An investigation of the relationship between eye and retinalinlage movement in the perception of movementl

ARIENMACK2NEW SCHOOL FOR SOCIAL RESEARCH

The question investigated was whetheror not eye movements accompanied byabnormal retinal image movements,movements t1uJt are either or both at adifferent rate or in a different directionthan the eye movement, predictably leadto perceived movement. Os reportedwhether or not they saw a visual targetmove when the movement of the targetwas either dependent on and simultaneouswith their eye movements or when thetarget movement was independent of theireye movements. In the main experiment,observations were made when the ratiobetween eye and target movement (em/tm)was 2/5. 1/5, 1/10, 1/20, and O. All theseratios were tested when the direction ofthe target movement was in the same (H+),opposite (H-), and at right angle's to(V+, V-) the movement of the eye~ t:vemovements, target movements, and reportsof target movement were recorded. Resultsindicate that a discrepancy between eyeand target movement greater than 20%predictably leads to perceived targetmovement, whereas a discrepancy of5% orless rarely leads to perceived movement.The results are interpreted as support forthe operation of a compensatorymechanism during eye movements.

What is the basis of perceiving themovement of a single object in anotherwise unstructured visual field? Themost obvious answer is that it is themovement of the retinal image that iscrucial,3 since with every movement of anobject there is a corresponding movementof the image of that object across theretina of the O. This answer is the moreattractive since there is now ampleevidence that vertebrate nervous systemsare well endowed with receptor cellsuniquely sensitive to retinal movements. If,however, retinal displacement were thecentral factor in perceiving objectmovement, it would follow that every timethere was retinal movement greater thanthe threshold of the detecting cellmechanisms, object movement· would beperceived, and in its absence, no suchmovement would be seen.

It is, of course, a well·known fact thatthe case is not so simple. There are toomany instances of perceived movement inthe absence of retinal movement and of the

correlative phenomenon of no perceivedmovement despite the occurrence of retinaldisplacement to maintain that retinal imagemovement is either a sufficient ornecessary condition for tile perception ofobject movement. An afterimage viewed inthe dark by an 0 moving his eyes appearsto move (Mack & Bachant, J969), butthere is no retinal movement. A movingobject in an otherwise empty fJeld,accurately pursued by the O's eyes, appearsto move, but there is no retinal movement.A stationary scene viewed by moving eyesappears stationary despite retinal imagemovements (position constancy). Theexistence of these and other phenomenasuggests that the relationship betweenretinal image movement and perceivedmovement is more complex. A theory thata ttempts to specify the relationshipbetween image or stimulus displacementsand perceived movement has beenvariously referred to as a cancellation,discounting, compensating, ortaking-into-account theory of movementperception (von Holst, 1954; Mackay,1967; Rock, 1966; Teuber, 1960).

If we are permitted to overlook thesubtle and not so subtle differencesbetween the various statements of thistheory, it becomes possible to set out itsmain outlines. All forms of this theory arein agreement that retinal image changesalone can neither account for theoccurrence of movement perception noralone determine whether the movementperceived resides in the object or in the O.It is a theory that accounts for perceivedobject motion as well as the failure toperceive object motion despitedisplacements or changes in the retinalimage. It assumes, in fact, that these arecorrelative events. The theory asserts thatthe basis of perceived object motion residesin the relationship between the behavior ofthe retinal image and the available sensoryinformation about the movement of the o.It assumes the existence of some kind ofcentral mechanism, a comparator(von Holst, 1954) that takes into account,matches, compares, or evaluates retinalimage shifts with reference to informationabout 0 movement. The theory assumesthat if not accounted for or matched byappropriate 0 movements, retinal imagedisplacements will invariably lead to the

perception of object mation andcorrelatively if eye movements are notmatched by retinal image displacements,mavement will be perceived. Thus. whenretimll: image changes are "matched,"constancy is predicted, and when thosechanges are not "matched," some changein visual perception is predicted. Constancyis predicted when retinal image changes areaccounted for in terms of 0 movementinformation. Constancy here signifies theperception of stability in the visual field,i.e., objects in the visual field appear tomaintain their positions despitedisplacements of the retinal images.Failures of constancy refer to thoseoccasions in which image displacements ofstationary objects lead to perceivedmovement in the visual field.

The postulation of some sort ofcomparator mechanism provides a conceptthat allows us to account for many of thefacts of movement perception. There isnow considerable evidence supporting theoperation of this sort of mechanism duringhead movements. Information about headmovements appears to mediate therelationship between visual displacementand perceived movement (Wallach &Kravitz, 1965; Wallach & Frey, 1969; Hay,1968; Rock, 1966). In these studies,abnurmal visual displacements,displacements that are either faster, slower,or in a different direction than the O's ownhead movements and do not simply reflectthe 0 movement, predictably led toperceived object motion. Until now therehas been no available evidence that acomparator operates during eyemovements. Since the eyes are rarely stilland since we generally see our visualenvironment as stationary despite theconstant displacements of the retinalimages produced by the fairly constantmovements of the eye, it would appearthat some kind of comparator mechanismin fact operates during eye movements.4

Yet in the only completely reportedexperiment examining the relationshipbet ween eye movemen ts, imagemovements, and the perception ofmovement (Wallach & Lewis, 1965) theauthors conclude that target movementsare not assessed in terms of eyemovements. By devising a situation inwhich the 0 saw a projection of his own

Perception & Psychophysics, 1970, Vol. 8 (SA) Copyright 1970, Psychonomic Journals, Inc., Austin, Texas 291

pupil, the authors produced abnormaldisplacements of a visual target dependenton eye movements. This technique made itpossible to produce a situation in whichthe target disk, the projection of the pupil,could be made to move at either a sloweror faster rate than the eye itself. Theirresults indicate that abnormal targetdis p lace men ts accompanying eyemovements, do not reliably cause reportsof perceived target movement. Theyconclude: "Our results are incompatiblewith the view that the apparent rest ofvisual objects whose images shift due toeye movements is to be explained by acompensating process which takes the eyemovements into account."

