JANETTE ATKINSONThe Psychological Laboratory, University of Cambridge, Cambridge, England

Visibility of an afterimage in the presence ofa second afterimage*

2

viewing, two black screens werepositioned so that the right eye sawonly one bar in the pattern and theleft eye saw only the second bar in thepattern. The S's eyes were horizontallyaligned with the fixation spot 28.5 cmfrom it. In the dichoptic viewingcondition, S triggered the flashgunhimself, so that his eyes wereaccurately lined up at the moment ofthe flash and the afterimage thatfollowed had the same relativepositioning properties as the stimuluspattern. Any S who had greatdifficulty in aligning the patternaccurately in the practice trials waseliminated from the experiments.Immediately after the flash, S closedhis eyes and kept them closedthroughout the trial time, thusmaintaining a constant backgroundluminance for viewing the afterimages.

The trial time was always 60 secimmediately following the flash. Salways had 15 min adaptation beforethe first trial and 5 min adaptationbetween trials. The S was asked not toblink, to keep his head still, and tomake as few voluntary eye movementsas possible.

Small silent two-way switches wereused by S to record visibility of the

1972). The afterimages were producedby a Braun electronic flashgun, withflash duration of approximately4 msec and light output of 160 J.Patterns were produced by cutouts in ablack mask which was placedimmediately in front of the flash.

For monocular viewing, S fixated aspot attached to the mask with hisright eye, his left eye being coveredthroughout the trial. For dichoptic

Visibility of an afterimage has been measured for a single rectangular bar onits own and in the presence of a second bar. Visibility of each bar is increased ifthe two bars are in the same orientation and closer than 1 deg (this effect beingcalled "facilitation") and is decreased if the bars are in different orientations(this effect being called "inhibition"). These interactions also depend on therelative size of the bars. If a pattern of two bars is viewed dichopticaUy, rivalrybetween the bars occurs causing a mutual decrease in visibility. The rivalry effectand the orientation-specific effects discussed above are additive in the dichopticviewing condition. The orientation-specific effects have been likened to thosefound in other psychophysical studies of angle detection.

The nonuniform fading of patternedafterimages was imt observed byHelmholtz (1866; see 1925 ed.).Quantitative studies of thefragmentation of afterimages havemore recently been made by Evans(1967), Evans and Wells (1967),MacKinnon, Forde, and Piggins(1969), and Atkinson (1970, 1972).The conclusion from all of thesestudies is that structuredfragmentation is a result of activity incontour-detecting units within thevisual system.

Pritchard, Heron, and Hebb (1960)studied the fragmentation of anoptically stabilized image, using acontact lens device, and found that afigure in one part of the stimulus fieldcould modify the visibility ofneighboring parts of the stimulus.However, they did not quantify thisGestalt type of "field effect."

The present experiment was carriedout to see whether one afterimagecould affect the visibility of anothernearby afterimage. The visibility of anafterimage was measured with andwithout a second afterimage. Only thevisibility of the entire image wastimed, no record was made offragmentary . appearances. Viewingconditions, orientation, size, andseparation of the two afterimages werevaried in different experiments.

Fig. 1. The four patterllll used in Experiment 1. x marks the fixation spot.Patterns 1 and 2 were aJao used in Experiment 2.

METHODApparatus and Procedure

The apparatus and procedure wereessentially the same as that describedin detail in a previous paper (Atkinson,

*Janette Atkinson was awarded theGeorge Henry Lewes studenlllbip from thePhysiolopcal LabQtatory. University ofCambridge. and. Ule Henry George PlimmerfeUowsbip from Ule Imperial College ofScience and ·rechnology. University ofLondon. to finance this research. TheauUlor would like to Ulank Dr. F. W.Campbell. Dr. L. A. RielS. and Dr. O. J.Braddick for their helpful criticlllms of thispaper.

3 4

Perception & Psychophysics, 1972, Vol. 12 (3) Copyright 1972, Psychonomic Society, Austin, Texas 257

1 2 3

Fig. 2. The three patterns used in Experiment 3. X marks the fixation spot.

