@Pergamon

VisionRes., Vol.36,No.17,pp.2721–2727,1996CopyrightCl1996ElsevierScienceLtd.All rightsreserved

PII: S0042-6989(%)00004-1 Printedin GreatBritain0042-6989/96$15.00+ 0.00

Temporal and Spatial Frequency Tuning of theFlicker Motion AftereffectPETER J. BEX,*~$ FRANS A. J. VERSTRATEN,$ ISABELLE MARESCHAL*

Received 18 September1995; in revisedform 10 December 1995

The motion aftereffect (MAE) was used to study the temporal and spatial frequency selectivity ofthe visual system at supra-threshold contrasts. Observers adapted to drifting sine-wave gratings ofa range of spatial and temporal frequencies. The magnitude of the MAE induced by the adaptationwas measured with counterphasing test gratings of a variety of spatial and temporal frequencies.Independently of the spatial or temporal frequency of the adapting grating, the largest MAE wasfound with slowly counterphasing test gratings (at approximately 0.125-0.25 Hz). The largestMAEs were also found when the test grating was of similar spatial frequency to that of the adaptinggrating, even at very low spatial frequencies (0.125 c/deg). These data suggest that MAEs aredominated by a single, low-pass temporal frequency mechanism and by a series of band-pass spatialfrequency mechanisms. The band-pass spatial frequency tuning even at low spatial frequenciessuggests that the “lowest adaptable channel” concept [Cameron et al. (1992). Vision Research, 32,561-568) may be an artifact of disadvantaged low spatial frequencies using static test patterns.Copyright @ 1996 Elsevier Science Ltd.

Adaptation Motionaftereffects Temporaltuning Spatialtuning

INTRODUCTION

It has been suggested that the early stages of humanvisual processing involve analysis by a parallel set of“channels” that may be defined in terms of their spatialand temporal tuning characteristics [see Graham (1989)for an overview]. Spatial frequency selectivity is wellestablished and has been demonstratedboth physiologi-cally (e.g. Enroth-Cugell & Robson, 1966) and psycho-physically (e.g. Campbell & Robson, 1968). The resultsindicate that spatialprocessing involvesa range of band-pass spatial frequency-selectivefilters each with a bandwidth of approximately one octave (e.g., Maffei &Fiorentini, 1973). Temporal frequency selectivity hasreceived less attention, but existing data suggest thatthere may be two or three temporal channels, one low-pass and one or two band-pass filters (Mandler &Makous, 1984;Hess& Plant, 1985;see also Fredericksen& Hess, 1996a,b).Physiologicaldata from the macaque(Foster et al., 1985) show that neurons in VI and V2 are

*McGill Vision Research, Department of Ophthalmology, McGillUniversity, 687 Pine Avenue West (H4 14), Montreal, Quebec,H3A IA1, Canada.

~Present address: Center for Visual Science, Universityof Rochester,274 Meliora Hall, Rochester, NY 14627,U.S.A.

~Vision Sciences Laboratory, Department of Psychology, HarvardUniversity, 33 Kirldand St, Cambridge,MA 02138,U.S.A.

$To whom all em-respondenceshouldbe addressed at present address[Fax +1 716271 3043;EmaiZbex@cvs.rochester.edu].

2721

broadly tuned for temporal frequency and may be eitherlow-pass or band-pass.

Motion aftereffects and channel theories

One interesting demonstration of the existence ofspatial frequency tuned mechanismsin the human visualsystem has been established using a well documentedvisual illusion known as the Waterfall Illusion or themotionaftereffect [MAE:see Wade (1994)for a selectiveoverview]. When a stationary image is examined afterprolonged viewing of a moving image, the stationaryimage appearsto move in the oppositedirectionto that ofthe inducing image: the MAE. MAEs are particularlyinteresting because they can be used for selectiveadaptation [see Sekuler & Pantle (1967) for the rationaleof this procedure].