From the Wallach and· Lewis results itmight be possible to argue that allde tection of retinal displacement issuppressed during eye movements. If thiswere the case, then, since no retinaldisplacements would be registered duringan eye movement, there would be noreason to expect a loss of positionconstancy during eye movements, andthere would be no reason to postulatesome kind of comparator that could matchretinal displacement to 0 movementinformation. That this cannot be the casemay be seen from the fact that there isevidence that an afterimage appears tomove even when it is viewed in completedarkness and the only 0 movement is 0eye movement (Mack & Bachant, 1969).This suggests that at least the absence ofany retinal displacement during an eyemovement constitutes a baSis formovement perception. If the absence ofdisplacement is crucial, then there must bea record of the occurrence ofdisplacements in order that these two kindsof events be distinguished. It thereforeappears that there must at least be amechanism that records retinaldisplacements and eye movements and iscapable of determining when these eventsoccur simultaneously, which determinationordinarily results in no perceived objectmovement. If no evidence for such amechanism is found to operate during eyemovements, then it would seem to meanthat no real (as opposed to illusory)movement could be perceived during aneye movement except when the image ofthe visual object is stable on the retina (asin the case of an accurately pursued targetor afterimage).

To our knowledge the only availableempirical evidence that supports the viewthat a comparator mechanism operatesduring eye movement appears in Yarbus(1967). In his book, Yarbus reports anex perimen t in which he producedabnormal image displacements by means ofsuction-eontact lenses. Os report perceiving

target motion. Unfortunately the accountgiven is brief and no details are included.

The contradiction between the Wallachand Lewis results and those brieflyreported by Yarbus made furtherexamination of the relationship betweeneye movement, image movement, andperceived movement necessary. Anotherreason for such a study is the fact that aneye- m()ve me n t/ im age -m ove men tcomparator capable of doing nothing morethan merely discriminating betweeninstances of eye movements that are eitheraccompanied or not accompanied byretinal displacements would make it verydifficult to account for any perception ofmovement during eye movements.

By a technique different from that usedby either Wallach and Lewis or Yarbus, asituation was created in which eyemovements were accompanied byabnormal image displacements, that is,displacements that did not directly reflectthe movements of the eye. Os reportedwhether or not the target at which theywere looking appeared to move.

METHODThe essential requirements for work of

this kind are: a visual target whosemovements are strictly related to themovements of the O's eyes, a means ofaltering the relation between eye and targetmovement, and a means of simultaneouslyrecording eye movements and the reportsof movement made by the O.

Lateral eye movements were recordedby means of electrooculography. Beckmanbipotential skin electrodes were attachedto the outer margins of the eyes and a thirdelectrode was attached to the center of theO's forehead and served as a ground. Theelectrodes were wired through a noise filterto one channel of a Sanborn four-channelpen recorder with sufficient dcamplification to represent the signal fromthe eyes accurately. The output of thischannel was wired in tum to an externalvolume control from which it was fed intoeither the horizontal or vertical inputchannel of either one or two TechtronixNo. 502A oscilloscopes. The volumecontrol permitted continuous control overvariations in gain. The effect of this was toafford continuous control over the amountof oscilloscope trace movement producedby an eye movement. The oscilloscopetrace served as the visual target. It was theonly object visible to the 0 during testingsessions. In order that the 0 should not seethe glow of the oscilloscope screen itself,Kodak Wratten fIlters, No. 47B, wereplaced over the screen and during testingall Os wore goggles containing KodakWratten ffiter No. 35. The movements ofthe oscilloscope trace, which was

approximately 2 mm in diam, wererecorded by one channel of the penrecorder. Simultaneous recording of eyeand target trace movemertt permitted acontinuous check on the relation betweeneye and target movement. The volumecontrol knob was calibrated from 0 to I.When this knob was set at 0, the trace wasnot moved at all by eye movements. Whenthis knob was set at 1, the trace moved thesame amount as the eyes and thus couldapproximate a stabilized image (Le., whenthe direction of trace displacement and eyemovement were the same). When the knobwas set at some point between 1 and 0, thetrace moved some fraction of the amountthe eyes moved. Thus, when the knob wasset at 2/5, the target trace moved 2/5 ofthe distance that the eyes moved. (Allcalibrations were frequently checked.) Theincrease in what is best described as tracejitter with increase in gain were mostlyeliminated by filtering. What jitterremained was accounted for in theexperimental design.

A switch on the control panel permittedrapid shifts from the horizontal to thevertical input channels of the oscilloscopes.This switch had the effect of making thetrace move either horizontally or verticallyin response to lateral eye movements of theO. A second switch, a plus-minus switch,changed the direction of trace movementaccompanying an eye movement. Whenthis switch was set at + and thehorizontal-vertical switch was set athorizontal, the trace moved in the samedirection as the eyes. If the eyes movedright, the trace moved right. If the +­switch was set at -, the trace moved leftwhen the eye moved right. When thehorizontal-vertical switch was set atvertical, then a + setting of the + - switchcaused the trace to move up when the eyesmoved right. A - setting caused the traceto move down when the eyes moved right.The various control switches permittedfour directions of target movement,horizontal with the eyes (H+), horizontalagainst the eyes (H-), vertical and up withthe eyes to the right (V+), and verticaldown with eyes right (V-), as well ascontinuous control over the ratio ofeye-to-target movement, em/tm. Thefollowing ratios of eye-to-target movement,em/tm, were repeatedly tested: 0, 1/20,1/10, 1/5, 2/5. Ratios of 1/2 and I weretested only under conditions of randomeye movements (to be explained).

