81 the role of feedback in visual attention and...

81 TheRoleofFeedbackinVisual

AttentionandAwareness

stephenl.macknikandsusanamartinez-conde

macknikandmartinez-conde:roleoffeedbackinvisualattentionandawareness 1165

abstract Themammalianvisualsystemincludesnumerousbrainareasthatareprofuselyinterconnected.Withfewexceptions,theseconnections are reciprocal. Anatomical feedback connections ingeneraloutnumberfeedforwardconnections,leadingtowidespreadspeculationthatfeedbackconnectionsplayacriticalroleinvisualawareness.However,evidencefromphysiologicalexperimentssug-geststhatfeedbackplaysamodulatoryrole,ratherthanadrivingrole.Herewediscusstheoreticalconstraintsonthesignificanceoffeedback’sanatomicalnumericaladvantage,andwedescribetheo-retical limits on feedback’s potential physiological impact. Theserestrictionsconfinethepotentialroleoffeedbackinvisualawarenessandruleoutsomeextantmodelsofvisualawarenessthatrequireafundamentalroleoffeedback.Weproposethatthecentralroleoffeedback is to maintain visuospatial attention, rather than visualawareness.Ourconclusionshighlightthecriticalneedforexperi-mentsandmodelsofvisualawarenessthatcontrolfortheeffectsofattention.Asamatterofclarityinthischapter:by“visualaware-ness”or“visibility”wemeantheconsciousperceptionthatastimu-lusisvisible.Thus,forthepurposesofthisdiscussion,weusethetermsvisual awareness, visibility,andconsciousnessinterchangeably.

Anatomical observations of feedback in the visual system

Thevisualareasofthebrainareinterconnectedinacomplexpattern of feedforward, lateral, and feedback pathways(Felleman & Van Essen, 1991). Feedback connections areubiquitousthroughoutthecortex,andsubcorticalregionsinascendinghierarchicalpathwaysalsoreceivealargeamountof feedback from cortical areas (Erisir, Van Horn, &Sherman, 1997; Fitzpatrick, Usrey, Schofield, & Einstein,1994;Guillery,1969;Sherman&Guillery,2002).

Anatomy of Feedback in the LGN Corticogeniculateinput is the largest source of synaptic afferents to the catlateralgeniculatenucleus (LGN).Whereasretinalafferentsonlyencompass25%ofthetotalnumberofinputstoLGNinterneurons,37%of the synapticcontactscome fromthecortex.Inthecaseofrelaycells,therespectivepercentagesare 12% versus 58% (Montero, 1991). Similar estimateshavebeencalculatedintheprimate,andthegeneralagree-mentisthatthecortical-to-retinalinputratioisbetween1:2

and 1:6 in both cats and primates (Erisir et al., 1997;Fitzpatricketal.,1994;Guillery,1969;Sherman&Guillery,2002;VanHorn,Erisir,&Sherman,2000).Boyapati andHenry(1984)concludedthatfeedbackconnectionsfromthecatvisualcortextotheLGNconcentratedalargerfractionoffineaxonsthanfeedforwardgeniculocorticalconnections,presumably resulting in comparatively slower conductionspeeds.

Theseandotherconsiderationsconcerning the synapticsize, efficacy, and contribution of feedback connectionsunderscorethepotentialmistakeinassumingthatanumeri-cally larger number of inputs means that those inputs arefunctionallymostimportant(Sherman&Guillery,2002).

Anatomy of Feedback in the Primary Visual CortexCortical feedforward pathways usually project from thesupragranular layers of visual areas early in the hierarchy(lessthan10–15%oftheconnectionsmayarisefromdeeplayers)andterminateinlayer4ofareaslaterinthehierarchy.In contrast, feedback projections usually arise from theinfragranular layersof laterareasandterminateoutsideoflayer4 in theearlyareas (Barone,Batardiere,Knoblauch,&Kennedy,2000;Felleman&VanEssen,1991;Hilgetag,O’Neill,&Young, 1996a, 1996b;Maunsell&VanEssen,1983).

