Conscious,preconscious,andsubliminalprocessing:atestableStanislasDehae - PDF document

Conscious,preconscious,andsubliminalprocessing:atestableStanislasDehaene,Jean-PierreChangeux,LionelNaccache,JemeSackurandClaireSergentINSERM-CEACognitiveNeuroimagingUnit,ServiceHospitalierFre´de´ricJoliot,Orsay,France`gedeFrance,Paris,France Correspondingauthor:Dehaene,S.(dehaene@shfj.cea.fr TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx ARTICLEINPRESS1364-6613/$-seefrontmatter2006ElsevierLtd.Allrightsreserved.doi:10.1016/j.tics.2006.03.007 EarlyvisualactivationisnotsufcientforconsciousreportWenowconsidertheneuralbasesofthesecond,transitivemeaningofconsciousness,whichweterm‘accesstoconsciousreport’.Howdoweconsciouslyperceiveavisualstimulus?Manyneuroimagingexperimentshavedemon-stratedatightcorrelationbetweentheconsciousvisualperceptionandtheactivationofstriateandextrastriatevisualareasareas.Forinstance,unmaskingofavisualstimulusincreasesactivityinextrastriateareasintightcorrelationwithsubjectivereportsofstimulusvisibilityvisibility.Furthermore,extrastriateregionsclearlyplayacausalroleinconsciousvisualperception,becausetheirselectivelesioningeliminatesthecorrespondingcontentsfromexperience–forinstancealesionofareaV4candestroycolorperceptioninthecontralaterall.Onthebasisofsuchdata,ZekiZekihasproposedthattheconsciousperceptionofagivenvisualattributeresidesintheextrastriateareaspecializedforthatattribute(e.g.areaMT/V5formotion,orareaV4forcolor).A‘micro-consciousness’wouldbeinvolvedwheneverthatareareceivesasufcientamountofactivation.Weargue,however,thatearlysensoryactivationisnecessarybutnotsufcientforconsciousaccess,becauseactivityinextrastriatevisualareasisfrequentlyobservedwhileparticipantsdenyhavingseenanystimulusstimulus23].Wheninvisibilityiscausedbymaskingmaskingorbydichopticstimulationstimulationthisstimulus-evokedactivityremainsweak,andonemightarguethatitssmallamplitudealonecouldexplaintheabsenceofconsciousconscious.However,thevisualactivationevokedbyinvisiblestimulicanalsobeverystrong,forinstancewheninvisibilityiscausedbyneglectneglectorinattentioninattention.Inarecentstudyoftheattentionalblink,weobservedthatuptoabout180msafterstimuluspresen-tation,theoccipito-temporalevent-relatedpotentialsevokedbyainvisiblewordwerelargeandessentiallyindistinguishablefromthoseevokedbyavisiblewordword.Yetoninvisibletrials,theparticipants’visibilityratingsdidnotdeviatefromthelowestvalue,usedwhennowordwasphysicallypresent.Thus,intenseoccipito-temporalactivationcanbeaccompaniedbyacompletelackofconsciousreport.Top-downamplication,long-distancereverberation,andreportabilityreportabilityandothersothershavesuggestedthat,inadditiontovigilanceandbottom-upactivation,athirdfactorunderlyingconsciousaccessistheextensionofbrainactivationtohigherassociationcorticesintercon-nectedbylong-distanceconnectionsandformingareverberatingneuronalassemblywithdistantperceptualareas.Whywouldthisbrainstatecorrespondtoconsciousaccess?Neurocomputationalsimulationsshowthatoncestimulus-evokedactivationhasreachedhighlyintercon-nectedassociativeareas,twoimportantchangesoccur:(1)Theactivationcanreverberate,thusholdinginformationon-lineforalongdurationessentiallyunrelatedtotheinitialstimulusduration;(2)Stimulusinformationcanberapidlypropagatedtomanybrainsystems.Wearguethatbothpropertiesarecharacteristicofconsciousinformationprocessingwhichinourviewisassociatedwithadistinctinternalspace,bufferedfromfastuctuationsinsensoryinputs,whereinformationcanbesharedacrossabroadvarietyofprocessesincludingevaluation,verbalreport,planningandlong-termmemorymemory.Empirically,accessofsensorystimulitoconsciousreportcorrelateswiththeactivationofhigherassociativecortices,particularlyparietal,prefrontalandanteriorcingulateareas.InfMRI,theactivationofthoseregionssystematicallyseparatesmaskedversusunmaskedpre-sentationsofwordswordsorimagesimages;undetectedversusdetectedchangesduringchangeblindnessblindness;extin-guishedversusseenvisualstimuliinneglectpatientspatients;ormissedversusreportedstimuliduringtheattentionalblink[9,22,23,30–32].Inmanyoftheseparadigms,anterioractivationisaccompaniedbyanamplicationandanincreaseinfunctionalcorrelationwithposteriorstimulus-specicareasareas.Suddenparieto-frontalactivationandtop-downamplicationaretwofrequentsignaturesofconsciousperception.IsattentionaconfoundoranecessityforconsciousSomehavearguedthatmanyoftheaboveneuroimagingparadigmsareinappropriatelycontrolledbecausecon-sciousperceptionisconfoundedwithincreasedattentionandmoreextendedstimulusprocessing.Forinstance,aconsciouswordcanbeattended,repeatedormemorizedwhileanon-consciouswordcannot.Suchconfoundswouldsufcetoexplainthegreaterparieto-prefrontalactivitytounmaskedwordswords.Forthisreason,Tseetal.al.havearguedthatoneshouldpreferexperimentaldesignsinwhichattentionisdrawnawayfromthestimulus.Theyshowthat,insuchasituation,correlatesofstimulusvisibilityarefoundsolelyinoccipitalareas,notinhigherassociativeregions,andthereforearguethatthemech-anismsofconsciousvisualperceptionlieinextrastriatecortex.Weobviouslyagreeononepoint:itisimportanttodesignparadigmsinwhichconsciousperceptionisnotconfoundedwithmassivechangesinovertorcovertbehaviour.However,thisgoalhasbeenachievedinseveralstudies.Inourrecentstudyoftheattentionalblinkblink,forinstance,subjectsviewedaxedstimulusandmadesimilarmotorgesturesonseenandnot-seentrials,yetthosewerestilldistinguishedbystrongparieto-frontalactivation.Wequestion,however,theproposalthatinattentionisanappropriatecontrol.Underconditionsofdivertedattention,suchasthosestudiedbyTseetal.al.,evenanunmaskedstimulusisnotguaranteedtobeconsciouslyperceived.Onthecontrary,considerableevidenceindi-catesthatwithoutattention,consciousperceptioncannotoccur.Intheinattentionalblindnessparadigm,evena700-msstimuluspresentedinthefovea,whenunat-tended,mightfailtobeseenseen.Duringtheattentionalblink,amildlymaskedstimulus,normallyquitevisible,becomesinvisiblewhenattentionisdivertedtoanother TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx ARTICLEINPRESS Therelationsbetweenstimulusstrength,attention,andconsciousperceptionarecomplexbecauseattentionmechanismscanalsobeactivatedautomaticallyinabottom-upmanner.Whenthestimulihavestrongenergy,sharponsetsorstrongemotionalcontent,theymighttriggeranactivationoffrontaleyeeldsoramygdalapathways,thuscausinganamplicationthatcanlowertheirthresholdforconsciousperceptionperception.Thus,bothbottom-upstimulusstrengthandtop-downattentionalamplication(whethertriggeredvoluntarilyorbyauto-maticattraction)arejointlyneededforconsciouspercep-tion,buttheymightnotalwaysbesufcientforastimulustocrossthethresholdforconsciousperception.Consciousperceptionmustthereforebeevaluatedbysubjectivereport,preferablyonatrial-by-trialbasis.Verifyingthatthestimulicanbeconsciouslyperceivedinaseparateexperimentalblockwheretheyareattended,asdonebyTseetal.al.,doesnotsufcetoguaranteeconsciousperceptioninadifferentblockwhereattentionisdiverted.Onecannotsimplyassumethat,byunmaskingstimuli,oneisstudyingtheneuralcorrelatesofconsciousprocessing.DistinguishingaccessibilityfromaccessTheabovedistinctionsleadustoproposalaformaldenitionoftwotypesofnon-consciousprocessesFigure1Subliminalprocessing.Wedenesubliminalproces-sing(etymologically‘belowthethreshold’)asaconditionofinformationinaccessibilitywherebot-tom-upactivationisinsufcienttotriggeralarge-scalereverberatingstateinaglobalnetworkofneuronswithlongrangeaxons.Simulationsofaminimalthalamo-corticalnetworknetworkindicatesthatsuchanon-linearself-amplifyingsystempossessesawell-deneddynamicalthreshold.Aprocessingstreamthatexceedsaminimalactivationlevelquicklygrowsuntilafull-scaleignitionisseen,whileaslightly TRENDS in Cognitive Sciences AbsentPresentPreconsciousConscious ¥ Very little activation¥ Activation is already weak in ¥ Strong feedforward activation¥ Orientation of top-down attention¥ Amplification of sensori-motor activity ¥ Intense activation, yet confined to