Small instant on-offlamps were attachedto both sides of one of the oscnJoscopescreens. Each light was a distance of6.25 cm from the center of the screen.Switches on the control panel turned oneither one or the other of these lights.Operations of one of these switches

292 Perception & Psychophysics, 1970, Vol. 8 (SA)

resulted in a light flash that lasted less than1/20 sec. The lamps were coated and tapedwith an opaque material so that the lightflashes were sufficiently dim that noafterimages were produced even indark-adapted eyes. The light switches werewired to an event marker on the penrecorder providing a permanent record ofthe onset of light flashes.

Each oscilloscope was housed in ashielded and light-tight cell in which therewas sufficient room for an a to sit. Thecenter of the oscilloscope screens coincidedmore or less with an eye level of the as.Bite bars were fixed 27.5 cm in front ofthe oscilloscope screens and preventedhead movements during testing. Positionedat an upright oblique angle and to the rightof the screens were panels consisting offour buttons placed at the four points of across. These were used by the a to reporttarget movements. The top button wasused to report upward movement, thebottom button for downward movement,the right button for rightward movementand the left button for leftward movement.The reporting buttons were wired throughthe dc amplifier to a channel on the penrecorder. Depression of one of thesebuttons moved the pen in that channel aspecific amount, so that the distance thepen moved indicated which button wasbeing pressed. The two oscilloscopes wererun simultaneously and at all timesdisplayed identical trace movementbecause the input to and the calibration ofthe oscilloscopes was always identical.

ProcedureThe as were run in pairs, one who

observed a target whose behavior dependedon his eye movements (experimental), andthe other who observed the same targetwithout its being dependent on his eyemovement (control). Each a was testedtwice, once as an experimental and once asa control O. The advantages of testing eacha twice and of running as in pairs are thatit provides control data for each testingsession, and it allows each a to act as hisown control, thus accounting for styles ofreporting. Since we obtained data fromtwo as simultaneously, the occurrence ofany electronic jitter with increases in gainwas not significant, because we were ableto analyze the difference between thereports of the two as. The experimental awore the eye electrodes and sat in theoscilloscope cell containing thelight-flashing apparatus. The control a satin the other oscilloscope cell, which wasidentical to that in which the experimentala sat except that there were no flashinglights. There were two parts to each testingsession. An entire testing session lastedabout 'Jll h. The two parts are designated

directed eye movements (DEM) andnondirected eye movements (NDEM).

DEM. The experimental a wasinstructed to look straight ahead at theoscilloscope trace, and bite on the bite bar.This signaled the onset of a trial.Immediately following instructions to lookstraight ahead and a check of theeye-movement record to establish that theo had properly responded, one of the lightswitches was flipped. The a then turnedhis eyes to where he had seen the lightflash and was required to keep his eyes inthis position until he was once againinstructed to look straight ahead. Duringthe interval between turning his eyes to theflash and returning his eyes to straightahead, a reported if there had been anytarget trace movement. He reportedwhether or not the trace moved when hemoved his eyes. All instructions were goneover carefully before testing occurred, andall testing was preceded by at least 10practice trials in which all the 0 wasrequired to do was to move his eyes to theflashed light and hold them in that positionuntil once again instructed to look straightahead. Despite the apparent complexity ofthe a's task, most as easily succeeded indoing what was required of them.

Control as wore no eye electrodes andtheir eye movements were not recorded.They were asked to look straight ahead andbite on the bite bar on instruction from theE (the same instructions were heard byboth as) and to report if the trace movedwithin the interval between thestraight-ahead instructions and theinstructions to the experimental a to onceagain look straight ahead. Both as weretherefore reporting on the behavior of thetrace during the eye movements of theexperimental O. Although no record of theeye movements of the control as weremade, it is probably safe to assume thatcontrol as' reports were made on the basisof a period in which their eyes moved verylittle since they had been instructed tolook straight ahead.

Each combination of displacement ratioand trace movement direction waspresented in blocks of four trials separatedby at least a brief rest period. Thus a blockof trials consisted of four trials in whichthe em/tm and the direction of targetdisplacement was constant. Several of theseblocks of trials were separated by longerrest intervals in which any dark adaptationwas erased. Each block of four trialsconsisted of two left and two right lightflashes presented in random order. In alltesting sessions, each combination of targetdisplacement direction and em/tm ratiowas presented for at least one block oftrials. Both ascending and descendingorders of em/tm were used. Blocks of trials

in which the em/tm was zero werepresented at least three times during eachtesting session. Zero testing trials wereusually presented after a block of trials inwhich the em/tm was either 2/5 or 1/20.Blocks of trials were ordered so that fourconsecutive blocks of trials were presentedin which the em/tm was constant while thedirection of target displacement was variedfrom one block of trials to the next.