Direct feedforwardprojections toprimateareaV1 (alsocalledprimaryvisualcortex,striatecortex,andBrodmann’sarea 17) originate from the pulvinar, LGN, claustrum,nucleus paracentralis, raphe system, locus coeruleus, andnucleus basalis of Meynert (Blasdel & Lund, 1983; Doty,1983; Fitzpatrick et al., 1994; Hendry & Yoshioka, 1994;Lachica&Casagrande,1992;Ogren&Hendrickson,1976;Perkel, Bullier, & Kennedy, 1986; Rezak & Benevento,1979). Direct feedforward projections from V1 extend toV2, V3, V5 or MT, MST, and FEF (Boussaoud, Unger-leider, & Desimone, 1990; Livingstone & Hubel, 1987;Lund,Lund,Hendrickson,Bunt,&Fuchs,1975;Maunsell& Van Essen, 1983; Shipp & Zeki, 1989; Ungerleider &Desimone,1986a,1986b).DirectfeedbackprojectionstoV1originatefromV2,V3,V4,V5orMT,MST,FEF,LIP,andinferotemporal cortex (Barone et al., 2000; Perkel et al.,1986; Rockland, Saleem, & Tanaka, 1994; Shipp & Zeki,

stephen l. macknik and susana martinez-conde BarrowNeurologicalInstitute,Phoenix,Arizona

Gazzaniga_81_Ch81.indd 1165 6/19/2009 10:08:47 AM

1166 consciousness

1989; Suzuki, Saleem, & Tanaka, 2000; Ungerleider &Desimone,1986a,1986b).DirectfeedbackprojectionsfromV1 extend to SC, LGN, pulvinar, and pons (Fitzpatricketal.,1994;Fries,1990;Fries&Distel,1983;Gutierrez&Cusick,1997;Lundetal.,1975).

Peters,Payne,andBudd(1994)showedthatonly1–8%ofthesynapticinputsintoalayer4CneuronfromprimateareaV1originateintheLGN.Theyconcludedthat“itisunlikelythattheresponsepropertiesofaparticularcorticalneuronaredominatedbyitsinputfromasinglegeniculateneuron”(p.215).However, thisconclusionwasbasedsolelyon theanatomicalnumbersof inputs,andnoton their functionalproperties,whichwediscussfurtherinthenextsection.

Physiological observations of feedback in the visual system

MethodologicalShortcomingsinPhysiologicalStudiesof Feedback Some visual physiology studies have foundthat feedback connections between the secondary andprimary visual cortices enhance or decrease neuronalresponsiveness without fundamentally altering responsespecificity (Martinez-Conde et al., 1999) (see figure 81.1).These studies were conducted by microinjecting smallamounts of neuronal modulators into area 18 of the catwhilerecordingfromthecorrespondingretinotopicpositionin area 17. This method is accurate in its assessment offeedbackeffectsbecauseitsequestersthesourceofneuronalenhancement and suppression to a small focal region thatcannotdirectlyaffecttheneuronalresponsesoftheneuronsbeingrecordedintheareaofinterest.Thustheonlypossiblecauseoftheresponsemodulationinarea17wasthefeedbackconnection from area 18. Another positive aspect of thistechniqueisthattheeffectsarefullyreversible,whichisnotafeaturesharedbytheablation(Super&Lamme,2007)andlesionmethods.

Otherstudieshaveproposedamoresignificantphysiolog-ical role for feedback in thevisual system.However, thesestudies have generally used alternative methods such asablation, cooling, transcranial magnetic stimulation, anddirectpharmacologicalmanipulationof theneuronsbeingrecorded. Such techniques are usually disadvantageous inthat they are nonfocal, nonreversible, and/or may haveunknown or poorly understood nonspecific effects on thephysiologicalmilieuoftheneuronsbeingdirectlyrecorded(such as by changing the pH, osmolarity, temperature, orother effects). Nonfocal and/or nonreversible techniquesmayalsoaffectthevasculaturefeedingthetargetedneuronsor fibers of passage with known or unknown connectivity(eitherdirectorindirect)tothetargetedneurons.Thustheresults obtained are more difficult to interpret, as theresponsesofthetargetedneuronsmayhavebeenaffectedinwaysunrelatedtoanyputativeroleoffeedback.

