JOHN HOGBENUniversity of Western Australia, Nedlands, Western Australia 6009

Interaction of forward and backwardvisual masking*

An investigation was conducted into the interaction of the forward and backwardmasking effects of unpatterned visual stimuli. It was found that detection of a test spotwas easier under conditions that should have provided both forward and backwardmasking than under either forward masking or backward masking alone. The implicationsfor an integration theory of masking are discussed, and the findings are contrasted withfindings on the interaction of forward and backward masking by dynamic visual noise.

Recently a number of studies haveappeared (e.g., Uttal, 1969a, b), employinga new technique-dynamic visual noise(DVN)-as a visual masking stimulus. DVNis produced by rapidly plotting randomdots sequentially on the face of a displayoscilloscope. The test stimulus is analphabetic character composed of dots ofsimilar size and intensity. Uttal claims thatthe importance of this technique is that itallows the isolation of two classes ofmechanisms that are confounded inexperiments with other types of maskingstimuli. One class of mechanisms operatesat the retinal level and concerns the inertiaof the transduction processes and lateralsummation or inhibition. The other classoperates at some higher level of the visualsystem and involves perceptual confusionsof pattern. At practical levels of DVN agiven retinal location is unlikely to bestimulated both by a signal dot and by anoise dot, and, hence, retinal mechanismsof masking will be minimized. Thiscontention is supported by an experiment(Uttal, 1970) which demonstrated strongmasking with dichoptic presentation of thetest stimulus and DVN. On the other hand,it has been shown (e.g., Battersby &Wagman, 1962; Schiller, 1969) that thedichoptic masking effects of anun patterned stimulus are very slight unlessthe contours of the test and maskingstimuli are adjacent. It is, therefore, likelythat different mechanisms underlie theeffects of DVN and the effects of anunpatterned mask that is substantiallylarger than the test stimulus.

In a study of the interaction of forwardand backward masking, Uttal (1969b)presented the test stimulus during aninterval between leading and trailing burstsof DVN. He found that forward andbackward masking by DVN combined toproduce a greater masking effect thaneither the forward-masking or thebackward-masking condition alone.

'Supported by A.R.G.C. Grant ARG17-226 toJ. Ross and V. Di Lalla. Grateful thanks are dueto Dr. V. Di Lalla for valuable criticisms ofearlier drafts of this paper.

Furthermore, the magnitude of thecombined effect was far greater than wouldhave been predicted by a simple additivemodel.

In contrast to this, a study of the effectof the intensity of a preadapting field on abackward-masking function (Hogben,1968), using a conventional Dodge-typetachistoscope, showed that forward- andbackward-masking effects did not combineto produce a total greater masking effect.Rather, in some complex manner, anadapting field of relatively high intensityappeared to inhibit the effect of abackward-masking stimulus. In that study,the task of the S was to identify variousgrid patterns presented in a small circularpatch masked by much larger overlappinghomogeneous white fields. The purpose ofthe present study was to develop theimplications of those findings with respeclto the detection of a spot of light and toex amine directly the interaction offorward-masking and backward-maskingeffects. The sequence of presentation ofstimuli was similar to the sequence inUttal's (1969b) experiment.

METHODSubjects

Ten first-year psychology students withuncorrected vision each attended oneexperimental session, lasting approximatelyI h. Half of the Ss were allocated randomlyto each of two experimental conditions.

ApparatusStimuli were presented in a Scientific

Prototype Model GB tachistoscope. A10-mL adapting field subtended 7 deghorizontally by 5 deg vertically. The teststimulus, which appeared in the center ofthe adapting field, was a 5-mL white disksubtending a visual angle of I deg on ablack background. High figure-groundcontrast was achieved by means of backlighting. A hand switch was provided for Sto trigger each stimulus presentation.Procedure

In Condition LDFL(light-dark-flash-light), the sequence of

events on a single stimulus presentationwas as follows. The adapting field wasswitched off and a dark interval of variableduration preceded the test stimulus. Theadapting field returned on the terminationof the test stimulus. For Condition LFDL,the order of occurrence of the dark intervaland the test stimulus was reversed. Tendurations of the dark interval were used ineach condition: from 0 to 200 msec in25-msec steps, and 500 msec.