Spatial frequency tuning. The contribution of spatialfrequency selectivemechanismsto motion detection hasbeen establishedby comparing MAE characteristics forgratingsof differingspatialfrequencies.In thisparadigm,observersview a driftinggrating(the adaptingpattern)ofa given spatialfrequency,then the magnitudeof the MAEis measured for static gratings (the test pattern) ofdiffering spatial frequencies. Using this procedure,several researchers have demonstrated that the strongestMAE is elicitedwhen the test and adaptinggratingsare ofsimilar spatial frequency (Over et al., 1973;Cameron etal., 1992). However, Cameron et al. showed that whenthe adapting grating is lower in spatial frequency thanabout0.5 cldeg, this relationshipbreaks down and a peak

.—

2722 P. J. BEX et al.

MAE is always found at around 0.5 cldeg. This relation-ship suggeststhat motion detection involvesmechanismsnarrowly tuned for spatial frequency (at least for spatialfrequencies above 0.5 c/deg).

Temporal frequency tuning. Support for the existenceof temporal frequency-selective mechanisms usingMAEs is more scarce. Most research has examined thetemporal parameters which result in the largest MAEs(e.g., Pantle, 1974; Wright & Johnston, 1985). Specifi-cally, Pantle (1974) studied whether a constant velocityor temporal frequency resulted in maximal MAEs for arange of spatial frequencies. The optimal parameterswere a productof spatialfrequencyand adaptingvelocity(i.e., temporal frequency), not velocity per se. Noexperiments have been reported which describe therelationshipbetween the temporal frequency of adaptingpatterns and the temporal modulation frequency of thetest pattern.This relationshipis exploredin experiment1.

In the research described above, the magnitude ofMAEs was measured using static test patterns. However,a recent distinction has emerged between MAEs whichare measured using static test patterns and thosemeasured with dynamic “flickering” test patterns (e.g.,Hiris & Blake, 1992).For example, in general no MAE isinduced by non-Fourier motion stimuli if tested withstatic patterns (Anstis, 1980; Derrington & Badcock,1985; Nishida et al., 1994), but a flickering test patternreliably reveals a MAE (McCarthy, 1993; Ledgeway,1994; Nishida et al., 1994). These differences in thepsychophysicaldata have led to the speculativesugges-tion that the two different types of MAE originate atdifferentsites along the path of visualmotionprocessing.Nishida & Sato (1995) suggested V1 as a possiblecandidatefor the staticMAE and area MT or MST for theflicker MAE (see also Ashida & Osaka, 1995).

Using flickeringtest patterns, Ashida & Osaka (1995)found that the optimal temporal frequencyfor inducingaMAE is found to be partiallyvelocity tuned,not temporalfrequency tuned, as is the case for static MAE. Theseresults support the proposal of two different sites ofMAE. Furthermore, Ashida & Osaka (1994) confirmedthat the magnitudeof static MAEswas greatest if test andmatch gratingswere of similar spatial frequency,but thisrelationship was not found if the test grating wasflickering. In this case no spatial frequency selectivitywas observed. This finding appears to contradict aprevious finding in which spatial frequency tuning wasshown using flicker MAE (von Gri.inau& Dub6, 1992).The different results were attributed to experimentaldifferences (Ashida & Osaka, 1994).

In the present experiments,we were interested in thespatial and temporal frequency tuning of flickerMAE. Inthe same way that spatial frequency tuning of the (static)MAE reveals spatial frequency-selectivemotion detec-tion mechanisms, we hypothesized that any temporalfrequency selectivity of motion detection mechanismswould be exhibited by temporal frequency tuning of theflickering MAE. We measured the magnitude of MAEselicited after adaptation to drifting sine gratings whose

Adaptation Period:(drift towards centre)

Test Period:~ (counterphase) ~

FIGURE1.Schematicdiagramillustratingthe geometryof the display.Sinusoidalgratingwere presentedin twohorizontalregions(7.5by 7.5deg), separated by a horizontal strip of 1 deg width with a centralfixation point. The sine gratings were all 50% peak contrast, theremainderof the displaywas at mean luminance(32 cd/m2).Duringa20 sec adaptationperiod, the sine gratings drifted towards the fixationpoint to aid steady fixation. During the test phase, the gratings weresinusoidallycounterphasedin each windowuntil the observer reported

the end of the MAE.

spatial and temporal frequencies were manipulated.Following Ashida & Osaka (1994), the magnitude ofthe MAE was estimated by recording the duration of theMAE. The test grating in each case was a sine grating ofthe same spatial frequency and peak contrast as theadapting grating, but whose contrast was counterphasedsinusoidally at between 0.125 to 16 Hz. A comparisonconditionwas recorded using stationarygratings (OHz).Using this protocol, we found no evidence for narrowtemporal frequency tuning of the flickering MAE. Forany spatial and temporal adapting frequencies, the peakflickerMAE was found at low counterphasefrequencies.