Also, as we will discuss more fully in a later section,it is critical that physiological measurements of feedback,as they relate to awareness, be conducted with carefulcontrolsfortheeffectsofattention,aswellasitsunderlyingcircuits.Thisisanecessaryprecaution,asthephysiologicalprocess of attention is differentiable from that of aware-ness (Koch & Tsuchiya, 2007). For a comprehensive dis-cussion of this issue, please see Koch (chapter 79, thisvolume).

What Do We Mean by Feedback? For the purposes ofthischapter,werestrictourdefinitionoftheword“feedback”tothelong-rangefibersthatconnectahigherbrainareatoalowerbrainarea,withinanascendingsensorysystem.Inthis definition, the same information arrives to the sameneuralcircuitatleasttwice:firstasitfeedsforwardthroughthesystemandlateragainasitfeedsback.

Other types of feedback loops in the brain are not dis-cussed in thischapter.For instance, informationmayflowup fromone thalamicnucleus to thecortex (i.e., from theLGNtoareaV1)andthenbackdowntoadifferentthalamicnucleus(i.e.,thepulvinar)(Rockland,1996).Inthiscase,itcouldbesaidthatthethalamusasawholesendsinformationto, and receives feedback from, area V1 (as Rocklanddescribesit).However,thefeedforwardandfeedbackprojec-tions are mediated by two separate thalamic nuclei. Thusthistypeofcircuitdoesnotmeetthedefinitionoffeedbackusedhere.

Herewewilldiscussspecificallythosereciprocalconnec-tionsbetweenvisualareasofthegeniculate-corticalpathway.Thusthefeedbackwewillconsiderentailsconnectionsfromneurons processing more complex visual information (andhavingmorecomplexreceptivefields)toneuronsprocessinglesscomplexvisualinformation(whichhavesimplerandlessselectivereceptivefields)(figure81.2).

Thischapteraimstodescribethepowerfulconstraintsonthefunctionalroleoffeedback,evenwithinsuchanordinaryand basic neural system. Figure 81.2A illustrates the basicconnectivitybetweenanareaofthegeniculocorticalpathwayand the next area up in the hierarchy (i.e., the LGN andareaV1).Thelower,simplerlevelofprocessingfeedsforwardtothehigherlevel.There,informationisfurtherprocessedbyneuralcircuitswithmorecomplexreceptivefields.Thehigher,morecomplexlevelthenfeedsbackinformationtothesimplerlevel.Ifsuchafeedbackconnectionisfunction-ally effective, the receptive fields from the lower level willacquirethespecificityandcomplexitythatcharacterizethehigher-levelreceptivefields(figure81.2B).Thisphysiologicalpredictionshouldapplytoanyfeedbackpathwaysthatarebothengagedandsignificantinstrength.

PhysiologyofFeedback intheLGN CorticogeniculateconnectionstotheLGNareretinotopicallyorganized,and



theypreferentiallyendonLGNlayerswiththesameoculardominanceasthecorticalcellsoforigin(Murphy&Sillito,1996).Althoughonly12–58%(Montero,1991)oftheinputstogeniculatecellsareretinalinorigin,thesesynapsesdrivethe primary responses of geniculate relay cells, whereasfeedbackinputsplayamodulatoryrole(Sherman&Guillery,1998,2002).

PhysiologyofFeedback inthePrimaryVisualCortexCortico-cortical feedback connections are also retinotopi-callyspecific(Salin,Girard,Kennedy,&Bullier,1992).Forinstance,thereisafunctionalprojectionfromarea18toarea17 neurons with similar retinotopic locations (Bullier,McCourt, & Henry, 1988; Martinez-Conde et al., 1999;Salinetal.,1992;Salin,Kennedy,&Bullier,1995).Girard,Hupe, and Bullier (2001) found that feedforward andfeedback connections between areas V1 and V2 of themonkeyhavesimilarlyrapidconductionspeeds.

