79

e No-Post-perceptual Cognition Paradigm in Action There is, however, a solution to be found in a recent experiment by Brascamp and colleagues [23]. (To avoid misunderstanding, note that this experiment did not involve nystagmus). In this study, the authors reasoned that the detection of prefrontal transitions in binocular rivalry might have to do with the attraction of attention to perceptual transitions rather than perceptual transitions per se. To circumvent this, they designed stimuli for which the transitions would be ‘inconspicuous’ and thus would not draw the subject’s attention. The stimuli were randomly moving dots (Figure 2). Frequently (every 300 ms) there were transitions in which each dot moved in a random direction and the coherence of the dots’ motion (the extent to which dots moved together) shifted. In one condition, the dots in the two eyes were of different colors, whereas in the other condition the dots were of the same color. The key idea is this: for the condition in which the eyes are shown different-colored dots, subjects noticed the rivalrous change of the dominant eye because the color changed. However, for the condition in which the dots were of the same color, the subjects were much less likely to notice the change of dominant eye. The explanation is that they had difficulty detecting whether a change was due to a change of dominant eye or to one of the frequent changes in which each dot moves in adifferent direction. Thus, theconscious changesduetoachangeindominanteyewereinconspicuous and could be expected not to draw attention and not be noticed. (Subjects are generally not aware of which of their eyes is dominating the perception in binocular rivalry). Thus, Brascamp et al. had created a case of binocular rivalry in which the rivalry did not draw attention. How doweknowthat the subjects were actually consciously experiencing two different perceptual contents when the different stimuli dominated? It might be said that both stimuli were experienced simply as randomly moving dots with constantly changing motion and coherence patterns. In that case, there wouldbenoconsciousdifference between one eye dominatingandtheothereyedominating. Brascamp 1008 Trends in Cognitive Sciences, December 2019, Vol. 23, No. 12Trends in Cognitive Sciences etal.addressedthisissue.Theywereabletoconfirm,usinganumberofdifferentmethods,thatperceptual binocular rivalry was occurring for the same-color patterns in Figure 2; that is, that the two displays triggered different perceptions. One such method involved a separate experiment varying the dot density in the twoeyes. Notethat, at any givenmoment,thearraysshowntothetwoeyesdifferedinthedirection of motionofeverydotandtheextenttowhichitsmotioncoheredwiththemotionofotherdots.Whenthe two eyes receive different inputs, there are only two alternative perceptual situations: (i) the percept can reflect some kind of combination or merger of the two inputs; or (ii) the percept can reflect rivalrous alternation of the sort described above. See Box 1 for an explanation of the factors that determine merger as opposed to rivalrous alternation. When Brascamp et al. asked for reports of density, they tended to get responses indicating that the perceived density reflected the number of dots shown to one eye rather than any sort of combination or merger of the two eyes. Thisresult confirmedthattherewaslittlemergingintheirprocedure,thusshowingthatoftensubjects must be experiencing real conscious rivalry. They further confirmed conscious rivalry by exploiting known temporal regularities of rivalry. Subjects were consciously experiencing repeated shifting of the patterns, but, as noted above, they had no way of distinguishing between the transitions that ref lected rivalry and those that reflected the regular change of patterns. Were the rivalrous switches inaccessible or unreportable? No– they were accessible but mostly not accessed. The rivalrous switches were (mostly) indistinguishable from the switches that were happening every 300 ms, so the rivalrous switches did not stand out and were not noticed and so did not produce differential cognitive states. Brascamp et al. state (p. 1674): ‘Based on the sensitivity index, d0, detection of switches in the samecolorcondition could not bedistinguishedfromchance,demonstratingjust how inconspicuousthese switchesare’. Thatis, thesubjectswere(approximately) at chanceon distinguishing the rivalrous switches from the run-of-the-mill