In a second experiment, we measured the spatialfrequency tuning of flicker MAE for low spatialfrequency adaptinggratings (0.125–2 c/deg). The resultsshowedclear spatialfrequencytuningat all spatialscales,in good agreementwith von Griinau& Dub6 (1992) andsuggest that the absence of such tuning reported byAshida & Osaka (1994) may be related to their stimulusparameters. The spatial frequency tuning at low spatialfrequencies shows that the lowest adaptable channel(Cameron et al., 1992) revealed using static MAE doesnot exist using flicker MAE.

EXPERIMENT1: TEMPORALFREQUENCYTUNINGOF FLICKERMAE

Methods

Apparatus and stimuli. Stimuliwere generated using aVSG 2/1 graphicscard (CambridgeResearch Systems)ina host PC microcomputer (DELL 333D) and werepresented on a Nanao Flexscan 6500 monitor with P4

TEMPORALAND SPATIALFREQUENCYTUNINGOF THE FLICKERMOTIONAFTEREFFECT 2723

r

\‘B: 4 e/deg0 .125.25 .5 1 2 4 8

r & A 16 Hz0 8 I+z~\

❑ 4 HzA 2 Hz● 1 Hz

7\N ■ o.5”-iiz.

\N

Countetphaae Frequeney (Hz)

FIGURE2. Magnitudeof MAEas a functionof the temporalfrequencyof the counterphasingtest grating. The data for the two observers areshown in separate columns and the data for the three adaptingspatialfrequencies (as well as test spatial frequencies) are shown in separaterows. The temporal frequency of the adaptinggrating is shown in thecaption and the temporalfrequencyof the adaptinggrating is showninthe legend. The spatial frequency of the test grating was the same asthat of the adaptinggrating in each case. The temporalfrequencyof thetest grating is shownon thex-axis with semi-logcoordinates(to permitthe inclusionof the OHz data, where the test grating was static). Theduration of the MAE is shown on the y-axis. Each data point is the

mean of at least four observations.Error bars show f S.E.

phosphor and with a frame rate of 118 Hz. The meanluminanceof the displaywas 32 cd/m2.The luminanceofthe display was linearized using an ISR attenuator(Pelli& Zhang, 1991)and calibratedusing a UDT Photometer.The image was 16 deg horizontally(512 pixels) by 13.4deg vertically (428 pixels) and was viewed from adistance of 118 cm. Subjects viewed the screenbinocularly in a dim room. The spatial layout of thedisplay is shown schematicallyin Fig. 1. There were twosquare windows on the screen, each subtending7.5 degby 7.5 deg. The windows were separated horizontallybya 1 deg strip of mean luminance, in the center of whichwas a prominent fixation point. The remainder of thedisplay was blank and at the mean luminance.

Adapting and test stimuli were vertical sinusoidalgratings of 50% Michelson contrast, which werepresented in the square windows. The adapting gratingsdrifted towards the fixationpoint. The test gratingsweresinusoidally counterphase flickering. The spatial fre-

quencyof the adaptinggratingswas either 1,2 or 4 cldeg.The temporal frequency of the adapting gratings wasvaried between 0.125 and 16 Hz, in steps of one octave.The spatialfrequencyof the test patternwas also 1,2 or 4c/deg and in experiment 1 was always the same spatialfrequency as the adaptinggrating which had preceded it.The test gratings were counterphased at a temporalfrequency between 0.125 and 16 Hz, in steps of oneoctave. An additional conditionwas measured in whichthe test gratingwas static (OHz counterphasefrequency).The startingphase of all gratingswas randomizedbeforeeach presentation.