Inthecatvisualcortex,electricalstimulationfromareasarea 18 and area 19 demonstrated 50% of monosynapticconnectionswithsuperficiallayersofarea17,inregionswithsimilar functional properties, such as retinotopic location(Bullieretal.,1988).MignardandMalpeli(1991)alsofoundthat inactivation of area 18 in the cat led to decreasedresponses in area 17. Martinez-Conde and colleagues(1999) found that focal reversible inactivation of area 18produced suppressedorenhancedvisual responses inarea17 neurons with a similar retinotopy. In most area 17neurons,orientationbandwidthsandotherfunctionalchar-acteristics remained unaltered, suggesting that feedbackfromarea18modulates area17 responseswithout funda-mentallyalteringtheirspecificity.

Inthesquirrelmonkey,SandellandSchiller(1982)foundthat most area V1 cells decreased their visual responseswhen area V2 was reversibly cooled, although a fewcellsbecamemoreactive.Orientationselectivityremained

Figure81.1 Reversibleremovaloffeedbackfromarea18toarea17inthecat.(A)Orientationtuningcurveofacellfromlayers2/3of area 17. (B) Orientation tuning curve of the same cell duringGABAapplicationinarea18.Notethat,althoughthefiringrateofthisarea17neuronincreasessignificantly,itsorientationselec-tivityisvirtuallyunchangedintheabsenceoffeedbackfromarea18. Thus feedback modulates the magnitude of the neuronalresponsesbutdoesnotaffecttheirfunctionalspecificity.(C )Orien-

tation tuning curve of the same cell after area 18 blockade. (D)Solidline,controltuningcurveofanarea18cellrecordedsimul-taneously.Dottedline,tuningcurveofthesamecellafterblockade.Inset:receptivefieldsofbothcells.Forclarity,onlythepreblockadepost-stimulus time histograms (PSTHs) are shown in D. Thenumberof spikes forPSTHsofareas17and18are indicatedatbottomofB andD, respectively.Bin size:100ms.Timebase:1s.(ReprintedfromMartinez-Condeetal.,1999.)


1168 consciousness

unchanged,althoughdirectionselectivitydecreasedinsomeinstances.Bullier,Hupe,James,andGirard(1996)reportedinthecynomologousmonkeythat,followingGABAinacti-vation of area V2, V1 neurons showed decreased orunchangedresponsesinthecenteroftheclassicalreceptivefield,but increasedresponses in the region surrounding it.TheseresultsweresupportedbysubsequentfindingsinareasV1, V2, and V3 following area MT inactivation (Hupeet al., 1998). More recently, Angelucci and colleagues(Angelucci & Bressloff, 2006; Angelucci, Levitt, & Lund,2002) have suggested that area V1 extraclassical receptivefieldpropertiesarisefromareaV2feedback.

Insummary,physiologicalstudiesasawholesuggestthatfeedbackconnectionsinthevisualsystemmayplayamodu-latoryrole,ratherthanadrivingrole,inshapingtheresponsesofhierarchicallylowerareas.Thisevidenceagreeswiththe“no-strong-loops”hypothesisformulatedbyCrickandKoch(1998b). The no-strong-loops hypothesis proposes that allstrong connections in the visual system are of the feed-forwardtype.Thatis,“thevisualcortexisbasicallyafeed-forwardsystemthatismodulatedbyfeedbackconnections,”whichis“nottosaythatsuchmodulationmaynotbeveryimportant for many of its functions” (p. 248). Crick andKocharguedthat“althoughneuralnetscanbeconstructedwithfeedbackconnectionsthatformloops,theydonotworksatisfactorilyiftheexcitatoryfeedbackistoostrong.”Simi-larly,iffeedbackconnectionsformed“strong,directedloops”in thebrain, the cortexwouldas a result “go intouncon-trolledoscillations.”Therefore,therelativenumberoffeed-back versus feedforward anatomical connections to anygivenvisualareamaybemisleadingastotherespectiveroles

ofsuchconnections.Forinstance,thefactthatthecatLGNreceives substantially larger numbers of synapses from thecortexthanfromtheretina(Montero,1991)doesnotneces-sarily mean that corticogeniculate connections are moreimportantthanretinogeniculateconnectionsindeterminingtheresponsecharacteristicsofLGNneurons.