Figure1.Proposeddistinctionbetweensubliminal,preconscious,andconsciousprocessing.Threetypesofbrainstatesareschematicallyshown,jointlydenedbybottom-upstimulusstrength(ontheverticalaxisatleft)andtop-downattention(onthehorizontalaxis).Shadesofcolorillustratetheamountofactivationinlocalareas,andsmallarrowstheinteractionsamongthem.Largearrowsschematicallyillustratetheorientationoftop-downattentiontothestimulus,orawayfromit(‘task-unrelatedattention’).Dashedcurvesindicateacontinuumofstates,andthicklineswithseparatorsindicateasharptransitionbetweenstates.Duringsubliminalprocessing,activationpropagatesbutremainsweakandquicklydissipating(decayingtozeroafter1–2seconds).Acontinuumofsubliminalstatescanexist,dependingonmaskingstrength,top-downattention,andinstructions(seeBox1).Duringpreconsciousprocessing,activationcanbestrong,durable,andcanspreadtomultiplespecializedsensori-motorareas(e.g.frontaleyeelds).However,whenattentionisorientedawayfromthestimulus(largeblackarrows),activationisblockedfromaccessinghigherparieto-frontalareasandestablishinglong-distancesynchrony.Duringconsciousprocessing,activationinvadesaparieto-frontalsystem,canbemaintainedadlibiduminworkingmemory,andbecomescapableofguidingintentionalactionsincludingverbalreports.Thetransitionbetweenpreconsciousandconsciousissharp,asexpectedfromthedynamicsofaself-ampliednon-linearsystem TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx3 ARTICLEINPRESS weakeractivationquicklydiesout.Subliminalproces-singcorrespondstothelattertype.Notethat,underourhypothesis,subliminalprocessingisnotconnedtoapassivespreadingofactivation,independentofthesubject’sattentionandstrategies,aspreviouslyenvisaged.Onthecontrary,whichevertaskandattentionalsetarepreparedconsciouslycanorientandamplifytheprocessingofasubliminalstimulus,evenifitsbottom-upstrengthremainsinsufcientforglobalignition.Inagreementwiththisanalysis,manytop-downinuencesonsubliminalprocessinghavenowbeenexperimentallyobserved(Box1Preconsciousprocessing.FreudFreudnotedthat‘someprocesses[]mayceasetobeconscious,butcanbecomeconsciousoncemorewithoutanytrouble’,andheproposedthat‘everythingunconsciousthatbehavesinthisway,thatcaneasilyexchangetheunconsciousconditionfortheconsciousone,isthereforebetterdescribedas“capableofenteringconsciousness”orasHerewefurtherspecifythelatterterm.Weproposeto(orpotentiallyconscious,orP-conscious)aneuralprocessthatpotentiallycarriesenoughactivationforconsciousaccess,butistemporarilybufferedinanon-consciousstorebecauseofalackoftop-downattentionalamplication(forexample,owingtotransientoccupancyofthecentralworkspacesystem).Asshownbytheattentionalblinkandinattentionalblindnessparadigms,evenstrongvisualstimulicanremaintemporarilypreconscious.Theyarepotentially(theycouldquicklygainaccesstoconsciousreportiftheywereattended),buttheyarenotconsciouslyaccessedthemoment.Attheneurocomputationallevel,preconsciousproces-singisproposedtoinvolveresonantloopswithinmediumrangeconnectionswhichmaintaintherepresentationofthestimulustemporarilyactiveinasensorybufferforafewhundredmilliseconds.Apreconsciousstimulusmightultimatelyachieveconsciousaccessoncethecentralworkspaceisfreed(asexempliedbythepsychologicalrefractoryperiodparadigmparadigm,inwhichonetaskisputonholdwhileanothertaskisbeingprocessed).Itmightnevergainaccesstoconsciousprocessingifthepreconsciousbufferiserasedbeforetheorientingoftop-downattention(asachievedbymaskingintheattentionalblinkparadigm).AccountingforconictingneuroimagingdataInexperimentalstudiesofconsciousperception,precon-sciousprocessing,asanintermediatecategory,hassome-timesbeenconfoundedwithsubliminalprocessing,andsometimeswithconsciousprocessing.Wenowshowhowthisdistinctioncanprovideasimpleaccountofconictingneuroimagingresults(Figure2(1)Maskingwhenstimuliareattended.Someexper-imentsrequireparticipantstoattendtomaskedstimuliwhicharemadevisibleorinvisiblebychangingthemaskingstrength.Inourtaxonomy,thoseexperimentscontrastsubliminalversuscon-sciousstimuli–amajorcontrastwhichshouldrevealbothearlystimulusprocessingregionsandadis-tributedparieto-frontalworkspacesystem.Indeed,empirically,bothearlyextrastriateandlateparietalandprefrontaldifferenceshavebeenobservedobserved.