NDEM. This condition was included asan attempt to verify the differencebetween "incidental" and "voluntary" eyemovements reported by Wallach and Lewis(1965). They found that abnormal imagedis p lacemen ts accompanying eyemovements sometimes led to movementreports when eye movements werevoluntary but rarely when eye movementswere incidental. In this condition both Osreceived exactly the same instructions.They were directed to look continuously atthe target trace and to indicate if it movedby pressing the right button of the buttonpanel for as long as the trace continued tomove. Unlike the DEM reports, only thepresence or absence of movement and notits direction was reported. The task ofconstantly monitoring a small and dimtarget light was sufficiently taxing withoutrequiring as to also report direction oftarget movement.

The following em/tm ratios were testedin combination with the four directions oftrace movement: 1, 1/2, 2/5, 1/5, 1/10,1/20, and O. Each combination of em/tmratio and direction of target displacementwas presented for a IS-sec interval at theend of which the em/tm ratio, thedirection of target displacement, or bothwere changed. as were allowed to restapproximately every minute.

Half the Os began testing with DEM,while the remaining half began withNDEM. Half the as were experimental Osbefore they were tested as controls. Theremaining half was tested in the reverseorder.

SubjectsFourteen Ss (seven pairs) were each run

twice_ These Ss were Stanford Universitystudents who were paid for theirparticipation.

RESULTSDEM

Data from the DEM condition wereanalyzed in terms of the percent of correctresponses in each block of four trials. Forexample, if a S reported target movementto the left following eye movements to aright light flash, and target movement tothe right following eye movements to a leftlight flash in three of the four trialsconstituting the block of trials in which the

Perception & Psychophysics, 1970, Vol. 8 (SA) 293

H V Grand Mean

EMrrM E C E C E C

2/5 89.64 98.39 94.64 100 92.11 99.201/5 84.11 92.86 87.68 99.11 85.89 95.981/10 40.18 85.72 46.43 92.86 43.30 89.291/20 10.18 76.25 12.33 64.46 11.25 70.36

CV =48.83, P =.01 CV =23.56, P =.01CV =40.49, P =.05 CV =20.72, P =.05

Table IMean Percent Correct Movement Responses: DEM

H- H+ V+ V- Grand XEM/TM E C E C E C E C E C

2/5 87.5 96.8 91.8 100 94.6 100 94.6 100 92.1 99.21/5 78.5 91.1 89.6 94.6 87.5 100 87.5 98.2 85.7 95.91/10 41.1 83.9 39.2 87.5 51.8 92.9 41.1 92.3 43.3 89.11/20 13.2 64.6 11.4 64.3 12.5 75.7 7.8 76.9 11.2 70.30 93.2 81.7

CV = Critical Value, E = Experimental, C = Control, H = Horizontal, V = Vertical

Table 3Mean Scores and Critical Values for Estimating the Significance

of Difference Between Means (Scheffe Procedure)

during a saccade as target movement that is2/S as great as the eye movement.

An analysis of means obtained in controlconditions reveals the only significantdifference to be that between thecondition in which em/tm was 1/20 and allthe remaining em/tm conditions (seeTable 3). Very small target displacementsappear to be difficult to detect regardlessof whether or not the;' are simultaneouswith an eye movement.

Statistical comparisons were made inwhich each experimental S was comparedwith himself as a control 0 (see Table 3).The following differences were found to besignifican t: all differences betweenexperimental and control Os in conditionsin which the em/tm equals 1/10(significant at greater than .01), and alldifferences between experimental andcontrols for em/tm equals 1/20 (significantat greater than .01). Thus, in the O. 2/5,and l/S conditions, there are no significantdifferences between experimental andcontrol Os in their ability to perceive targetmovement, although the differences thatdo appear between these groups are allcharacterized by the fact that the controlmeans are greater. When the em/tm was1/10 or 1/20 Os were able to moreaccurately detect target displacement inthe control conditions than were those Osin the experimental conditions who werereporting on target displacementsconcurrent with eye movements.

An analysis of the errors made byexperimental Os reveals no particulartrend. This analysis is potentially mostinteresting in the case of errors made whenthe em/tm level is zero, since these errorsmight possibly reveal a tendency to see astationary target moving in a directionpredictably related to the direction of theeye movement. Errors were extremely fewin this condition and reveal no trend. Mostcases in which the percent correct scoreswere less than 100% were the result of afailure to report any movement rather thanreports of movement in an objectivelyincorrect direction.

An analysis was made of reaction time(RT) to the light flashes in order todetermine if on any trials the light wasvisible when the eyes began to move. Anytrials in which this had occurred wouldhave to have been ignored since reports ofmovement derived from these trials mighthave been based on the availability ofobject-relative displacement information}In no case did RT approximate theduration of the light flash, which was1/20 sec. Typically, RT was slightly lessthan 1/2 sec. The only explanation we canoffer for the prolonged duration of the RTis that the complexity of the O's tasktended to make his responses to the light

.05

<.001

<.001

the immediately adjacent em/tm level.These comparisons were again madeseparately for experimental and control Os(see Table 3). For experimental Os, themean percent correct responses for the 2/5em/tm level is not significantly differentfrom the mean percent correct responsesfor em/tm equals lIS. Movement wasreported as frequently when the targetdisplaced 1/5 the distance the eyes movedas when the target displaced 2/S of thatdistance. The following differences aresignificant: The difference between meansin the l/S em/tm condition and means inthe 1/10 em/tm condition and thedifference between means in the 1/10 andmeans in the 1/20 em/tm condition are allsignificant at greater than .01. For allem/tm levels less than l/S, the smaller theem/tm ratio the more difficult it is for theo to detect target displacement during aneye movement. This may be stated anotherway. The smaller the absolute extent oftarget displacement during an eyemovement the less likely it is to bedetected. The difference between means inconditions in which the em/tm is 2/5 and 0are not significant. Ss appear to be asaccurate in reporting no target movement