Although the role of feedback modulation in ourvisual perception remains unclear, one possibility is thatfeedback may be involved in attentional mechanisms(Martinez-Condeetal.,1999).Wewilldiscussthisideamorefullyinthenextsection.

The role of feedback in attention

Based on the evidence we have reviewed, one potentiallyimportant role for feedback may be to carry attentionalmodulation signals. Other modulatory roles for feedbackremainpossible,butnoneareasclearlyestablished.Thusitmaybethatthesoleeffectofallfeedbackconnectivityistofacilitateandsuppressattention.Atfirst,giventhemassiveamountofanatomicalfeedbackversusfeedforwardconnec-tions, this possibility may seem unlikely. Indeed, the greatextent of feedback connectivity suggests to some thatfeedback must have a large number of roles (Sherman &Guillery, 2002; Sillito & Jones, 1996). However, we willarguehere that thegreatnumberof feedbackconnectionsmay potentially be explained by the need for top-downattentional modulation alone. Ascending circuits in thevisual system primarily form a labeled-line hierarchy, andso feedback connections necessarily require more wiringthanfeedforwardconnectionstosendbackeventhesimplestsignal.

To illustrate the logic of this argument, let us considertheanatomicalconnectivitybetweentheLGNandV1(figure81.3).Aspreviouslydescribed,LGNrelaycellsreceivemorenumerousfeedbackconnectionsfromthecortexthanfeed-forwardinputsfromtheretina.However,becauseV1recep-tivefieldsareorientationselectiveandLGNreceptivefieldsarenot,anyfunctionallysignificantfeedbackfromV1toagivenretinotopiclocationintheLGNmustrepresentmany,or all, orientations. (One should note that Vidyasagar &Urbas,1982,foundslightorientationbiasesinLGNrecep-tivefields; thesebiasesweremuch smaller than the strongorientationselectivityfoundinV1.)Thatis,foreachunori-ented feedforwardconnection fromtheLGNtoV1, theremust bemanyorientedfeedbackconnectionsfromV1totheLGN,eachwithadifferentorientation,sothatthesumofallfeedbackprojectionsspansalltheorientationspace.Other-wise, if the orientation space of the feedback connectionswerenotfilledcompletely,LGNreceptivefieldswouldshowa substantial orientation bias. Thus anatomical feedbackconnectivitymustbelargesoastorepresenttheentireori-entationspaceateachretinotopiclocation.However,because

A

B

Figure81.2 Ageneralizedmodelof theeffectof feedback inahierarchy of simple to complex neural processing. (A) In a func-tionalhierarchy,informationprocessingbecomesmorecomplexasone ascends in the pathway. (B) When feedback is engaged, thelowerlevelsofthehierarchytakeonthemorecomplexpropertiesoftheupperlevels.



oftheirorientationselectivity,onlyafractionofallfeedbackconnectionswillbeactiveatanygiventime,dependingontheorientationof thevisual stimulus,whereas the feedfor-ward connections will be active irrespective of stimulusorientation.Insummary,themassivefeedbackversusfeed-forwardconnectivityratiocanbemisleading:thislargeratiodoes not necessarily mean that feedback signals are moreimportantormorephysiologicallyrelevantthanfeedforwardsignals.Becausehighervisualareasaremoreselectivethanlower visual areas, only a relatively small fraction of thefeedbackmaybeexpectedtobeactiveatanygivenmoment.Thusfeedbackconnectionsmayneedtotiletheentirespaceofreceptivefieldpropertiesofthehigherlevel;otherwise,thefeedbackwouldimposehigh-levelreceptivefieldcharacter-isticsonthereceptivefieldsoflowerareas.

