch as manual prehension directed at that object Ofcourse the two systems work together in controlling therich stream of behavior that we produce as we live ourcomplex lives Their respective roles in this control differhowever The perceptual representations constructed bythe ventral stream are part of a high-level cognitivenetwork that enables an organism to select a particularcourse of action with respect to objects in the worldthe visuomotor networks in the dorsal stream (and asso-ciated cortical and subcortical pathways) are responsible

for the programming and on-line control of the particu-lar movements that the selected action entails

Dissociations Between Perception and Action inNormal Observers

If visual perception and the visual control of actiondepend on different neural mechanisms in the humancerebral cortex it should be possible to demonstrate adissociation between these two kinds of visual process-ing in neurologically intact individuals In other wordseven in normal observers the visual information under-lying the calibration and control of a skilled motor actdirected at an object might not always match the per-ceptual judgments made about that object The trick isto nd the right task to demonstrate this dissociationThe present study used one particular paradigm a size-contrast illusion to do this

Most of the studies examining such dissociations innormal observers have focused on the spatial frames ofreference used by visual perception and the visual con-trol of action and have not looked at the effects offrames of reference on object size (for review see Goo-dale amp Haffenden in press Milner amp Goodale 1995) Thebulk of this work on spatial location suggests that themechanisms mediating the perception of object locationoperate largely in allocentric coordinates whereas thosemediating the control of object-directed actions (egsaccadic eye movements and aiming movements withthe limb) operate in egocentric coordinates In otherwords perception uses a coordinate system that isworld-based in which objects are seen as changing loca-tion relative to a stable or constant world the systemscontrolling action systems however cannot afford thesekinds of constancies and must compute the location ofthe object with respect to the effector that is directedat that target In short visuomotor control demandsdifferent kinds of visual computations than visual per-ception As we suggested earlier it is this difference inthe computational requirements of the visual control ofaction and visual perception that explains the divisionof labor between the two streams of visual processingin the primate cerebral cortex

Just as the perception of object location appears tooperate within relative or allocentric frames of refer-ence so does the perception of object size Although weoften make subtle judgments about the relative sizes ofobjects we rarely make judgments of their absolute sizeIndeed our judgments of size appear to be so inherentlyrelative that we can sometimes be fooled by visual dis-plays in which visual stimuli of the same size are posi-tioned next to comparison stimuli that are either muchsmaller or much larger than the target stimuli Suchsize-contrast illusions are a popular demonstration inmany introductory textbooks in psychology and percep-tion One such illusion is the so-called Ebbinghaus Illu-sion (or Titchener Circles Illusion) in which two target







The Present Study













RESULTS






















DISCUSSION





























METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES




































The Present Study













RESULTS






















DISCUSSION





























METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES








































RESULTS






















DISCUSSION





























METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES

































RESULTS






















DISCUSSION





























METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES













































DISCUSSION





























METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES






































DISCUSSION





























METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES
































DISCUSSION





























METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES























































METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES


















































METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES












































METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES




































METHOD

Subjects



















The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES












































The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES

































The Visuomotor Task




The Perceptual Task





Acknowledgments



REFERENCES
































Acknowledgments



REFERENCES































Documents

The Effect of Pictorial Illusion on Prehension and Perceptiontres/neuroII/yanagi.pdf · The Effect of Pictorial Illusion on Prehension and Perception ... haus Illusion. Aglioti,