Fig. 3. The four patterns used in Experiment 4. X marks the filiation spot.

afterimage. The S depressed a switchwhen a particular afterimage wasvisible and released the switch whenthe image disappeared. Visibility wasmeasured in seconds of trial time forwhich the afterimage was visible, asrecorded on counters connected to theswitches. No knowledge of results wasgiven during testing.

Orders of presentation were usedwhich balanced for practice effectsbetween different stimuli and viewingconditions.

SubjectsAll Ss were students at the

University of Cambridge. They werebetween 17 and 35 years of age andwere naive regarding the nature of theexperiments. All bad normal or

corrected-to-normal vision.

Experiment 1In this experiment and in

subsequent experiments involving thesame patterns, the term "a pattern oftwo vertical bars" will mean a patternin which the two bars are aligned sothat one bar is directly above the other(as in Fig. 1, Pattern 1). "A pattern oftwo horizontal bars" will mean twoparallel bars horizontally oriented, oneabove the other (as in Fig. 1,Pattern 2). It is obvious that these twopatterns do not have the sameconfigurational properties, nor do theycover the same area of visual field.

The two additional patterns used inExperiment 1 are shown in Fig. 1. Inall patterns, the separation of the bars

was constant at 1 deg of visual angle.The fixation point was always 2 degfrom .the bars, as indicated in Fig. l.

For all patterns, both monocularand dichoptic viewing conditions wereused. In the dichoptic condition, S'sright eye viewed the top bar only andthe left eye the bottom bar only.

For each of the patterns, twocomparisons were made: (1) Betweenthe visibility of the top bar whenviewed on its own (the bottom barbeing concealed for that trial) andthe visibility of the same bar whenboth bars were present. (2) Betweenthe visibility of the bottom bar whenviewed on its own and the visibility ofthe same bar when both bars werepresent.

Thus, the visibility of an afterimagewas compared with and without asecond afterimage. Single verticalimages were compared only with othervertical images at the same retinallocus, and horizontal images werecompared with horizontal images.

Eleven Ss recorded the totalvisibility for each of the bars in eachpattern. Each S had five trials on eachcondition, the f"mt being a practice.

Experiment 2Patterns 1 and 2 from Fig. 1 were

used again in Experiment 2. Thedistance separating the bars was varied,being 0.5, 1, 2, or 3 deg. Each bar wasviewed alone and with the second bar,making four experimental conditionsin all for each separation. Of course,separation is meaningless for single barpatterns, but each single bar patternacted as a control at the sameeccentricity 88 the bars of the relevanttwo-bar pattern.

Eleven Ss completed five trials oneach condition, the first trial in eachcondition being a practice. For eachpattern, the mean visibility for the topand bottom bars were calculated.

Experiment 3The patterns used are shown in

Fig. 2.

258 Perception & Psychophysics, 1972, Vol. 12 (3)

Fig. 4. Visibility (in seconds) for a horizontal bar alone and accompanied byanother horizontal bar or by a vertical bar (top graph). Visibility (in seconds) fora vertical bar alone and accompanied by another vertieal bar or by a horizontalbar (bottom graph).

Horizontal bar

DoMonocular

Dichopttc

-I

-I

significantly different on theNewman-Keuls and Tukey (a) tests(p < .01). This lowering of visibility

I

I-

-

II

-

I

-v.rtical bar

30

10

20

30

'lIGi!i!y(in secOnd!)

(In WOOds!

alone compared with a vertical bar'svisibility when accompanied by ahorizontal bar (see Fig. 4) was

RESULTSExperiment 1

A preliminary analysis was carriedout to determine whether therequirement to respond to twoseparate afterimages (in the same trialtime), rather than one, introduced anybias into the results which would maskany potential differences betweenpatterns. Trials given in Experiment 1,in which two bars were in the stimuluspattern, were split into two type•. InType I, the S bad only to record thevisibility of one bar, although a secondbar was present. In Type 2, thevisibilities of two afterimages wererecorded simultaneously by the S,uaing two switches (one in each hand).The results for Type 1 and Type 2were compared, using analysis ofvariance, and no significant differencewas found between the two [F(I,llO)= 0.3, p > .05]. This meant that therequirement to respond to twoafterimages rather than one was notaffecting the amount of visibilityrecorded. The results for the two typesof trial were therefore pooled forfurther analysis.