Figure81.4 illustrates this idea in termsof the feedbackfromdichoptictomonopticlevelsofthevisualpathway.To

beclearaboutthejargon:“monocular”means“withrespecttoasingleeye,”and“monoptic”meanseither“monocular”or“notdifferentbetweenthetwoeyes.”“Binocular”means“withrespecttobotheyes,”and“dichoptic”means“differ-ent in the two eyes.” Thus subcortical levels of the visualsystem are monoptic (because the cells are monocular),whereas cortical visual areas that have binocular circuitsmaypotentiallyprocessdichopticinformation.

To summarize: because receptive fields in ascendingpathwaysbecomemoreselective,larger,andmorecomplexintheirpropertiesasonerisesthroughthehigher levelsofthe brain’s hierarchies, anatomical feedback connectionsmust be more numerous than feedforward connections.Otherwise, the hierarchical nature of the visual systemwouldbediminished(figure81.2).Moreover,thenumerical

If a subset of orientations are fed back to the LGN

In absence of feedback

Every feedforward connection must have many oriented feedback connections; or geniculate cells would be oriented.

Figure81.3 AmodeloftheeffectsoffeedbackfromareaV1onanLGNneuron.NumerousV1-oriented cellsmust feedback toevery feedforward LGN cell in order to account for the lack ofsignificantorientationbiasinLGNreceptivefields.

A

B

Figure 81.4 The effects of feedback from dichoptic levels tomonopticlevelsofvisualprocessing.(A) Ageneralmodelofearlyvisualbinocularintegrationintheabsenceoffeedbackconnections.(B) If significant feedback existed between dichoptic levels ofprocessing and earlier monoptic levels, the earlier levels shouldacquire the properties of the dichoptic levels (i.e., they shouldbecome dichoptic by virtue of the feedback). (Reprinted fromMacknik,2006.)


1170 consciousness

advantageoffeedbackoverfeedforwardconnectionsshouldbeexpectedevenifthereisjustasinglefunctionalroleforfeedback(i.e.,attentionalmodulation).

IfwecombinetheseideaswiththeCrickandKoch’sno-strong-loopshypothesisandthephysiologicalfindingsindi-catingthatfeedbackplaysamodulatoryratherthanadrivingrole,wemayconcludethatfeedbackinputshavemoremod-eratephysiological effects than feedforward inputs,despitebeing anatomically more numerous. This concept is sup-ported by the known physiology: besides their lack oforientation selectivity, another feature that distinguishesLGNfromV1receptivefieldsistheirsmallersize(Allman,Miezin,&McGuinness,1985;Desimone,Schein,Moran,&Ungerleider, 1985; Kastner, Nothdurft, & Pigarev, 1999;Knierim&VanEssen,1991;Zeki,1978a,1978b).If feed-backconnectionsfromV1totheLGNwerefunctionallyasstrong as their feedforward counterparts, LGN receptivefieldswouldbeaslargeasV1receptivefields,buttheyarenot.Thatis,becauseLGNreceptivefieldsaresmallerthanV1receptivefields,feedbackfromV1totheLGNmustbeweakerthantheretinalinputs.

It follows fromthese ideas thatwhenfeedback isopera-tional, some receptive field properties, such as size, whichcontinues to increase throughout the visual hierarchy(Allmanetal.,1985;Desimoneetal.,1985;Kastneretal.,1999;Knierim&VanEssen,1991;Zeki,1978a,1978b)willbefedbackfromhighertolowerlevels.Thuswemaypredictthat,ifattentioniscarriedbyfeedbackconnections,earlierreceptive fields should increase in size when attention is

appliedactively.Thispredictionhasbeenconfirmedexperi-mentally (He,Cavanagh,& Intriligator, 1996;Williford&Maunsell, 2006). Chen and colleagues (2008) recentlyshowed that attentional modulation of V1 neurons in theawakemonkeyisspatiallyspecific:increasingtaskdifficultyenhancedV1neuronalfiringrateat the focusofattentionand suppressed it in surrounding regions, in support ofDesimone and Duncan’s (1995) center-surround model ofattention (figure 81.5). Moreover, response enhancementandsuppressionweremediatedbydistinctneuronalpopu-lations that differed in direction selectivity, spike width,interspike-intervaldistributionandcontrastsensitivity.Thisfinding suggested that attentional feedback facilitates andsuppresses distinct populations of neurons in the primaryvisualcortex.