(2)Stimulipresentedatthreshold.Evenwhenattended,stimulipresentedatsensorythresholdmayormaynotbeperceived.Inourtheorythisisagainacontrastbetweensubliminalandconsciousstimuli.Aspre-dicted,neuroimagingexperimentsrelyingonthiscontrasthaveyieldedbothearly(e.g.P100)andlate(e.g.P300)correlatesofconsciousperceptionerception.Thetheorycanalsoexplainwhyconsciousaccessuctuateseventhoughthestimulusremainsconstant.Simulationsshowthatthethresholdforglobalignitioncanvarybothwithvigilanceandwiththeamountofspontaneousactivitybeforestimulusstimulus.Severalexperimentsconrmthattheperceptionofnear-thresholdstimulicanbepredictedbytheprestimulusstate,inbothhumansandnd.(3)Maskingwhenstimuliarenotattended.Ifstimuliarenotattended,thenthecomparisonofmaskedand Box1.Whyattentionandconsciousnessaredifferent:top-downinuencesonsubliminalprocessing Subliminalprocessingisfrequentlythoughttobeautomaticandindependentofattention.However,thepresentframeworkimpliesthattop-downattentionandtasksetcanhaveaneffectonsubliminalprocessing(seeFigure1inmaintext,toprow).Thispredictionhasbeenveriedinseveralrecentreports.ModulationofsubliminalprimingbytemporalattentionInanumericalmaskedprimingparadigm,Naccacheetalalrstshowedthatsubliminalprimingwaspresentwhensubjectscouldallocateattentiontotheprime-targetpair,butvanishedwhenstimulicouldnotbetemporallyattended.KieferandBrendelBrendelobservedasimilareffectinanexperimentinvestigatingtheN400potentialelicitedbymaskedwords.UnseenmaskedwordselicitedamuchlargerN400whentheyweretemporallyattendedthanwhentheywerenot.Modulationbyspatialattentionetal.al.rstreportedthatblindsightpatientGYcoulduseconsciouslyperceivedcuestoenhanceunconsciousprocessingofvisualtargets.Whenatargetwaspresentedinhisblindvisualeld,GYrespondedfasterandmoreaccuratelywhenitwasvalidlycuedbyaconsciouslyperceptiblearrowpointingtoit,thanwhenitwasinvalidlycued.Inbothcases,hestillclaimedthathecouldnotseethetarget.Modulationofsubliminalprimingbyspatialattentionwasalsoobservedinnormalsubjectssubjects.ModulationbystrategiesTaskinstructionsalsoalterthefateofsubliminalstimuli.Forinstance,maskedprimescanelicitinstruction-dependentactivationinmotorcortexcortex,suggestingthatarbitrarystimulus–responsemappingsconveyedbyconsciousinstructionscanalsoapplytonon-consciousstimuli.Theinuenceisalwaysunidirectional:onceastrategyorresponsemappingisconsciouslyadopted,itextendstonon-consciousprimesprimes.Kundeetal.al.studiedthe‘Grattoneffect’,astrategicincreaseinexecutivecontrolthatfollowsStroopinterferencetrials.Theyobservedthiseffectfollowingconsciousconicttrials,butnotfollowingsubliminalconicttrials.Onceestablished,however,theincreaseincontrolappliedtobothsubliminalandsupraliminaltrials–anotherinstanceofatop-downeffectonsubliminalprocessing. TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx ARTICLEINPRESS unmaskedstimulibecomesacontrastbetweensub-liminalandpreconsciousprocessing.Aspredicted,onlytheearlycomponentsofoccipito-temporalacti-vationareseenseen.Accordingtoourterminology,thesearethecorrelatesofpreconsciousvisualproces-sing(potentialvisibility,yetnoconsciousaccess).(4)Stimulimadeinvisiblebyinattention.Someexper-imentshavecontrastedconsciouslyperceivedstimuliwithstimulimadeinvisiblebydivertingtop-downattention(attentionalblink,changeblindness,inat-tentionalblindness).Thisisacontrastbetweenpreconsciousandconsciousprocessing.Asexpected,thedifferenceappearslate(200–300msafterthestimulus)andinvolvesparieto-prefrontalactivationaswellaslateamplicationofposterioractivityy.ConclusionInsteadoftheclassicalbinaryseparationbetweennon-consciousandconsciousprocessing,weintroducehereatripartitedistinctionbetweensubliminal,preconscious,andconsciousprocessing.Thekeyideaisthat,withinnon-consciousstates,itmakesamajordifferencewhetherstimuliinvisibilityisachievedbyalimitationinbottom-upstimulusstrength,orbythetemporarywithdrawaloftop-downattention.Therstcasecorrespondstosub-liminalprocessing,thesecondtopreconsciousprocessing.Wehaveshownhowthisdistinctionistheoreticallymotivatedandhelpsmakesenseofneuroimagingdata.Isourtaxonomycomplete?Box2brieydiscussesthreeothertypesofnon-consciousknowledgeinthenervoussystem:latentconnectivitypatterns,distributedringpatterns,andfunctionallydisconnectedbrainsystems.Altogether,thesedistinctionsmightsufcetocapturetheknownexperimentalconditionsinwhichinformationescapesconsciousreportability.Theproposedtaxonomyistestable,notonlywithneuroimagingtools,butalsousingelectrophysiologicaltechniquesintheawakemonkey,providedthattaskssimilartotheattentionalblinkandpsychologicalrefractoryperiodcanbedevelopedinthesespecies(seeBox3Ourproposalcouldalsoleadtoareconciliationofseveralmajortheoriesofconsciousperception.Thedistinctionbetweenpreconsciousandconsciousprocessingisconsist-entwithLamme’sproposalofaprogressivebuild-upofrecurrentinteractions,rstlocallywithinthevisualsystem,andsecondmoregloballyintoparieto-frontalal.ItisalsoconsistentwithZeki’shypothesisofanasynchronousconstructionofvisualperceptioninmultipledistributedsitesbeforebindingintoa (a)(b) Conscioushigh strengthand attention T1 versus T3: unmasked or masked stimuli(both attended)T1 versus T2: accessed versus non-accessed stimuliUnmasked words (T1) Preconscioushigh strength,no attentionT3Subliminalweak strengthMasked words (T3)Unmasked words (T2)�masked words (T3)(both used as unattended primes)Seen stimuli (T1�) missed stimuli (T2) during the attentional blin k Figure2.Resolvingcontradictionsinneuroimagingstudies.Schematicrepresentationoftheglobalneuronalworkspacemodel.AvisualtargetT1(ingreen)isconsciouslyaccessedwhenitactivates,inasynchronized,reciprocalandlong-lastingmanner,asetof‘centralworkspace’neuronsparticularlydenseinparietal,prefrontalandcingulatecortices,andwhoselong-distanceconnectionsenablebroadcastingtomanydistantareas.Astimuluscanfailtobecomeconsciousfortworeasons:(1)itmightnothaveenoughbottom-upstrength,forexample,owingtolow-levelmaskingorpresentationclosetothreshold(subliminalstimulusT3,inred);or(2)itmighthaveenoughstrengthtobe,butstillfailtobebylosingthecompetitionforcentralaccessrelativetootherconcurrentstimuliortasksets(preconsciousstimulusT2,inorange).Reinterpretationofneuroimagingexperimentsinthisframework.Whenmaskedandunmaskedstimuliarecontrastedwhilesubjectsareattending(topright),amajordifferenceinbrainactivationisseen,withbothearlysensoryandlateparieto-frontalenhancementsforseenstimuli(illustrationreproducedwithpermissionfromfrom;seesee).Whenmaskedandunmaskedstimuliarecontrastedwhileattentionisdrawnelsewhere(middleright),theeffectofmaskingisconnedtoearlyoccipitemporalcortices(fMRIdatabyKouiderandDehaene;seealsoalso).Whenstimuliareabovethemaskingthreshold,andconsciousaccessismanipulatedbydrawingattentiontoorawayfromthestimuli(bottomright),thedifferenceinactivationislateandconnedtohigherassociationcortices,particularparietalandprefrontalregions(illustrationsfromtheattentionalblinkparadigmreproducedwithpermissionfromfrom–leftimage,andand–rightimage);seealso TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx5 ARTICLEINPRESS .Ouronlysourceofdisagreement–butanimportantone–residesintheirattributionof‘phenomenalconsciousness’or‘micro-consciousness’towhatwehavetermedpre-consciousprocessing.Rememberthat,insuchastate,onlyafewhundredsofmillisecondsafterastimuluswaspresentedandyieldingstrongvisualactivity,participantsdenyperceivinganythingnything.Whethertheyactuallyhadaconsciousphenomenalexperiencebutnopossibilityofreportingit,doesnotseemtobe,atthisstage,ascienticallyaddressablequestion.Theonlyrationaleforattributingphenomenalconscious-nesstopreconsciousprocessingseemstobetheintuitionthatvisualexperienceinvolvesarichnessofcontentthatgoesbeyondwhatwecanreporteport.However,thisintuitionitselfcanbeexplainedasakindofillusion–wethinkthatweseemorethanweactuallydo(Box4Tofurtherexplorethesedifcultissuesinthefuture,itwillbecrucialtodevelopbettermethodsfortheformalcollectionandquanticationofintrospectivereportseports,aswellasforthestudyofthespontaneousowofconsciousprocessesprocesses.Meanwhile,theproposeddistinctionbetweensubliminal,preconsciousandcon-sciousprocessing,andtheidenticationofconscious Box2.Whydoessomeknowledgeremainpermanentlyinaccessible?Ahypotheticaltaxonomy