Table 2Analysis of Variance Outcome (DEM)

F DF

177.546 1/416

46.269 1/416

1.044 1/416

121.147 31416

em/tm level was 1/5 and the direction oftarget movement was H-, he received ascore of 75%. Table 1 presents the meanpercent correct scores of all experimentaland control Os for all combinations oftarget-movement direction and levels ofem/tm. An analysis of variance wascomputed for this data. The results of thisanalysis appear in Table 2. All three mainfactors had a significant effect on theperception of movement. Using the Scheffeprocedure for all possible a posterioricomparisons (Winer, 1962) the differencesbetween means obtained in each conditionwere evaluated. The outcome of thisanalysis is found in Table 3. Statisticalcomparisons between means obtainedwhen target movement was horizontal andwhen it was vertical and when the em/tmwas constant yield no significance (seeTable 3). These comparisons were madeseparately for experimental and control Os.The direction of target displacement doesnot appear to affect the perception ofmovement significantly under theseconditions.

Again using the Scheffe procedure, allmeans obtained at one level of em/tm werecompared with the means obtained from

Factors

GroupsExperimental and Control

DirectionHorizontal and Vertical

DirectionPositive and Negative

EM/TM215, I/S, 1/10, 1/20

294 Perception & Psychophysics, 1970, Vol. 8 (SA)

Table 4Range (in Degrees) of Target Movement

Scores calculated on basis of percent of timeduring each 15-sec trial in which movementis reported, expressed as a percent of 15 sec.

Table 5Mean Percent Movement Reports: NDEM

E C

1 69.6 65.21/2 55.0 59.82/5 52.5 561/5 35.0 40.31/10 31.3 41.11/20 18.8 27.1

NDEMThe results of this condition are not

simply described. Only two quantitativedescriptions of the data are possible; one isthe time dUring each IS-sec trial in whichthe 0 reports seeing target motion,expressed as a percent of 15 sec. The otheris the number of movement reportsfollowing eye movements of a certain

and significant difference betweenexperimental and control Os. Also in bothconditions displacement ratios less than1/10 invariably lead to a significantdecrement on the part of experimental Osin the detection of target movement.

In the NDEM condition for bothexperimental and control Os a decrease inem/tm level produced a correspondingdecrease in the amount of movementreported or in the number of movementreports. This can be seen in the OEMresults as well. A clear difference, however,between this data and the data from theOEM condition is the marked degree ofsimilarity between the control andexperimental scores. The decrease in theamount of movement reported byexperimentals and controls with decreasesin em/tm level appears almost as striking inthe control as in the experimental data. Weattribute this similarity betweenexperimental and control scores to thedifficulty in detecting any targetmovement in this condition. This may beexplained by the fact that the smaller theem/tm ratio the less the actual targetdisplacement, and the less the actual targetdisplacement the less likely was theexperimental 0 to make eye movementsthat were sufficiently large to displace thetarget a detectable amount. It is thereforenot surprising that both experimental andcontrol Os show a decrease in ability todetect movement with a decrease in levelof em/tm that appears to be independentof whether or not target displacementsoccur during the O's eye movements.

Despite this similarity between thepatterJ1 of experimental and control scores,the pattern of significant differences is nottotally unlike that found in the DEM data.As in the DEM results, the only significantdifference that appears in the controlscores is that between the scores obtainedwhen the em/tm equaled 1/20 and thescores obtained from all other em/tmlevels. For experimental Os, the scoresobtained from trials in which the em/tmratio was 2/5 or larger are significantlydifferent from scores obtained on trials inwhi~h the em/tm was less than 2/5. Thisdifference is not found in the DEM resultsin which the first significant break isbetween scores obtained from trials inwhich this ratio is less than 1/5. NDEMresults show no significant differencebetween scores from the 1/5 and 1/10conditions. Thus, with spontaneous eyemovements, target movement is no moredifficult to perceive when that movementis dependent on the eye movement and 1/5as great as the eye movement than when itis 1/10 as great. As in the DEM results,there is a significant difference between allexperimental scores obtained when the

extent, for example, larger than 3 deg,expressed as the percent of the totalnumber of eye movements of that extentor larger occurring in any given IS-sec trial.Both ways of analyzing the data were used.

One of the drawbacks of a condition ofspontaneous eye movements for aninvestigation of this kind is that Os tend tomake very few spontaneous eyemovements that are larger than I deg underthese experimental conditions, Le., whenthe only visual target is a dimly glowingspot of light which moves only as somefunction of the O's own eye movementsand which they are instructed simply toobserve. When the em/tm level is low, theam ou n t of target displacementaccompanying a relatively small eyemovement is infinitesimal. Even with eyemovements as large as 3 deg, the amount oftarget displacement is almost negligible.Thus there is only a slim chance that thereare a sufficient number of occasions inwhich even a control 0 is able to detecttarget movement. Furthermore, in thiscondition, Os are required to monitor thebehavior of the target light constantly,which is likely to cause fatigue and

, consequent inaccuracies.Table 5 depicts the results of the NDEM

condition. Despite these drawbacks, theresults obtained generally resemble thoseof the OEM condition. Comparing each 0with himself as a control, the differencebetw~ the amount, the percent of eachIS sec in which movement was reported,for trials in which the em/tm ratio was 1/2,2/5, 1/5, 1/10, and 1/20 are as follows:3.6%, 4.0%, 5%, 10%, and 12%,respectively. In the 1/2, 2/5, and 1/5conditions, the differences in amount ofmovement reported by both control andexperimental Os are trivial, whereas thedifferences between experimental andcontrol scores when the em/tm was either1/10 or 1/20 are significant. This samepattern appears if a comparison is madebetween the number of movementresponses made by experimentals andcontrols following eye movements 3 deg orlarger at each of the em/tm levels (againtreating each 0 as his own control). Thesedifferences are as follows for the 2/5, 1/5,and 1/10 em/tm trials: 11%, 11.4%, and17.7%, respectively. The only significantdifference is between experimental andcontrol scores in the 1/10 em/tmcondition.