Toconclude,feedbackconnectionsmaypotentiallyhaveno other function than to modulate (facilitate or suppress)feedforwardsignalsasafunctionofattentionalload.

The role of visual masking, binocular rivalry, attention, and feedback in the study of visual awareness

Let us assume that visual awareness is correlated to brainactivity within specialized neural circuits, and that notall brain circuits maintain awareness. It follows that theneuralactivity that leads to reflexiveor involuntarymotoractionmaynotcorrelatewithawarenessbecauseitdoesnotresidewithinawareness-causingneuralcircuits(Macknik&Martinez-Conde,2009).

DA B C

Figure 81.5 V1 attentional response modulation in awakemonkey single cells during hard and easy tasks. (A) Temporalstructureofatrial.Tworhesusmonkeysweretrainedtofixateonasmallcrosswhilecovertlyattendingtoaspatiallocationthatwascuedat thebeginningof each trial.Thecuewasa thin red ringwithadiameterthatwasthreefoldlargerthanthediameteroftheneuronalreceptivefield (RF).Followingthecue,driftinggratingswerepresentedsimultaneouslyatfivedifferentspatiallocationsfor1.5–3s.Followingarandomizedperiodoftime,oneofthegratingschangedcolor/luminance,andtheanimalwastaskedwithdetect-ing the change by releasing a bar within 0.5s. The attentionalmodulationsweremeasuredatthelastcycleofthedriftinggratingbefore the color change. (B) The color change could be easy or

hard todetectandcouldoccur insideoroutsideof thereceptivefield.(C)Thenumberofcellsthatweresignificantlymodulatedbyattention (red) was much lower during the easy task (top) thanduringthehardtask(bottom).Asubsetof8cellswassignificantlymodulated by attention during both the easy and the hard task(P<0.05).(D) AnincreaseintaskdifficultyleadstoanenhancementofV1visualresponsesatthefocusofattentionandasuppressionoutsidethefocus.Difficulty-enhancedV1neuronshavepoordirec-tional selectivity and broad interspike interval distributions (sus-tainedresponses).Difficulty-suppressedV1neuronsaredirectionalselectiveandhave tight interspike intervaldistributions (transientresponses).(ReprintedfromChenetal.,2008.)



Letusalsoproposethatthereisa“minimalsetofneuralconditions” necessary to achieve conscious visibility (seeChalmers,2000,foranexcellentreviewofthisidea).Suchconditionstaketheformofaspecifictype(ortypes)ofneuralactivitywithinasubsetofbraincircuits.Theminimalsetofconditions will not be met if the correct circuits have thewrongtypeofactivity(toomuchactivity,toolittleactivity,sustained activity when transient activity is required, etc).Moreover, if the correct type of activity occurs, but solelywithincircuitsthatdonotmaintainawareness,visibilitywillalso fail. Finding the conditions in which visibility fails iscriticaltotheresearchdescribedhere;althoughwedonotyetknowwhattheminimalsetofconditionsis,wecannev-erthelesssystematicallymodifypotentiallyimportantcondi-tions (changeneural circuits,modify levelsof activity) andsee if theyresult instimulus invisibility.If so, themodifiedconditionwouldbepart,potentially,of theminimal setofneuralconditionsnecessarytomaintainvisibility.

Toestablishtheminimalsetofconditionsforvisibilityweneedtoansweratleastfourquestions(Macknik,2006).Thequestionsandtheir(partial)answersareasfollows:

1. Whatstimulusparametersareimportanttovisibility?A. Thespatiotemporaledges(alsocurvesandcorners)

of stimuli are the most important parameterstostimulusvisibility(Macknik,Martinez-Conde,&Haglund,2000;Troncoso,Macknik,&Martinez-Conde,2005;Troncosoetal.,2007).