Thisarticlediscussesamodelofhowvisualprocessingcanremainnon-consciousfordynamicalreasonsofinsufcientstrengthorconcurrentattentionalload.AcompletetaxonomysuchasthatproposedinTableI,however,shouldalsocapturethemanytypesofpermanentlynon-consciousknowledgestoredinthenervoussystem.Theglobalneuronalworkspacehypothesisstipulatesthatinfor-mationisconsciouslyaccessibleifitisexplicitlycodedintheringofgroupsofexcitatoryneuronswithbidirectionallinkstoadistributednetworkofworkspaceneurons.Accordingly,informationmightremainpermanentlynon-consciousforatleastthreereasons[6](i)InformationisnotencodedinneuronalringKnowledgestoredinalatentformassynapticefcaciesremainsinaccessibleuntilitisusedtorecreateevokedpatternsofneuralring.Thisconstraintmayexplaininstancesofimplicitlearning,andwhywedonothaveconsciousaccesstomostofourmentalalgorithms.Inthefewcaseswherewedo(e.g.whenwedescribethestepsneededforlongdivision),themodelpredictsthateachstepshouldbeexplicitlycodedintheringofworkspaceneurons.Indeed,experimentally,prefrontalneuronscodingforintentions,plans,ordinalsteps,evaluations,intermediatedecisions,anderrorshavebeenidentied.(ii)InformationisnotrepresentedinexplicitringformformForanaspectofthevisualscenetobeconsciouslyaccessible,theworkspacehypothesisstatesthatitmustberepresentedbysmallgroupsofneuronswhoseringprovidesanunambiguousindexoftherelevantattribute,andwhichwouldbeampliedbytop-downattention.Forinstance,althoughthecollectiveringofV1neuronsencodesallaspectsofthevisualscene,includingthepresenceoffacesorcolor,thoseattributeswouldnotbeconsciouslyseenunlesstheextrastriateareasinvolvedintheirextractionareintact.Atahighercognitivelevel,whenwegainconsciousaccesstoapreviouslysubliminaldistinction(e.g.developmentof‘phonemicawareness’inchildren),neuronalpopulationsselectiveforthislearneddistinctionshouldbefound.(iii)Informationiscodedbyneuronsfunctionallydiscon-nectedfromtheworkspaceEveninformationinexplicitringformcanremainnon-consciousiftherelevantneuronslackthebidirectionalprojectionsappropriatetoestablishareverberantassemblywithparietalandprefrontalcortices.Thisfunctionaldisconnectionhypothesismightexplainwhywehavenoconsciousaccesstothestateofactivityofsubcorticalsystemssustainingbasicmaintenanceprocesses(res-piration,ingestion,posture,etc).Patientswithwhitematterlesions,includingcallosallesions,canalsoloseconsciousaccesstoword,colororobjectinformationthatisstillextracted,yetfunctionallydisconnected. Box3.Questionsforfurtherresearch Canonedesignattentionalblink,psychologicalrefractoryperiod,andpartialreportparadigmsfornon-humanprimates?Cantheybeusedtodissecttheneuralmechanismsofthe‘preconsciousbuffer’?Doesthispreconsciousstateengagesolelylocaloccipito-temporalDoalldemonstrationsofnon-consciousinformationprocessinginhumansfallintooneofthecategoriesoftheproposedtaxonomy?Inparticular,canoneidentifymodelcaseswherethenon-consciousinformationisdemonstrablyencodedinsynapticweights,orinneuralsystemsfunctionallydisconnectedfromparieto-frontalCantranscranialmagneticstimulation(TMS)beusedtodisruptparieto-frontalcircuitsandprobetheircausalinvolvementinconsciousvisualperception?Wouldoccipito-temporalTMSsimi-larlydisruptthepreconsciousbufferduringthepsychologicalrefractoryperiod?Canonendexperimentalmeansoftestingwhetheranysubjectivecontentisassociatedwithpreconsciousstates?Oristheexistenceofnon-reportableconsciousstatesuntestablebyCanbetternon-verbalmethodsbedevelopedforthequanticationofintrospectivereports,bothinhumansandinnon-humanWhattypeofneuralactivitypatternsunderliesintrospectivereports,asopposedtoothermoredirectsensory-motordecisions? TableI.Atheoreticaltaxonomyofconsciousandnon-consciousinformationencodinginthebrainInformationencodingMainfeaturesNon-consciousLatentconnectivitypatternsInformationisencodedinlatentformasmatricesofsynapticweightsDistributedringpatternsInformationisencodedinthedistributedringofmanyneurons,notcondensedinsmallspecializedgroupsofneuronsFunctionallydisconnectedInformationisencodedintheringofneuronsfunctionallydisconnectedfromtheworkspaceSubliminalprocessingProcessingisconnedtoabrieftravellingpulseofringPreconsciousprocessingProcessinginvolveslocalresonantringloops,buttop-downattentionisfocussedonanotherstimulusortaskset.ConsciousprocessingProcessingreceivestop-downamplicationandexpandsintoaglobalparieto-frontalreverberantstate. TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx ARTICLEINPRESS contentswithgloballyaccessibleones,appeartobeproductiveavenuesforscienticresearch.1Paus,T.