Analysis of variance on the resultsfor mean visibilities in Experiment 1showed that: (1) There was asignificant difference betweenmonocular and dichoptic viewingconditioJlSjviaibility was always higherfor monocular viewing [F(I,10) =70.3, p < .01]. (2) There was nosignificant difference between thetotal visibilities of vertical andhorizontal barli [F(I,10) 2.9,p> .05] for monocular or dichopticviewing. (3) Visibility was altered bythe 'presence· and . orientation ofanotheratteri.mage iii the stimuluspattth [F(S,$O)· 52.1, p < .01]. Themean visibility for each bar, with andwithout alleCODd bar, is shown inFig. 4 for each pattern.

Monoculor viewing. The differencebetween the visibility for a vertical bar

The total visibility was measured foreach bar with and without a secondbar. Three different separations wereused, these being 0.5, I, and 2 deg.Each of the five Ss had three trials oneach condition, the first being apractice.

Experiment 4In Experiments 1-3, all the bars

were 2 del in length; in Experiment 4,the relative sizes of the bars werevaried as shown in Fig. 3. Monocularand dichoptic viewing conditions wereused for each pattern. Fifteen Sscompleted three trials on eachcondition, the first being a practice.The visibility of each bar wasmeasured for the bar on its own andaccompanied by the second bar ineach pattern.

Perception. Psychophysics, 1972, Vol. 12 (3) 259

Table 1

One Horizontal Bar One Vertical Bar

Pattern 2 Pattern 3 Pattern 4 Pattern 1 Pattern 3 Pattern 4

MV F MV MV I MV F MV MV I

Mean Visibility ±23.7 - 1.0 19.0 + 5.0 19.3 + 5.0 25.1 - 1.0 17.8 + 5.0 18.3 + 5.0bUdbition/Facilitation

Total (in Sec) 22.7 24.0 24.3 24.1 22.8 23.3

Fig. 5. Facilitation (l.e., the amount of increase, in seconds, of visibility dueto the presence of a second bar in the same orientation) as a function ofseparation of the bars (in degrees). The dotted line shows the separation up towhich facilitation is a significant effect.

3

discussed above. The secondaltemative assumes that the two factors(orientation and rivalry) are additive.

If the first hypothesis is correct,then the interactive factor for ViewingCondition by Pattern would besignificant. This interactive factor wasnot significant at the .01 level, but wasat the .05 level. As the .01 level of phas been used for all experimentalresults in this study, this particularresult will be considerednonsignificant, and, as such, the resultis support for the hypothesis that theori en ta tion-specific effect istransferred in dichoptic viewing and issuperimposed on thenonorientation-speeific rivalry effect.

From the monocular results, theorientation-specific effect was foundto be approximately 5 sec ofinhibition for a pattern of two bars ata right angle to one another and 1 secof facilitation for two bars that are inthe same orientation. If these amountsfor facilitation and inhibition aresu btracted from the dichopticvisibility for each bar, the calculatedvisibility for each bar without theorientation-specific effects is much thesame for all patterns (approximately24 sec visibility per bar). Thesecalculations are shown in Table 1.From this calculation, a value ofapproximately 4 sec reduction invisibility is found to be due to rivalrywhich is not orientation specific.

Experiment 2The effect of one bar on the

visibility of another was calculated foreach bar at each separation bysubtracting the mean visibility of a baron its own from the mean visibility ofthe same bar in the presence of thesecond bar. Only patterns of twovertical or two horizontal bars wereused in this experiment. Facilitation,at each separation, is shown in Fig. 5.Facilitation occurred when either twovertical or two horizontal bars wereclose together, but not when theywere more than 1 deg apart. A smallbut statistically reliable facilitationwas found using analysis of variance[F(I,10) = 10.9, P < .01]. On furthercomparison of the means for variousseparations, using the Newman-Keulsand Tukey (a) tests, the facilitation wassignificant for only the two smallerseparations of the bars (0.5 and 1 deg).