switches that were happening all the time. All switchesrivalrous and nonrivalrous– were noticeable, accessible, and reportable, but the subjects mostly could not pick out the rivalrous switches from the ones that were happening all the time. So, what did Brascamp et al. find? Using fMRI, they could detect prefrontal differences in the case of different color dots but not in the case when the dots were of the same color. As they say, prefrontal differences in activation for the inconspicuously different stimuli were ‘altogether undetectable in our procedure’. They conclude that ‘when viewing a conflicting or ambiguous Box 1. Rivalry versus Merger Whendifferent stimuli are presented to the two eyes, in some cases the perceptions merge and in other cases theyalternate in dominatingtheperception.Whatdeterminesthedifference?FigureIshowsamasculineanda feminine face presented, each to a different eye. If the subject attends to local features or to parts of the faces, suchastheeyes,thesubjectexperiencesstandardbinocularrivalry,asindicatedinthediagramby‘Alternating percepts’. Standard binocular rivalry also occurs if the faces are presented upside down. However, if the subjects attend to holistic features such as gender, there is fusion instead of rivalry: the subject sees an androgynous face that blends masculine and feminine features [25]. What determines whether there is rivalry or merger? We can call rivalrous stimuli ‘incompatible’ and merging stimuli ‘compatible’. Mere luminance differences do not contribute to incompatibility as long as the items of different luminance are of the same shape and contrast polarity. (Contrast polarity is a matter of being lighter ordarkerthanthebackground.BoththeXandtheOofFigureIinBox2compriseelementsthatarelighterthan the background,sotheyarethesameincontrastpolarity). The items of Figure Iin Box2 arecompatiblelocally because the squares are of the sameshapeandcontrast polarity but differ globally (X vs O). All of the subjects in [26] experienced fusion of these items rather than rivalry. Other experiments using more complex setups have also reported that holistic differences do not matter if there are no relevant local differences [27]. The explanation of the primacy of the local is that local properties tend to be represented in monocular parts of the visual cortex. There is more binocular processing in higher vision where receptive field sizes are much larger than in monocular areas. Binocular rivalry usually depends on monocularly represented features, features that differ in the twoeyes, whereaswhatwenaivelythinkofasthecontentsofperception(e.g.,seeinganXvsseeinganO)arerepresentedbinocularly,mainlyinhigherareas.SeeBox2foranapplicationofthispointtothepredictivecodingparadigm. Trends in Cognitive Sciences, December 2019, Vol. 23, No. 12 1009stimulus,aswitchinperceptionmayariseinthevisualsystem,butnoticingthechangemayrelyon brainregionsdedicatedtobehavioral responses’ ([23], seep.1677).Theupshot isthat itmaybe noticingthatbrings inprefrontallyrepresentedcognitiveconcepts; theperceptions incases that donotdrawattentionarebasedinperceptualareas inthebackandmiddleof thehead. NotethatalthoughBrascampetal. justifiedthemethodof inconspicuousswitchesbyappealingto theneedtoavoidattractingattentiontotheswitches, thereisanotherbenefitof theprocedure, andonethat isthefocushere: thismethodavoids thesystematicchangeofcognitivestates like Trends in Cognitive Sciences FigureI.BinocularRivalryStimuli. Themasculinefaceispresentedtothelefteyeandthefemininefacetotherighteye.Ifthesubjectisattendingto localfeaturesorpartssuchastheeyes,standardbinocularrivalryensues.Thisisindicatedbythelabel‘alternating percepts’andillustratedasfirstamasculineface, thenafeminineface, thenamasculineface. If thesubject is attending toholistic features suchas gender or the identityof theperson, the subject sees apersisting morphedandrogynousface(‘stablemorph’),aspictured.ThankstoChrisKlinkforthisfigure.See[25]. 