Procedure

The subjectwas instructed to maintain steady fixationduring adaptation and testing and initiated each trial bypressing a mouse button. This was followed by a 20 secadaptationperiod duringwhich the adaptingsine gratingwas presented.The adaptinggrating was always driftingtowards the center of the screen to facilitate steadyfixation. The adaptation period was immediately fol-lowed by a brief tone and the test period. During the testperiod, the counterphasingtest grating was presented inboth windows.The subjectwas requiredto press a mousebutton when the MAE had finished. If the subjectexperiencedno MAE, the durationwas recorded as zerosec. Subjectspracticed the task many timesbefore formaldata collection. The direction of the MAE was alwaysseen in the opposite direction to that of the adaptinggrating (in this case it always appeared to move awayfrom the fixationpoint)and it was not necessaryto recordthe perceived direction of MAE. Several studies (e.g.Georgeson & Harris, 1978) have found that counter-phasing gratings viewed parafoveally appear to driftaway from the center even withoutadaptation:the foveo-fugal drift effect (FFDE). However, the FFDE wouldresult in motion away from fixation which neverterminated whereas the MAE measured in the presentstudy reliably halted. Furthermore, on some trials,observers reported that they experienced no MAE evenafter adaptation, again inconsistentwith the intrusion ofFFDE into our results.Subjectshad normal or eorrected-to-normalvision.

Each trial was followed by a inter-trial recoveryintervalof not less than 1 min. The whole procedurewasrepeated for each of the combinations of spatial andtemporal frequencies measured. The presentation se-quence for the various spatial and temporal frequencieswas randomizedand the datawere collectedover a periodof severalweeks. The mean and standard deviation of atleast four estimatesof MAE duration for each conditionwere recorded.

Results

Estimates of the MAE duration as a function of testtemporal frequency are shown for the two observers inFig. 2. The top panels represent the results where theadaptingand test spatialfrequencieswere 1c/deg.For themiddlepanels the spatialfrequencieswere 2 c/deg and in

2724 P. J. BEX et al.

201 T

0 PB: 2 dsglsec

O 2 tidsq❑ 1 c/d~A 0.5 e/decI● 0.25 c/dw

%

■ 0.12s Clde

IM : 1Hz

.1 1 10.1 1Test spatial Fraquency (c/deg)

FIGURE3. Magnitudeof MAE as a function of the spatial frequencyof the counterphasingtest grating. The data for the two observers areshown in separate columns and the data for the separate temporalfrequencies are shown in separate rows. The spatial frequency of theadaptinggrating is shown in the caption. In the top row, the temporalfrequency of the adapting grating was 1 Hz (its speed varied), in thebottom row, the speed of the adapting grating was 2 degjsec (itstemporal frequency varied). The spatial frequency of the adaptinggrating is shownon thex-axiswith semi-logcoordinates.The temporalfrequencyof the test gratingwas 0.25Hz in each case. The durationofthe MAEis shownon they-axis. Each data point is the mean of at least

four observations.Error bars show ~ 1 S.E.

the bottompanels theywere 4 c/deg. In all cases, it can beseen that the longest MAE was found when the testgrating was counterphasingat a low temporal frequency.Three combinations of adapting spatial and temporalfrequency resulted in no or negligible MAE (1 c/degadapting temporal frequency 16 Hz, 4 c/deg adaptingtemporal frequency 1 or 0.5 Hz).

The results of the temporal frequency tuning of theflicker MAE clearly show that the most pronouncedMAEs are found when the test grating is slowlycounterphasing.This effect is found for both observersand for all conditions measured. The effects of theadapting frequency are not as clear. The data show weakband-pass tuning of the MAE, but there is no clearsupport that the MAE is tuned to either the temporalfrequency or the speed of the adapting grating. Togetherwith different observations of temporal frequency anddrift speed determinantsof MAE (Pantle, 1974;Ashida&Osaka, 1995),these data suggestthat both may contributeto MAE magnitude in the same way that both contributeto the perceived speed of moving images (Smith &Edgar, 1994).