\ 2- /

./SEPARAT 'ON

These monocular effects of"facilitation" and "inhibition" aredependent on relative orientation ofthe bars and can therefore be said tobe "orientation specific."

Dichoptic uiewing results. For all 'four patterns, visibility was lower fordichoptic than for monocular viewing,as shown in Fig. 4. For dichopticviewing, the drop in visibility causedby the presence of a second afterimagecould be due to either a type ofbinocular rivalry that is dependent onthe relative orientation of the bars, orto rivalry between bars (which is equalin amount for any pattern) plus themonocular orientation-specific effect

DEGREES OF

1

2u-e~

.nQ

ZoUw.n

~

o

...-c...

zo

by a differently oriented bar will becalled "inhibition." The term is merelyused as a description and is not meantto imply anything about underlyingmechanisms. The same effect wasfound for the visibility of a horizontalbar accompanied by a vertical bar.

Visibility was raised when a verticalbar was accompanied by a secondvertical bar, although this effect wassmall and did not reach significance onthe Newman-Keuls or Tukey (a) tests.The same was true for patterns of twohorizontal bars. This small increase invisibility produced by the presence ofa second parallel bar will be referred toas "facilitation."

260 Perception & Psychophysics, 1972, Vol. 12 (3)

Fig. 6. Inhibition (Le., the amount of decrease, in seconds, of visibility of abar, due to the presence of a second bar in a different orientation) as a functionof separation of the bars (in degrees).

Fig. 7. Orientationilpecific effects (facilitation and inhibition) for differentbars in the four patterns of Experiment 4.

bar

..

Inhibition ofsmall sa~e

r/

I/

//

/I

II

I

VD'ChOPt"

-

by another small square (Fig. 3,Pattern 4). However, this result couldbe explained merely in terms of thedifference in orientation that waspresent in Fig. 3, Pattern 2, but notpresent in Fig. 3, Pattern 4. In Fig. 3,Patterns 1 and 2, the orientation ofthe two bars was at 90 deg, but inPattern I, the bars were also the samesize. Inhibition of the horizontal barby the small square was smaller thaninhibition of the horizontal bar by thevertical bar of the same size. Thus, it isseen that the relative .size, as well asthe relative orientation, has anappreciable effect on the amount ofinhibition.

DISCUSSIONFacilitatory and inhibitory effects

of one afterimage OIl another havebeen demonstrated. These effectsdepend on the relative orientation,size, and spacing of the two images.For two identical images in the sameorientation, as close or closer than1 deg, there is a small facilitatoryeffect. For two bars in differentorientations, there is a large mutualinhibition (app r oximately 5 secinhibition). The results ofExperiment 3 showed a higher

-6

-8

-10

-12

Inhibition(in seconds)

r

.-- vertical

ohor;zontol

_·_··.·.5loping 145·)

/'I, /

\\ /,.\' / /\' i I

\~'//

0 .......-----........-

-2

-8

-4

·6

Inhibition Ofhorizontal bar

Facilitation and Inhibition- - -I • - .,<D ~ Po 0

//

-1//

/ -2/

//

_..0' _4----o-:

(In degrees)

facilitation depended upon the relativesize of the bars in each pattern. Thesmall square, as seen in Fig. 3,Pattern 2, was inhibited by thehorizontal bar but was not inhibited

"barsSeparation of

2',.

-2

V,

-4

-6

-6 /

Inhibition

0 .......----.....