(2000)Functionalanatomyofarousalandattentionsystemsinthehumanbrain.Prog.BrainRes.126,65–772Zeki,S.(2003)Thedisunityofconsciousness.TrendsCogn.Sci.214–2183Lamme,V.A.(2003)Whyvisualattentionandawarenessaredifferent.TrendsCogn.Sci.7,12–184Dehaene,S.andChangeux,J.P.(2005)Ongoingspontaneousactivitycontrolsaccesstoconsciousness:aneuronalmodelforinattentionalblindness.PLoSBiol.3,e1415Dehaene,S.etal.(2003)Aneuronalnetworkmodellinkingsubjectivereportsandobjectivephysiologicaldataduringconsciousperception.Proc.Natl.Acad.Sci.U.S.A.100,8520–85256Dehaene,S.andNaccache,L.(2001)Towardsacognitiveneuroscienceofconsciousness:basicevidenceandaworkspaceframework.Cognition79,1–377Rees,G.etal.(2002)Neuralcorrelatesofconsciousnessinhumans.Nat.Rev.Neurosci.3,261–2708Crick,F.andKoch,C.(1995)Areweawareofneuralactivityinprimaryvisualcortex?375,121–1239Marois,R.andIvanoff,J.(2005)Capacitylimitsofinformationprocessinginthebrain.TrendsCogn.Sci.9,296–30510Balkin,T.J.etal.(2002)Theprocessofawakening:aPETstudyofregionalbrainactivitypatternsmediatingthere-establishmentofalertnessandconsciousness.125,2308–231911Laureys,S.(2005)Theneuralcorrelateof(un)awareness:lessonsfromthevegetativestate.TrendsCogn.Sci.9,556–55912Gusnard,D.A.andRaichle,M.E.(2001)Searchingforabaseline:functionalimagingandtherestinghumanbrain.Nat.Rev.Neurosci.2,685–69413Bar,M.andBiederman,I.(1999)Localizingthecorticalregionmediatingvisualawarenessofobjectidentity.Proc.Natl.Acad.Sci.U.S.A.96,1790–179314Moutoussis,K.andZeki,S.(2002)Therelationshipbetweencorticalactivationandperceptioninvestigatedwithinvisiblestimuli.Natl.Acad.Sci.U.S.A.99,9527–953215Ress,D.andHeeger,D.J.(2003)Neuronalcorrelatesofperceptioninearlyvisualcortex.Nat.Neurosci.6,414–42016Pins,D.andFfytche,D.(2003)TheneuralcorrelatesofconsciousCereb.Cortex13,461–47417Grill-Spector,K.etal.(2000)Thedynamicsofobject-selectiveactivationcorrelatewithrecognitionperformanceinhumans.3,837–84318Tse,P.U.etal.(2005)Visibility,visualawareness,andvisualmaskingofsimpleunattendedtargetsareconnedtoareasintheoccipitalcortexbeyondhumanV1/V2.Proc.Natl.Acad.Sci.U.S.A.17178–1718319Zeki,S.(1993)AVisionoftheBrain,Blackwell20Dehaene,S.etal.(2001)Cerebralmechanismsofwordmaskingandunconsciousrepetitionpriming.Nat.Neurosci.4,752–75821Vuilleumier,P.etal.(2001)Neuralfateofseenandunseenfacesinvisuospatialneglect:acombinedevent-relatedfunctionalMRIandevent-relatedpotentialstudy.Proc.Natl.Acad.Sci.U.S.A.22Marois,R.etal.(2004)Theneuralfateofconsciouslyperceivedandmissedeventsintheattentionalblink.41,465–47223Sergent,C.etal.(2005)Timingofthebraineventsunderlyingaccesstoconsciousnessduringtheattentionalblink.Nat.Neurosci.24Lamme,V.A.andRoelfsema,P.R.(2000)Thedistinctmodesofvisionofferedbyfeedforwardandrecurrentprocessing.TrendsNeurosci.25Baars,B.J.(1989)ACognitiveTheoryofConsciousness,CambridgeUniversityPress26Haynes,J.D.etal.(2005)VisibilityreectsdynamicchangesofeffectiveconnectivitybetweenV1andfusiformcortex.811–82127Beck,D.M.etal.(2001)Neuralcorrelatesofchangedetectionandchangeblindness.Nat.Neurosci.4,645–65028Pessoa,L.andUngerleider,L.G.(2004)Neuralcorrelatesofchangedetectionandchangeblindnessinaworkingmemorytask.14,511–52029Rees,G.etal.(2002)Neuralcorrelatesofconsciousandunconsciousvisioninparietalextinction.Neurocase8,387–39330Gross,J.etal.(2004)Modulationoflong-rangeneuralsynchronyreectstemporallimitationsofvisualattentioninhumans.Proc.Natl.Acad.Sci.U.S.A.101,13050–1305531Kranczioch,C.etal.(2005)Neuralcorrelatesofconsciousperceptionintheattentionalblink.24,704–71432Vogel,E.K.etal.(1998)Electrophysiologicalevidenceforapost-perceptuallocusofsuppressionduringtheattentionalblink.J.Exp.Psychol.Hum.Percept.Perform.24,1656–167433Mack,A.andRock,I.(1998)InattentionalBlindness,MITPress34Sergent,C.andDehaene,S.(2004)Isconsciousnessagradualphenomenon?Evidenceforanall-or-nonebifurcationduringtheattentionalblink.Psychol.Sci.15,720–72835Anderson,A.K.andPhelps,E.A.