EXPERIMENT2: SPATIALFREQUENCYTUNING OFFLICKER MAE

In experiment2, the spatial frequencytuning of flickerMAE was measured. The procedurewas the same as for

experiment 1 except that the spatial frequency of theadaptinggratingswas between 0.125 and 2 c/deg in stepsof one octave. We included these low spatial frequenciesbecause Cameron et al. (1992) presented evidence thatthe spatial frequency tuning breaks down for spatialfrequencies below 0.5 c/deg. Although the results ofAshida & Osaka (1994) argue against spatial frequencyselectivity for flicker test stimuli, to our knowledgenobody has reported whether the flicker MAE at lowspatial scales is spatial frequency tuned. Moreover, thedifferencein resultsbetweenAshida & Osaka (1994)andthose reported by von Griinau and Dub6 (1992: experi-ment 4) requires investigation.The latter found spatialfrequencyselectivitywith flickeringtest stimuli,whereasAshida & Osaka (1994) did not. Differing techniqueshave been suggestedas the sourceof the discrepanciesinthe results (Ashida & Osaka, 1994).

The temporal frequency of the adapting patterns waseither 1 Hz or was varied with spatialfrequencysuch thatdrift speed was 2 deg/sec. For each adapting spatialfrequency, MAEs were measured using five test spatialfrequencies, one of the same spatial frequency, two ofhigherand two of lower spatialfrequency,in stepsof oneoctave. This range was employed except at the lowestspatial frequenciesmeasured,which would have resultedin too few visiblecycles of the grating. In these cases thelowest frequency measuredwas 0.125 c/deg. The resultsof experiment1 showedthat a maximalMAE occurswitha counterphase frequency of around 0.125-0.25 Hz,independently of spatial frequency. Test gratings werecounterphased at a temporal frequency of 0.25 Hzbecause this was near or at the peak of the temporalfrequencytuningcurve.The startingphase of all gratingswas randomizedbefore each presentation.

Results

Estimates of the MAE duration as a function of testspatial frequencyare shown for the two observersin Fig.3. In the top panels, the temporal frequency of theadapting grating was 1 Hz, which means that the speedvaried. In the bottom panels the speed of the adaptinggratingwas 2 deg/see,so temporalfrequencyvaried. Thespatial frequency of the adapting grating is shown alongthe x-axis.

The results are unambiguous: in all cases, it can beseen that the longest MAE is found when the adaptinggratingand test grating are of the same spatial frequency.The data showclear evidencefor spatialfrequencytuningeven at the lowest spatial frequenciesmeasured.

GENERALDISCUSSION

In this paper we have investigated both spatial andtemporal frequency tuning of the flicker motion after-effect. We found no evidence for band-pass temporalfrequency tuning of the flicker MAE, but clear evidencefor band-pass spatial frequency tuning was revealed bymeasuring MAE duration. The data from experiment 1show that maximum MAEs were found using flickeringtestpatterns,counterphasingat low temporalfrequencies.

—.

TEMPORALAND SPATIALFREQUENCYTUNINGOFTHEFLICKERMOTIONAFTEREFFEm 2725

The tuning was independent of the spatial or temporalfrequency of the adapting grating.

The low-passtemporal tuning of flickerMAE suggeststhat flickerMAE may be dominatedby a single low-passtemporal mechanism. The low-pass mechanism must bebroadly tuned because test patterns of high temporalfrequencycan produce robustMAEs, but the peak tuningof the MAE is always at a low counterphase temporalfrequency.It shouldbe emphasizedthat the resultsdo notpreclude the existence of additional temporal mechan-isms which are band-pass and tuned to higher temporalfrequencies. Instead, the results suggest the contributionof such mechanismsto flickerMAE maybe substantiallyless than that of a single, low-pass temporal mechanism.

A comparison of the MAE for the static and counter-phasing test patterns shows no clear distinctionbetweenthe two. Instead, there is a steady transition of MAEmagnitudefrom high temporalfrequencycounterphasinggratings to gratings counterphasing at zero Hz (static).This is supported by the subjective impressions of theMAEs which were approximately the same under thevarious conditions, although of differing duration. Thedata provide no evidence to suggest that the two types ofMAE may be mediated by separate mechanisms. Thisdoes not imply that there are no such mechanisms. Forexample, it is known that the MAE direction oforthogonallydirected transparentmotion (see Verstratenet al., 1994a) can change drastically depending onwhether the test pattern is dynamic or static (Verstratenet al., 1994b). Moreover, differences found in recoveryfrom adaptation with static and dynamic stimuli favor atsvo mechanism interpretation (Verstraten et al., 1996).Also, Culham & Cavanagh (1994) have shown that afterattentive trackingof a radial grating, a MAE is perceivedfor a counterphasing test grating, not for a stationarygrating. MAE studies using inter-ocular transfer techni-ques (IOT) also showgreatdifferencesbetween staticanddynamic test patterns (Raymond, 1993; Nishida et al.,1994; Steiner et al., 1994). In sum, there is plenty ofevidence for different gain controls along the path ofmotion processing. However, the temporal tuningcharacteristicswe report here do not justify the conclu-sion as drawn by Ashida & Osaka (1994).