-8

-4

-6

Inhltlltlon-2

(in seesJ

Experiment 3Ai; in the previous two experiments,

the visibility was calculated for eachbar with and without a second bar; thedifference between these twovisibilities is shown in Fig. 6. Theamount of inhibition varied with theseparation of the bars, being highestwhen the bars were 1 deg apart[F(2,8) = 63.1, p < .01]. This wastrue for both the right angle andoblique bar patterns. The extent ofinhibition did not vary significantlyfrom pattern to pattern, i.e., there wasno greater inhibition between bars atright angles to one another thanbetween bars at 45 deg to one another[F(5,20)=0.9,p> .05].

Experiment 4The inhibition and facilitation were

calculated for each pattern (see Fig. 3for patterns used) and are plotted inFig. 7. From analysis of variance,significant differences were found(1) between monocular and dichopticviewing conditions [F(1,14) = 38.1,P < .01] and (2) between visibilities ofa bar with and without a second bar[F(6,84) = 11.1, P < .01].

For dichoptic viewing, all visibilitieswere found to be shiftedapproximately 4 sec in the inhibitorydirection, although the absoluteamount of visibility varied frompattern to pattern. From the analysisof variance, there was no significantinteraction between viewingconditions and the effects [F(6,84) =2.9, p> .05). This result confirms theresult of Experiment 1 and shows thatthe two factors (rivalry andorientation-specific effect) are separateand additive.

The amount of inhibition or

Perception &.Psychophysics, 1972, Vol. 12 (3) 261

inhibition at 1-deg separation than at0.5- or 2-deg separations.

These results can be accounted forif it is postulated that for any two-barpattern there is a small facilitation ofvisibility if the bars are close together,and superimposed on this effect thereis an orientation-specific inhibitionwhich operates only if the bars are in'different orientations. Both effectsdecrease in magnitude as the spacingof the bars is increased, the facilitatoryeffect declining at a greater rate thanthe inhibitory effect. Now, at smallseparations (0.5 deg), both facilitatoryand inhibitory effects will be operatingin opposition to one another (for anypattern where the bars are in differentorientations); at 1-deg separation, botheffects have declined but are stilloperating and the facilitatory effect ismuch weaker than it was at 0.5-degseparation, so that the final value forinhibition appears higher at 1 than at0.5 deg. At a separation of 2 deg, thereis no facilitation but the inhibitoryfactor is still operating; because of theincrease in separation, the inhibition isweaker at 2- than at 1-deg separation.

A previous study by Atkinson(1970) showed that the unitaryvisibility of two bars at a right anglewith no gap between their ends wasthe same as the unitary visibility oftwo bars in the same orientationjoined end to end (contour lengthbeing equal for both patterns). It isobvious, therefore, that theo ri e n tati on -s p e e i fic inhi bi ti ondemonstrated in the aboveexperiments does not operate unlessthere is a space between the ends ofthe bars. The size of this space,necessary to cause inhibition, may berelated to the distance over whichperceptual "closure" of the space ispossible. Several informal observationsmade on two experienced Ss revealedthat there was a definite minimum sizeof gap at which the bars inhibited oneanother, and if the gap was smallerthan this critical size, "closure" wasoften observed. The minimum size ofgap was approximately 0.25 deg. Itseems likely that the minimum gapsize relates in some way to the size ofoverlap of receptive fields of the unitsin the visual system responsible for theperceptual phenomena, but it wouldseem presumptuous to try to specifythe exact mechanism any further atthis point.

Blakemore, Carpenter, andGeorgeson (1970) and Bouma and

Andriessen (1970) have reportedstudies in which a change in perceivedorientation of a line has been broughtabout by a second line at an angle tothe first. This orientation-specificeffect might be similar to the effectreported in the present study onafterimages ; the orientation-sensitiveunits that are operating in both theseexperimental situations laterallyinhibit one another if they are tunedto different, but proximal,orientations. There is as yet noelectrophysiological evidence toprovide a basis for such inhibitionbetween cortical units.