Each experimental session began with10 min of practice, which was followed bya determination of test stimulus durationthreshold for each dark interval in randomorder. The duration threshold wasdetermined by means of a temporalforced-choice tracking procedure. Eachtrial consisted of two presentations, ononly one of which the disk wasilluminated. There were 2-3 sec betweenpresentations, and S was required tonominate whether the disk occurred duringthe first or the second. Duration wasinitially set at 1.0 msec and increased insteps of 1.0 msec until three successivecorrect responses were recorded. It wasthen reduced to the level at which the lasterror had occurred, and increased in stepsof 0.2 msec, again to a criterion of threesuccessive correct responses. If no erroroccurred with this smaller step size,duration was again decreased by 1.0 msecand an ascending series begun with thesmaller step size.

RESULTSFigure I shows the duration threshold of

the test flash as a function of duration ofthe dark interval for the two conditions.Analysis of variance showed a significanteffect of dark interval (F =9.35, df =9,72;P < 0.01) with the effects of conditions(F = 1.22, df = 1,8) and the interaction(F =1.26, df =9,72) not significant.

As is clear from the graph, the majorpart of the main effect variance iscontributed by the increase in thresholdbetween the intervals of 0 and 25 msec,but inspection of the graph suggests alikely quadratic component (inverted Ushape). Accordingly, a trend analysis wasperformed on the first nine dark intervals.The linear (F =35.37, df =1,64) andquadratic (F = 28.90, df= 1,64)components of the main effect variancewere highly significant, as expected. Inaddition, cubic components of the maineffect (F = 8.07, df = 1,64; P < 0.0 I) andof the Conditions by Dark Intervalinteraction (F = 5.47, df = 1,64; P < 0.0 I)were detected. These cubic componentswere clearly contributed by the LFDLcondition, which shows a reliable decreasein threshold at the interval of 125 msec. Incontrast, the LDFL condition shows nosuch dip in the curve.

Perception & Psychophysics, 1971, Vol. 9 (6) Copyright 1971, Psychonomic Journals, Inc., Austin, Texas 487

OMit INTEIIlVAL (_)

Fig. 1. Test stimulus duration threshold as a function of dark interval duration for thetwo conditions of the experiment.

processes, it has been found that theaddition of another masking stimulusactually facilitates performance ascompared to a case of forward maskingalone (Condition LFDL) or a case ofbackward masking alone (LDFL). Theunderlying processes are evidently far morecomplex than might be suggested by ahypothesis of straightforward temporalintegration of luminance. The disruptiveeffect of the insertion of a dark intervalcompares with the results of Kahneman(J 966) who found that the perception ofan acuity target was much more difficult ifan adapting field was turned off during teststimulus presentation than when the teststimulus was superimposed on a steadyadapting field. These results provideevidence for the importance of darkintervals in masking phenomena but leaveunresolved the question of what features ofthe dark interval may be important. Muchof the work done in connection with theearly stages of light and dark adaptation(e.g., Baker, 1963; Boynton & Miller,1963) demonstrates that any change, eitherupward or downward, in the luminance ofan adapting field may produce an elevationof the threshold for detection of a target.

REFERENCESBAKER, H. D. Initial stages of dark and light

adaptation. Journal of the Optical Society ofAmerica, 1963,53,98-103.

BATTERSBY, W. S., & WAGMAN, I. H. Neurallimitations of visual excitability: IV. Spatialdeterminants of retrochiasmal interaction.American Journal of Physiology, 1962, 203,359-365.

BOYNTON, R. M., & MILLER, N. D. Visualperformance under conditions of transientadaptation. Illuminating Engineering, 1963,58,541-550.