(2001)Lesionsofthehumanamygdalaimpairenhancedperceptionofemotionallysalientevents.411,305–30936Freud,S.(1940)AnOutlineofPsychoanalysis,W.W.Norton37Sigman,M.andDehaene,S.(2005)Parsingacognitivetask:acharacterizationofthemind’sbottleneck.PLoSBiol.3,e3738Pashler,H.(1984)Processingstagesinoverlappingtasks:evidenceforacentralbottleneck.J.Exp.Psychol.Hum.Percept.Perform.39Palva,S.etal.(2005)Earlyneuralcorrelatesofconscioussomatosen-soryperception.J.Neurosci.25,5248–525840Super,H.etal.(2003)Internalstateofmonkeyprimaryvisualcortex(V1)predictsgure–groundperception.J.Neurosci.41Linkenkaer-Hansen,K.etal.(2004)Prestimulusoscillationsenhancepsychophysicalperformanceinhumans.J.Neurosci.24,10186–10190 Box4.‘Phenomenalconsciousness’withoutreportability? FollowingWeiskrantzntz,weconsiderthatsubjectivereportsaretheprimarycriterionthatcanestablishwhetheraperceptisconsciousornot.Suchreportsneednotbeverbal.Manyneuroimagingexper-imentsrelyonmanualreportsofconsciousperception,whichcanbemademoreprecisebytheuseofacontinuousvisibilitymeasuremeasure.ThephilosopherNedBlock,however,hassuggestedthatthereportabilitycriterionunderestimatesconsciouscontentscontents.Whenweviewacomplexvisualscene,weexperiencearichnessofcontentthatseemstogobeyondwhatwecanreport.ThisintuitionledBlocktoproposeadistinctstateof‘phenomenalconsciousness’priortoglobalaccess.ThisproposalreceivesanapparentconrmationinSperling’siconicmemoryparadigm.Whenanarrayoflettersisashed,viewersclaimtoseethewholearray,althoughtheycanlaterreportonlyonesubsequentlycuedroworcolumn.Onemightconcludethattheinitialprocessingofthearray,beforeattentionalselectionofaroworcolumn,isalreadyphenomenallyconsciousconscious.However,theseintuitionsarequestionable,becauseviewersareknowntobeover-condentandtosufferfroman‘illusionofseeing’eeing’.Thechange-blindnessparadigmdemonstratesthis‘discre-pancybetweenwhatweseeandwhatwethinkwesee’see’.Inthisparadigm,viewerswhoclaimtoperceiveanentirevisualscenefailtonoticewhenanimportantelementofthescenechanges.Thissuggeststhat,atanygivenmoment,verylittleofthesceneisactuallyconsciouslyprocessed.Interestingly,changesthatattractattentionoroccuratanattendedlocationareimmediatelydetected.Thus,theillusionofseeingmightarisebecauseviewersknowthattheycan,atwill,orientattentiontoanylocationandobtainconsciousinformationfromit.Insummary,accordingtothepresentterminology,thewholevisualsceneisprocessedpreconsciously.Subjectsreport(correctly)thatallofitsdetailsare–butthechange-blindnessparadigmrevealsthattheyarenotnecessarilyconsciously TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx7 ARTICLEINPRESS 42Block,N.(2005)Twoneuralcorrelatesofconsciousness.TrendsCogn.9,46–5243Naccache,L.etal.(2002)Unconsciousmaskedprimingdependsontemporalattention.Psychol.Sci.13,416–42444Kiefer,M.andBrendel,D.Attentionalmodulationofunconscious‘automatic’processes:evidencefromevent-relatedpotentialsinamaskedprimingparadigm.J.Cogn.Neurosci.(inpress)45Kentridge,R.W.etal.(1999)Attentionwithoutawarenessinblind-sight.Proc.Biol.Sci.266,1805–181146Kentridge,R.W.etal.(2004)Spatialattentionspeedsdiscrimi-nationwithoutawarenessinblindsight.Neuropsychologia47Lachter,J.etal.(2004)Forty-veyearsafterBroadbent(1958):stillnoidenticationwithoutattention.Psychol.Rev.111,880–91348Dehaene,S.etal.(1998)Imagingunconscioussemanticpriming.395,597–60049Eimer,M.andSchlaghecken,F.(1998)Effectsofmaskedstimulionmotoractivation:behavioralandelectrophysiologicalevidence.J.Exp.Psychol.Hum.Percept.Perform.24,1737–174750Greenwald,A.G.etal.(2003)Long-termsemanticmemoryversuscontextualmemoryinunconsciousnumberprocessing.J.Exp.Psychol.Learn.Mem.Cogn.29,235–24751Kunde,W.etal.(2003)Consciouscontroloverthecontentofunconsciouscognition.Cognition88,223–24252Crick,F.andKoch,C.(2003)Aframeworkforconsciousness.6,119–12653Weiskrantz,L.(1997)ConsciousnessLostandFound:ANeuropsycho-logicalExploration,OxfordUniversityPress54O’Regan,J.K.andNoe,A.(2001)Asensorimotoraccountofvisionandvisualconsciousness.Behav.BrainSci.24,939–97355Simons,D.J.andAmbinder,M.S.(2005)Changeblindness:theoryandconsequences.Curr.Dir.Psychol.Sci.14,44–48 TRENDSinCognitiveSciencesVol.xxNo.xxMonthxxxx ARTICLEINPRESS