In experiment 2, we tested a range of spatialfrequencies for spatial frequency tuning of flickerMAE. The results show that the maximum MAE wasfound using flickering test patterns of similar spatialfrequency to that of the adaptingpattern. This result is ingood agreement with von Grunau & Dub6 (1992),notwithstanding the fact that they used a differenttechnique. However, the results are not consistent withthose of Ashida & Osaka (1994), even though they alsoused MAE duration as the dependent variable. Thisinconsistencycontributesto a growing body of evidenceshowing that MAE varies with a numberof experimentalparameters, including the stimulus geometry and themethod of MAE magnitude estimation (Wade, 1994).One key parameter which may contribute to thesedifferences is the use of sub-optimaltemporalconditions

for the flicker frequency of the test pattern. In ourexperiments, we used the optimal temporal frequencydetermined in experiment 1, which should reveal tuningdifferencesmore clearly.

More interesting,perhaps, is the comparison betweenthe results of experiment 2 and those of Cameron et al.(1992). Using static test patterns, these researcherspresented evidence for a “lowest adaptable channel” of0.5 cldeg. This hypothesiswas based on the observationthat the peak MAE for a 0.25 c/deg grating was foundwhen testedwith a static0.5 c/deg grating(Fig. 2, p. 516).This is apparentlynot the case for dynamic “flicker” testpatterns, even at lower spatial frequencies than thosemeasured by Cameron et al. Part of this difference mayarise from the particular method of MAE magnitudeestimation. Cameron et al. used a tracking procedure,where a subject manually matched the speed anddirection of the MAE over a fixed period, using apotentiometer.These data were later combined to give amean velocity estimation. Using this procedure, theseauthors found “no measurable MAE at spatial frequen-cies lower than 0.25 c/deg” (p. 561).Usingour methodofMAE durationestimation,we found robustMAEs at verylow spatial frequencies (we measured as low as 0.125c/deg), even using static test gratings. We have alsoverified that MAEs can be measured using the durationmethod with the same stimulus geometry used byCameron et al. (two horizontal adapting fields, oneabove and one below fixation).This shows that stimulusgeometryalone cannotaccountfor these differences.It istempting to conclude that it might not be possible torecord MAEs using the tracking methodology for lowspatial frequency gratings. Consequently, with thisprocedure it is not possible to record a MAE for lowspatial frequencies or to measure any spatial frequencytuning.

This finding is redolent of early observationsthat thelowest spatial frequency channelwas originallybelievedto be around 1–3 c/deg (Blakemore & Campbell, 1969;Campbellet al., 1981).However, when larger field sizesand higher temporal frequencies were used, separatelyadaptableand maskablemechanismswere found to existdown to 0.2 c/deg (Stromeyer & Julesz, 1972;Tolhurst,1973; Kranda & Kulikowski, 1976; Stromeyer et al.,1982). This suggests that an absence of temporalmodulation,combinedwith a MAE magnitudeestimationtechnique which does not detect MAEs for very lowspatial frequencies,may have contributedto the differentresults of Cameron et al. (1992) and our experiment 2.

Our results and the suggestionsby Ashida & Osaka(1994)may appearto complicatethe understandingof thespatial and temporal tuning of MAEs. Ashida and Osakasuggest that since spatial frequency selectivity is aproperty of mechanisms (channels) at a relatively earlystage of the visual system, the absence of spatialfrequency tuning is evidence for higher level MAEs.However,the narrow spatialfrequencytuningreported inthis paper and by von Griinau & Dub6 (1992) make thisargument disputable. We show that selecting optimal

2726 P. J. BEX et al.

conditionsfor maximizingMAEs can avoid confoundingvariables and reveal the tuning characteristics of themotion mechanisms in human vision.