1010 TrendsinCognitiveSciences,December2019,Vol.23,No.12 TrendsinCognitiveSciencesthoughtandjudgmentthatcanaccompanyrivalrouschanges. Ifonefirstthinksthatoneisseeinga gratingmovingupwardandthenthinksthatoneisseeingagratingmovingdownward,thatchangeof thought,basedintheprefrontalcortex,can, ifcorrelatedwiththepercept,seemtoindicateaprefrontalchangeeveniftheprefrontalcortexplaysnoroleinconsciousnessatall. Inprecludingsuch asequenceofevents, theirmethodology isnot justano-reportmethodology: it isano-differential-post-perceptualcognitionmethodology.Theupshotisthat, inatleastoneinstanceofano-differential-post-perceptualcognitionparadigm, theremaybenoprefrontalcomponenttoconscious change. AnticipatinganObjection IcanimagineanobjectionbeingraisedthattheunivariatefMRIusedinthestudybyBrascampetal.is notsensitiveenoughtocapturethedifferencesbetweenthetwopercepts.Caseinpoint,Odegaard etal. [24]arguedthatfailuretofindprefrontaldifferenceswithfMRImaymissrealeffectsthatwould bedetectedbyothermeans.TheymentionthatECog(electrocorticography) recordings inwhich Box2.RivalryandPredictiveCoding Thelocalnatureofbinocularrivalrycreatesdifficultiesformanytheoriesofperception.Forexample,the‘predictiveprocessing’approach[28]regardsperceptionasamatterof‘controlledhallucination’inwhichhypothesesinvolvinggeneralknowledgeofwhatmightbeseenconfrontthedata.Asmentionedabove,acommonly usedbinocularrivalrystimulusisoneinwhichafaceisshowntooneeyeandahousetotheother.Whatthe subjectexperiencesisanalternationbetweenafaceandahouseperceptioninwhicheachfillsthewholevisual fieldforabriefperiod.Advocatesofthepredictiveprocessingapproachseetheissueintermsofconflictbetweenafacehypothesisandahousehypothesis,fedbythegeneralknowledgethatnothingisbothafaceanda house.AndyClark[29]explainsthereasoning, following[30]: ‘Butwhy,undersuchcircumstances,dowenotsimplyexperienceacombinedor interwovenimage:akindof house/facemash-upforexample?.Suchmash-upsdonotconstituteaviablehypothesisgivenourmoregeneral knowledgeaboutthevisualworld.Foritispartofthatgeneralknowledgethat,forexample,housesandfacesdo notoccupythesameplace,atthesamescale,atthesametime..This,indeed,maybethedeepexplanationofthe existenceofcompetitionbetweenthehigher-levelhypothesesinthefirstplace–thesehypothesesmustcompete becausethesystemhaslearnedthat‘‘onlyoneobjectcanexistinthesameplaceatthesametime’’…’ However,aswehaveseen,besidesthehigh-levelattentionofthesortmentionedinconnectionwithFigureIin Box1,generalknowledgehypothesesaboutfacesbeingdifferentfromhousesandonethinginoneplaceat onetimeplaylittleroleindeterminingthedifferencebetweencompatibleandincompatiblestimuli.Themost importantdeterminantsarelow-level local features. Trends in Cognitive Sciences FigureI.StimulusShowingtheLocalNatureofRivalry Thesearesamplestimuliusedinabinocular rivalryexperiment.Eachimageisprojectedtoadifferenteye. ThankstoThomasCarlsonforthisfigure.See[26]. TrendsinCognitiveSciences,December2019,Vol.23,No.12 1011 TrendsinCognitiveSciencesTrends in Cognitive Sciences there is direct intracranial electrophysiological recording in human surgical epileptics (in which the skull is opened and electrode grids are placed on the cortex) showed differential prefrontal activity even when subjects were not required to report the stimulus. Odegaard et al. are correct in that ECogmaypick upchanges missed by fMRI, but using patients to test rivalry without doing anything to keep them from thinking about what they are experiencing does not satisfy the point made here that the subjects maybethinking abouttheperceptualcontents evenwhennotrequiredtoreporton them.Still, the basic point that Odegaardetal. weremakingdoesapplytotheBrascampetal.experiment since the rivalry changes in that experiment, involving only the direction and coherence of motion of dots, might require temporal and spatial resolution too fine grained for fMRI. Further, when one rivalrous percept is replaced by the other, a conscious change occurs even if the subject does not noticeit. Whenonerivalrousperceptisreplacedby theother,thefirstisweakenedbyadaptation andthenreplacedbyastrongerpercept.This changeinlevelofperceptual activation might showup in fMRI but would certainly be detected by ECog. Thus, although the Brascamp et al. experiment is evidence against prefrontalism, further experiments with ECog could support prefrontalism. Concluding Remarks In sum,the‘‘no-report’’ methodology was anoverlysimple approach to themethodologicalproblem of isolating the neural basis of consciousness. The real methodological problem is how to distinguish the neural basis of conscious perception from the neural basis of thought, judgment, and reasoning (see also Outstanding Questions). There may be no general solution to this problem, but at least in the caseofbinocular rivalry a solution may be at hand.Ibelieve that theexperiment reported inBrascampetal. does provide some support for noncognitive theorists, but the main point of this opinion articleistohomeinonthemethodological issueratherthansupportingoneside