Here, as in the OEM condition, directionof target displacement did not significantlyaffect movement reports. More important,in both conditions there is a significantdifference between experimental andcontrol Os where the em/tm level is equalto 1/10, and for both conditions this is theHrst level of em/tm that produces a clear

Range

6.2-4.22.1-3.11.6-1.1.8- .5

EM/TM

2/51/51/101/20

flashes more deliberate than they otherwisemight have been.

An analysis of the size of the eyemovements made in response to the lightflashes indicates that eye movements variedin size between 10.5 and 15.6 deg, with theaverage eye movement covering about13 deg of visual angle. An eye movementof 13 deg did, in fact, carry the eye fromits focus at the center of the oscilloscopescreen to either light, a distance of62.5 mm. Variations in the size of an eyemovement did not appear to affectmovement reports. Reports of movementwere no more frequent or accurate wheneye movements were on the large end of theeye-movement range than when they wereon the small end of this range. Variationsin the size of the eye movements did, ofcourse, affect the absolute amount oftarget movement that accompanied anygiven eye movement. Since these variationsin the size of eye movements did not seemto affect movement reports, theconcommitant variations in the absoluteamount of target movement also seem tohave had no affect on these reports. Theability to perceive target movement seemstherefore to depend on the em/tm ratiorather than on the absolute amount oftarget movement. Table 4 describes therange of variation in the amount of targetmovement produced by variations in eyemovements in the various em/tmconditions.

Perception & Psychophysics, 1970, Vol. 8 (SA) 295

Table 6Mean Percent Correct Reports: OEM

(Closer Viewing Distance)

V+ V- H+ H- Grand X

2/5 98.8 96.3 95.0 87.5 94.41/5 92.5 86.1 71.9 58.9 77.21/10 72.2 55.1 43.3 44.4 53.51/20 35.7 33.5 7.5 20 24.1

em/tm ratio was 1/20 and when the ratiowas 1/10 or greater.

That our results are not limited to eyemovements of a particular absolute size isclear both from the results of the NDEMcondition and from another study whichpredated the one here described. Thisearlier study resembled the DEM conditionof the study reported with the followingexceptions: The Os sat closer to theoscilloscope screen; the bite bar was placed12.5 em from the screen so that an eyemovement of 26 deg was required to movethe eye from the center of the screen toone of the lights; and no control conditionwas employed. Each 0 was run for twoexperimental testing sessions but Os werenot tested in pairs. In every other way, thisexperiment was identical to the DEMcondition of the experiment reported. TenOs were tested. The results appear inTable 6. With a few exceptions, theyresemble those results already reported.The following are the exceptions: V 1/5scores are notably greater than the scoresin the H 1/5 conditions, although thisdifference is not significant at .05. (Thecritical value for a difference that issignificant at .05 is 35.165, Scheffe, 1953.The difference value obtained is 23.3.) Thistrend, which was absent in the datapreviously described but present in all butthe 2/5 condition in this experiment,conforms to data reported by Wallach,Frey, and Rodney (1969) for headmovement accompanied by abnormaldisplacements. Displacements that are atright angles to head movement appear tobe more easily perceived. It is not clearwhy this tendency is restricted to this setof data. Another exception is that themean scores in the 2/5 condition aresignificantly greater than those in the 1/5condition. (Critical value for a difference at.05 is 16.3. The difference obtained is17.2.) A difference between a displacementratio of 2/5 and one of 1/5 does appear inthe NDEM data. The means obtained withan em/tm of 1/20, with the exception ofthe H+ 1/20, are greater than the analogousscores in the previously reported DEMresults. This may reflect the fact that in thepreviously reported results when theem/tm was 1/20 an average eye movementdisplaced the target .78 deg, whereas herethe target was displaced 1.3 deg.

The point to be made is that despite

some differences all results indicate thatwhen target displacements are 2/5 as greator greater than a concurrent eye movement(in the main experiment, when targetdisplacements are 1/5 as great or greater),that displacement is predictably perceived,and when the displacement is only 1/20 asgreat as the eye movement thatdisplacement is generally not perceived.

DISCUSSIONWhen the eye scans a stationary object,

the image of that object displaces acrossthe retina of the eye in the same directionand to the same extent that the eye hasmoved. This is the normal condition forperceiving position constancy and isidentical with the experimental conditionin which the em/tm level is zero and thereis no objective target displacement. A lookat the data obtained in the DEM conditionwhere the em/tm is zero indicates that, infact, movement is generally not reportedunder these circumstances. Targetdisplacements 1/5 as great or greater thanthe accompanying eye movement were asreliably perceived as the identical targetdisplacements that were independent ofeye movements. This result conclusivelyestablishes the fact that targetdisplacements concurrent with eyemovements are not suppressed and thatthere is no complete saccadic suppressionof movement perception during saccadiceye movements. Results in the DEMconditions support the conclusion thatdirection of movement perceivedcorresponds to the actual direction oftarget displacement.