CONCLUSION

We used the flicker MAE to study the temporal andspatial frequency tuning of the visual system at supra-threshold contrasts. The results show that the magnitudeof the flicker MAE is dependent on the temporalfrequency of the counterphasing test grating, such thatlowest temporal modulation frequenciesgive the largestMAEs. The relationshipis independentof the spatial andtemporalfrequencyof the adaptinggrating.This suggeststhat the flicker MAE is dominated by a single, low-passtemporal mechanism. The data show that the magnitudeof flickerMAE is also dependenton the spatialfrequencyof the test grating, such that the largest MAE is foundwhen the adapting and test patterns are of similar spatialfrequency, even at very low spatial frequencies. Thisrelationship suggests that the flicker MAE involves aseries of band-pass spatial frequency-selectivemechan-isms. The differences between our data and someprevious data may be based on the use of sub-optimalconditions, resulting in weak or non-measurableMAEs,the tuning of which are consequently difficult todetermine.

REFERENCES

Anstis, S. M. (1980). The perception of apparent movement.Philosophical Transactions of the Royal Socie@ of London, B,290, 153–168.

Ashida, H. & Osaka, N. (1994). Difference of spatial frequencyselectivitybetweenstatic and flickermotionaftereffects.Perception,23, 1313-1320.

Ashida, H. & Osaka, N. (1995). Motionaftereffect with flickeringteststimuli depends on adapting velocity. Vision Research, 35, 1825–1833.

Blakemore,C. & Campbell,F. W. (1969).ONthe existenceof neuronsin the human visual system selectively sensitive to the orientationand size of retinal images.Journal ofPhysiology, 203, 237–260.

Cameron, E. L., Baker, C. B. & Boulton, J. C. (1992). Spatialfrequency selective mechanisms underlyingthe motion aftereffect.VisionResearch, 32, 561–568.

Campbell,F. W., Johnstone,J. R. & Ross,J. (1981).An explanationforthe visibility of low frequency gratings. Vision Research, 21, 723–730.

Campbell, F. W. & Robson, J. G. (1968). Application of Fourieranalysisto the visibilityof gratings.Journal ofPhysiology (London),197, 551–566.

Culham, J. C. & Cavanagh, P. (1994). Attentive tracking of acounterphase grating produces a motion aftereffect. InvestigativeOphthalmology and Visual Science, .?5, 1622.

Derrington, A. M. & BadCock,D. R. (1985). Separate detectors forsimple and complex grating patterns?. Vision Research, 25, 1869–1878.

Enroth-Cugell,C. & Robson,J. G. (1966).The contrast sensitivity ofretinal ganglion cells of the cat. Journal of PhysioZo~ (London),187, 51’7-552.

Foster, K. H., Gaska, J. P., Nagler, M. & Pollen, D.A. (1983).Spatialand temporalfrequencyselectivityof neuronesin visual cortexareasV1 and V2 of the macaque monkey.Journal of Physiology, 365,331–363.

Fredericksen, R. E. & Hess, R. F. (1996a). Temporal detection in

humanvision–I:Impulseresponse. Journal ofthe Optical Socie~ ofAmerica A, submitted.

Fredericksen, R. E. & Hess, R. F. (1996b). Temporal detection inhuman vision-II: Mechanisms.Journal of the Optical Society ofAmerica A, submitted.

Georgeson, M. & Harris, M. (1978). Apparent foveofugal drift ofcounterphasegratings Perception,7, 527–536.

Graham, N. V. S. (1989). Visual pattern analyzers. Oxford: OxfordUniversityPress.

von Griinau, M. & Dubcf,S. (1992). Comparing local and remotemotion aftereffects. Spatial Visio~ 6, 303–314.

Hess, R. F. & Plant, G. T. (1985).Temporal frequencydiscriminationin human vision: Evidence for an additionalmechanism in the lowspatial and high temporal frequency region. Vision Research 25,1493-1500.

Hiris, E. & Blake, R. (1992).Anotherperspectiveon the visual motionaftereffect. Proceedings of the National Academy of Science, USA,89, 9025–9028.

Kranda, K. & Kulikowski,J. J. (1976).Adaptationof coarse gratingsunder scotopic and photopic viewing conditions. Journal ofPhysiology, 257, 35P.