The inhibitory effects found fordichoptic viewing have been explainedin terms of a combination of themonocular orientation-specific effects(reported in the preceding section) anda rivalry effect which is independentof relative orientation. Binocularrivalry between two opticallystabilized images has been reported byKrauskopf and Riggs (1959) andDitchburn and Pritchard (1960).Kaufman (1963) and Levelt (1965),using nonstabilized images, also foundthat rivalry depended on the relativepositioning of the two images.However, none of these studiescompared dichoptic results withmonocular results in the same way asin the present afterimage study, and soit is impossible to compare thepossible effect of orientationdifferences across studies.

Cohen (1961), using a stabilizedvertical line as stimulus to one eye,found that a similar nonstabilized lineto the other eye increased the visibilityof the stabilized line. His results arecontrary to what would be expectedfrom the present result on afterimages,which would predict a decrease invisibility on presentation of any othercontour as a nearby stimulus to theother eye.

In conclusion, the results ofExperiments 1-4 show that afterimagesmutually interact, causing changes intheir total visibilities. The nature ofthese interactions depends on therelative orientation, size, and spacingof the two images and also on theviewing condition used. A type ofbinocular rivalry which is constant forall patterns is superimposed uponorientation-specific effects, producinglower visibilities in the dichopticviewing condition than in themonocular.

The orientation-specific effect for

afterimages has been likened to thatfound in other psychophysical studieson angle estimation. Thus, the notionof inhibition betweencontour-detecting units, which hasbeen put forward to account for thelatter results, may also be applicable tothe interactions of afterimages.

REFERENCESATKINSON, J. A study of perceptual

analysis using stabilized images.Unpublished doctoral dissertation,University of Cambridge, 1970.

ATKINSON, J. The effect of size, retinallocus, and orientation on the visibility ofa single afterimage. Perception &Psychophysics, 1972, 12 (2B). 213-217.

BLAKEMORE, Coo CARPENTER. R. H. S.,& GEORGESON, M. A. Lateral inhibitionbetween orientation detectors in thehuman visual system. Nature, 1970. 228.37-39.

BOUMA, H., & ANDRIESSEN, J. J.I n d uc ed changes in the perceivedorientation of line segments. VisionResearch, 1970. 10, 333-349.

COHEN, H. B. The effect of contralateralstimulation on visibility with stabilizedretinal images. Canadian Journal ofPsychology, 1961. 15. 212-219.

DITCH BURN, R. M., & PRITCHARD, R.W. Binocular vision with two stabilizedretinal images. Quarterly Journal ofExperimental Psychology, 1960, 12,26-65.

EVANS, C. R. Further studies of patternperception and a stabilized retinal image:The use of prolonged afterimages toac hieve perfect stabilization. BritishJournal of Psychology, 1967, 58,315-327.

EV ANS, C. R., & WELLS, A. M.Fragmentation phenomena associatedwi th binocular stabilization. BritishJournal of Physiological Optics. 1967, 24,45-50.

HELMHOLTZ, H. von. Handb ucb derphYlliologischen Optik. Vol. II. (Trans. J.C. Southall). Optical Society of America,1925. P. 231.

KAUFMAN, L. On the spread ofsuppression and binocular rivalry. VisionResearch. 1963, 3, 401-415.

KRAUSKOPF, J., & RIGGS, L. A.Interocular transfer in the disappearanceof stabilized images. American Journal ofPsychology. 1959, 72. 248-252.

LEVELT, W. J. M. On binocular riualrv:Soesterberg, The Netherlands: NationalDefence Research Organization, 1965.

MacKINNON, G. E., FORDE. J., &PI GGINS, D. J. Stabilized images.steadily fixated figures and prolongedafterimages. Canadian Journal ofPsychology, 1969, 23, 184-195.

DITCHBURN. R. W.• & PRITCHARD, R.M. Binocular vision with two stabilizedretinal images. Quarterly Journal ofExperimental Psychology, 1960. 12.26-32.

PRITCHARD, R. M., HERON. W .• & HEBB,D. O. Visual perception approached bythe method of stabilized images.Canadian Journal of Psychology, 1960,14.67·77.

(Accepted for publication June 9,1972.)

262 Perception & Psychophysics, 1972, Vol. 12 (3)