A second conclusion to be drawn is thatthese results support the concept of acomparator mechanism that matchesavailable target displacement informationwith available information about 0 eyemovements.

If this hypothetical comparator were tooperate with extreme precision, only aretinal-image displacement that exactlymatched an eye movement wouldconstitute the condition for perceivedposition constancy, that is, no movementperceived despite retinal-image movement.Thus, if the eye turned 13 deg to the right,only a 13-deg rightward displacement overthe retina would constitute aposition-constancy match. It is apparentfrom the data that eye-movement/image­movement comparisons are not so precise.I n most instances when the targetdisplacement is 1/5 as large as the eyemovement, regardless of the direction, thedisplacement is perceived. Thus, if, forexample, the eye moves 13 deg right andthe target shifts 2.6 deg in the same (H+)or the opposite (H-) direction, the imageof the target will displace either 11.4 or

15.6 deg to the right over the retina, andmovement will be seen. This seems toindicate that when the discrepancybetween eye and image movement is 20%or greater, object movement will always beseen. If, however, the eye moves and thetarget displaces 1/20 of that distance,movement is generally not seen. Thus, ifthe eye moves 13 deg to the right and thetarget moves .65 deg either in the same(H+) or the opposite (H-) direction, theimage will displace either 12.35 or13.65 deg right over the retina. Underthese circumstances, movement will notgenerally be seen, indicating that adiscrepancy of 5% between eye and imagedisplacement is acceptable and constitutesgrounds for the perception of a stable,unmoving visual environment. That a 5%target displacement is equivalent to targetimmobility can be seen by comparing thescores in the OEM condition when theem/tm was 1/20 with those scores obtainedwhen the em/tm was zero. To make thetwo scores comparable, we can subtract thescores in the 1/20 condition from 100%,arriving at a mean score of no movementreports amounting to 88.2%, which isentirely comparable to the score of 93.2 inthe 0 condition in which there is actuallyno target displacement.

The fact that target displacements thatare 1/10 as large as concurrent eyemovements are not perceived as frequentlyas the identical displacements occurringindependently of eye movements, but aremore frequently seen than displacementsthat are only 1/20 as large as the eyemovement, suggests that a discrepancy asgreat as 10% between eye and imagedisplacement is close to the threshold ofwhat constitutes an acceptable basis forperceiving stability during eye movements.

It might be argued that these results aresimply to be interpreted as an index of thecapacity for movement detection duringsaccadic eye movements. Looked at in thisway the results indicate that the directionof target displacement during a saccade isnot a factor in determining whether or notthat displacement will be perceived, and,further, that there is the expected raising inthe movement detection threshold duringan eye movement. While this account ofthe results is legitimate, it in no wayinvalidates or contradicts an account of theresult in terms of a comparator mechanism.To insist on an account of the data strictlyin terms of thresholds and to rule out asillegitimate any other kind of account is toconfuse description and explanation. Theconcept of a cOIi~parator or compensatorymechanism is explanatory. The results ofthese experiments support this kind ofexplanation. To speak of threshold issimply to employ a well understood and

296 Perception & Psychophysics, 1970, Vol. 8 (SA)

efficient terminology for describing results.To infer from these results to the operationof a comparator mechanism is an attemptto account for them.

A puzzling aspect of the results is thatthey fail to show any consistent effect ofthe direction of target displacement. Forthe most part, the results indicate that theem/tm is the central variable. Since undernormal position-constancy conditions, theimage of a stationary object displaces inthe same direction over the retina as theeye moves, it might be reasonable toexpect that experimental conditions inwhich the image of the target displaces inthe same direction as the eye might leadless frequently to perceived movement, allother things being equal, than those casesin which the image displaces in some otherdirection. The only results that evensuggest that this may be the case are fromthe early DEM study in which the 0 satrather close to the oscilloscope screen. Asreported, the mean percent correct scoresat each level of em/tm are here somewhathigher for the two V conditions than forthe two H conditions. This is exactly whatone might expect, since in both Hconditions, regardless of the level ofem/tm, the image of the target displaced inthe same direction as the eye moved,whereas in the V conditions the target wasalways displacing at right angles to the eyeso that the image was always displacing atan angle to the direction the eye moved.No difference could be expected betweenthe H conditions, and generally noneappears. Further testing is required inwhich image displacement is directlyopposite to the direction of eye movementbefore any conclusions are drawn aboutthe relevance of the direction of imagedisplacement for the perception of targetmovement.

Another question left unanswered bythis work is whether it is the percent ofdiscrepancy between eye and imagemovement or the absolute discrepancybetween these two inputs that determinesif object movement will be seen. It ispossible that whenever there is adiscrepancy between eye and imagemovement that is larger than 10%, targetmovement will generally be seen-or it ispossible that whenever the discrepancybetween eye and image movement isgreater than, for example, 1.5 deg,movement will be seen. The fact that theresults of the various conditions show thesame pattern suggests that it is the percentof discrepancy which is crucial, but morework is required.

One could ask why the postulatedcomparator operation during eyemovements is not more precise. Thisquestion is unresolved by this study, but

several possibilities suggest themselves.Perhaps the fact that a 5% mismatchbetween eye and visual displacementinformation is an acceptable basis forposition constancy during eye movementsis a function of a saccadic inhibition ofvisual displacement detection during eyemovements such that a 10·degdisplacement is indistinguishable from a9-deg displacement.