ERIKSEN, C. W. Temporal luminancesummation effects in backward and forwardmasking. Perception & Psychophysics, 1966, I,87-92.

HOGBEN, J. H. Effect of an adapting field on abackward visual masking function.Unpublished thesis, University of WesternAustralia, 1968,

KAHNEMAN, D. Time-intensity reciprocityunder various conditions of adaptation andbackward masking. Journal of ExperimentalPsychology, 1966,71,543-549.

SCHILLER, P. H. Behavioral andelectrophysiological studies of visual masking.InK. N. Leibovic (Ed.), Informationprocessing in the ne/volls system. New York:Springer-Verlag, 1969.

UTTAL, W. R. Masking of alphabetic characterrecognition by dynamic visual noise (DVN),Perception & Psychophysics, 1969a, 6,121·128.

UTT AL, W. R. The character in the holeexperiment: Interaction of forward andbackward masking of alphabetic characterrecognition by dynamic visual noise (DVN).Perception & Psychophysics, 1969b, 7.177·181.

UTTAL, W, R. On the physiological basis ofmasking with dotted visual noise. Perception &Psychophysics, 1970,7,321-327.

(Accepted for publication October /8. 1970.)

200 500150

terminated. If summation were to occur, atrailing field that immediately followed thetest stimulus should increase the thresholdconsiderably, whereas the threshold for theO-msec condition is actually lower. Thisresult is not in accordance with theIum in anee -summation/contrast-reductiontheory of visual masking (Eriksen, 1966).This theory proposes that masking isbrought about by reduction of contrast inthe test stimulus, due to temporalintegration of luminance in overlappingretinal areas. Masking should increase as adirect result of the temporal proximity ofspatially overlapping areas of luminance,whereas the present study shows decreasedmasking under just these conditions.

The present results contrast with theresults of Uttal (1969b), who found thatleading and trailing bursts of DVNcombined in their masking effects upon atest stimulus; the results of this study showclearly that leading and trailingunpatterned fields do not combine in theirmasking effects. Indeed, the case in whichthe test stimulus was preceded andfollowed immediately by the unpatternedfield was by far the easiest of allconditions. It is probable that thecontrasting results of these two studiesreflect fundamental differences in theprocesses of masking by pattern andmasking by light.

Quite a new problem is raised by theresults of this experiment: Far fromdemonstrating the summation of masking

10050o

'U' '"'\ 0•• 3 \• 6-......\ I

0

~\ I 6

/\ /1/1

"III 'tIt II-

ZI

0 I~ 2u I ~ LDFLIIIl- I LI'DLIII -- .0

II

I.

DISCUSSIONThere are two prominent aspects to the

results: (I) the overall similarity of thethreshold curves for the two conditions,especially the large increase in thresholdbetween 0 and 25 msec (Fig. I); (2) thedifference in level of threshold at about125 msec. Up to 100 msec, it makes littleor no difference whether the dark intervalis presented before or after the teststimulus. At greater intervals, a large andsignificant difference is evident in theshape of the threshold curves. Theou tstanding difference is the largedecrement in threshold at 125 msec for theLFDL condition. The group curve isrepresentative of the individual results offour of the Ss; the fifth S did not show thesharp threshold decrement at 125 msec.The reason for this set of results isunknown.

For both conditions, the main incrementin threshold occurs between 0 and25 msec. The threshold is much lowerwhen there is no dark interval than whenthere is a dark interval of any durationfrom 25 to 500 msec. This indicates thatthe leading and trailing fields do not simplysummate in their effects on the teststimulus. This point is most clearly seen incomparing the O-msec and 500-msecintervals in Condition LFDL. The500-msec interval may be regarded as acase of pure forward masking, since theadapting field does not return until500 msec after the test stimulus is

488 Perception & Psychophysics, 1971, Vol. 9 (6)