Ledgeway,T. (1994).Adaptation to second-ordermotion results in amotion aftereffect for directionally ambiguous test stimuli. VisionResearch, 34, 2879-2889.

Maffei, L. & Fiorentini, A. (1973). The visual cortex as a spatialfrequencyanalyzer. Vision Research, 13, 1255–1267.

Mandler, M. B. & Makous, W. (1984). A three channel model oftemporal frequencyperception. Vision Research, 24, 1873-1880.

McCarthy, J. E. (1993). Directional adaptation effects with contrastmodulatedstimuli. Vision Research, 33, 2653–2662.

Nishida, S., Ashida, H. & Sate, T. (1994). Complete inter oculartransfer of motion aftereffect with flickeringtest. Vision Research,34, 2707-2716.

Nishida, S. & Sate, T. (1995). Motion aftereffect with flickering testpatterns reveals higher stages of motion processing. VisionResearch, 35, 477490.

Over, R., Broerse, J., Crassini, B. & Lovegrove, W. (1973). Spatialdeterminantsof the aftereffects of seen motion.Vision Research, 13,1681-1690.

Pantle,A. (1974).Motionaftereffect magnitudeas a measureof spatio-temporalresponsepropertiesof direction-selectiveanalyzers. VisionResearch, 14, 1229–1236.

Pelli, D. & Zhang, L. (1991). Accurate control of contrast onmicrocomputerdisplays. Vision Research, 31, 1337–1350.

Raymond, J. E. (1993). Complete inter-ocular transfer of motionadaptationeffects on motioncoherencethresholds.Vision Research33, 1865–1870.

Sekuler, R. & Pantle, A. (1967).A model for the aftereffects of seenmovement. Vision Research, 7, 427439.

Smith, A. T. & Edgar, G. K. (1994). Antagonistic comparison oftemporal frequency filter outputs as a basis for speed perception.Vision Research, 34, 253–265.

Steiner, V., Blake, R. & Rose, D. (1994). Inter-ocular transfer ofexpansion,rotation, and translationmotionaftereffects. Perception,23, 1197–1202.

Stromeyer, C. F. & Julesz, B. (1972). Spatial frequency masking invision. Critical bands and spread of masking.Journal of the OpticalSociety of America, 62, 1221–1232.

Stromeyer,C. F., Klein,S., Dawson,B. M. & Spillman,L. (1982).Lowspatial channels in human vision: Adaptation and masking. VisionResearch, 22, 225–233.

Tolhurst,D. J. (1973).Separate channels for the analysis of the shapeand movementof a movingvisual stimulus.Journal of Physiology,231, 385-402.

Verstraten,F. A. J., Fredericksen,R. E. & van de Grind,W. A. (1994).The movementaftereffect of bi-vectorial transparentmotion. VisionReseszrch, 34, 349–358.

Verstraten,F. A. J., Fredericksen,R. E., van Wezel, R. J. A., Lankheet,M. J. M. & van de Grind, W. A. (1996).Recoveryfrom adaptationfor dynamic and static motion aftereffects: Evidence for twomechanisms. Vision Research, 36, 421-424.

TEMPORALAND SPATIALFREQUENCYTUNINGOF THE FLICKERMOTIONAFTEREFFECT 2727

Verstraten, F. A. J., van Wezel, R. J. A., Fredericksen,R. E. & van deGrind, W. A. (1994) Movementaftereffects of transparent motion: Acknowledgements-PB was supported by Canadian NSERC grantThe art of “test” noise. Investigative Ophthalmology and Visual OGPOO01978 to C. L. Baker, FV was supportedby a Niels StensenScience, 35 (suppl.), 1838. Foundationpost-doctoralaward and the NetherlandsOrganizationfor

Wade, N. J. (1994). A selective history of the study of visual motion Scientific Research (NWO-NATO research grant), IM receivedaftereffects. Perception, 23, 1111–1134. support from Canadian MRC grant to C. L. Baker. We thank Eric

Wright, M. J. & Johnstone, A. (1985). Invariant tuning of motion Fredericksen,Walter Makousand Curtis Baker for helpful comments.aftereffect. Vision Research, 25, 1947–1955.

-—