Still another possibility might be thatthe tolerance for a mismatch as a basis forposition constancy may be the result of afaulty memory for egocentric direction. Ifthe crucial components in the comparisonbetween eye and visual displacementinformation were eye·movementinformation on the one hand andinformation about the visual direction ofan object before and after an eyemovement on the other hand, rather thaninformation about retinal movement, thenthe tolerance for a mismatch might be theresult of an inherent inability to remembervisual directions accurately. If an objectwere straight ahead at the onset of an eyemovement and moved I deg during an eyemovement of 10 deg, perhaps the reasonthat target movement is not perceived isthe result of an inability to rememberprecisely where the object was at the onsetof the eye movement. If this were the case,a I-deg displacement might beindistinguishable from a O-degdisplacement and would account for thefact that it is not perceived. In consideringthis explanation it is necessary to assumenot only that memory for positionbecomes more inaccurate when theposition to be remembered precedes an eyemovement but, more important, thatretinal movement is still registered in thematch between information about eyemovement and change of visual direction,since otherwise we would have to concludethat no real movement is perceived duringan eye movement. If retinal movementinformation did not playa role, only theinference of movement based on a changein visual direction would be possible. Fromour own work, it would be extremely hardto argue that real movement is not seenduring a saccade, since Os reported nodifference in the appearance of targetmovement when they were tested underexperimental and control conditions.

Finally, it might be argued that theapparent imprecision in the match betweeneye and target movement is a function ofinaccurate or imprecise eye-movementinformation. If the eyes were to tum10 deg to look at something off to the side,and the information registered in thenervous system is that the eye has movedeither more or less than 10 deg, perhapssomewhere between 9 and 10 deg. then a

I-deg image displacement in either thesame or the opposite direction as the eyemoved would "match" the eye movement,and no movement should be perceived. Theresults of this experiment do not permitany evaluation of the relative merits ofthese possible explanations of the apparentimprecision in the "match" between eyeand image movement.

These results reconfirm the fact thatretinal displacement cannot be considereda sufficient condition for movementperception since in all cases in which targetdisplacements are in the same direction asthe eye movement (H+), the higher theem/tm level the less the retinal imagedisplacement, yet the lower the em/tmlevel the less likely is target movement tobe perceived. In conclusion, these resultsclearly contradict those reported byWallach and Lewis (1965) and extendthose of Mack and Bachant (1969). Theyconstitute evidence that "abnormal"retinal movements concurrent with eyemovements do lead to perceivedmovement.

REFERENCESBRINDLEY, G. S., & MERTON, P. A. The

absence of position sense in the human eye.Journal of Physiology, 1960, 153,127-130.

HAY, J. Visual adaptation to an alteredcorrelation between eye movements and headmovements. Science, 1968, 160,429-430.

MACK, A., & BACHANT, J. Perceived movementof the afterimage during eye movements.Perception & Psychophysics, 1969. 6,319-384.

MacKAY, D. M. Ways of looking at perception.In w. Wathen-Dunn (Ed.), Models for thepreception of speech and visual form.Cambridge: M.I.T. Press, 1961.

ROCK. I. The nature of perceptual adaptatzon.New York: Basic Books, 1966.

SHAFFER, 0., & WALLACH, H.Exten t-of-motion thresholds undersubject-relative and object-relative conditions.Perception & Psychophysics. 1966. I,441-451.

TEUBER. H. L. Perception. In J. Field (Ed.).Handbook of physiology-neurophysiology 11/.Baltimore: Williams & Wilkins, 1960.Pp. 1595-1668.

von HOLST. E. Relations between the centralnervous system and the peripheral organs.British Journal of Animal Behaviour, 1954, 2.89-94.

WALLACH, H., & FREY, K. Adaptation in theconstancy of visual direction measured by aone-trial method. Perception & Psychophysics,1910,5,249-252.

WALLACH. H., FREY. K., & RODNEY, G.Adaptation to field displacement during headmovement unrelated to the constancy of visualdirection. Perception & Psychophysics, 1969,5.253-256.

WALLACH, H.. & KRAVITZ, J. Themeasurement of the constancy of visualdirection and of its adaptation. PsychonomicScience, 1965. 2, 211-218.

Perception & Psychophysics, 1970, Vol. 8 (SA) 297

WALLACH. H•• & LEWIS, C. The effect ofab,nonnal displacement of the retinal imageduring .eye movements. Perception &Psychophysics. 1966. 1, 25-29.

WINER. B. 1. Statistical principles inexperimental design. New York: McGraw-Hili,1962.

YARBUS. A. L Eye movements and vision. NewYork: Plenum Press. 1967.

NOTES1. This research was done during an NIMH

postdoctoral fenowship at Stanford University. Iwould like to express my appreciation to

everyone who helped make this research possible.I am particularly grateful to Dr. Leo Ganz for hisgenerous help and to Robert Brown whodesigned and built some of the apparatus andassisted in the testing of Os. I would also like toexpress my gratitude to Hans Wallach. IrvinRock, and Leon Festinger for their carefulreading of an earlier version of this manuscript.

2. Address: New School for Social Research.New York, New York 10003.

3. All discussion of movement perception isrestricted to object movement. which is subjectrelative since. with movement that is objectrelative, the relative displacement of one object

in the visual Held with respect to another is apowerful indicator of movement.

4. We assume that the information about eyemovements that is matched against retinaldisplacement information is efferent rather thanafferent information (Brindley & Merton, 1960).

5. Data reported by Schaffer and Wallach(I966j indicates that Os are capable of reportingfar smaller subject-relative target displacements.Perhaps the low in tensity of the target in ourwork accounts for this difference in results.

(Accepted for publication January 28. /970.)

298 Perception & Psychophysics, 1970, Vol. 8 (SA)