2. What types of neural activity best maintain visibility(transient versus sustained firing, rate codes, burstsof spikes, etc.—that is, what is the neural code forvisibility)?A. Transient bursts of spikes best maintain visibility

(Macknik & Livingstone, 1998; Macknik et al.,2000;Martinez-Conde,Macknik,&Hubel,2000,2002).Seefigure81.6.

3. What brain areas must be active to maintainvisibility?A. Visual areas downstream of V2, lying within the

occipitallobe,mustbeactivetomaintainvisibilityof simple unattended targets (Macknik, 2006;Macknik&Martinez-Conde,2004a;Tse,Martinez-Conde,Schlegel,&Macknik,2005).

4. Whatspecificneuralcircuitswithintherelevantbrainareasmaintainvisibility?A. The specific circuits that maintain visibility are

currently unknown, but their responsivity ismodulated by lateral inhibition (Macknik, 2006;Macknik&Livingstone,1998;Macknik&Martinez-Conde,2004a,2004b;Mackniketal.,2000).

Wemustalsodeterminethesetofstandardsthatwillallowus toconclude thatanygivenbrainarea,orneuralcircuitwithin an area, is responsible for generating a conscious

Figure81.6 MultiunitrecordingfromupperlayersofareaV1inananesthetizedrhesusmonkey.Blackboxesbeloweachhistogramrepresentthetimecourseof themask (M)andtarget (T ).Noticethat under conditions that best correlate with human forwardmasking (interstimulus interval [ISI]of0ms,herecorrespondingtostimulusonsetasynchrony[SOA]of−100ms),themaineffectofthemaskistoinhibitthetransientonset-responsetothetarget.Similarly, in the condition that produces maximum backwardmasking in humans (stimulus termination asynchrony [STA] of100ms,herecorrespondingtoSOAof100ms),theafterdischargeis specifically inhibited.Eachhistogramisanaverageof50trialswithabinwidthof5ms.(ReprintedfromMacknik&Livingstone,1998.)


1172 consciousness

experience.ParkerandNewsomedevelopeda“listofideal-izedcriteria that shouldbe fulfilled ifweare toclaimthatsome neuron or set of neurons plays a critical role in thegenerationofaperceptualevent”(Parker&Newsome,1998,p.230). Ifonereplaces thewords“perceptualevent”with“consciousexperience,”ParkerandNewsome’s list canbeused as an initial foundation for the neurophysiologicalrequirementsneededtoestablishwhetheranygivenneuronor brain circuit may be the neural substrate of awareness(Macknik&Martinez-Conde,2007).ParkerandNewsome’slist(pp.230–231)follows:

1. The responses of the neurons and of the perceivingsubject shouldbemeasuredandanalyzed indirectlycom-parableways.

2. Theneuronsinquestionshouldsignalrelevantinfor-mationwhentheorganismiscarryingoutthechosenpercep-tualtask.Thustheneuronsshouldhavediscerniblefeaturesintheirfiringpatterns inresponsetothedifferentexternalstimulithatarepresentedtotheobserverduringthetask.

3. Differences in the firing patterns of some set of thecandidate neurons to different external stimuli shouldbe sufficiently reliable in a statistical sense to account for,and be reconciled with, the precision of the organism’sresponses.

4. Fluctuationsinthefiringofsomesetofthecandidateneurons to the repeated presentation of identical externalstimuli shouldbepredictiveof theobserver’s judgmentonindividualstimuluspresentations.

5. Directinterferencewiththefiringpatternsofsomesetof the candidate neurons (e.g., by electrical or chemicalstimulation)shouldleadtosomeformofmeasurablechangeintheperceptualresponsesofthesubjectatthemomentthattherelevantexternalstimulusisdelivered.

6. Thefiringpatternsoftheneuronsinquestionshouldnotbeaffectedbytheparticularformofthemotorresponsethattheobserverusestoindicatehisorherpercept.

7. Temporary or permanent removal of all or part ofthe candidate set of neurons should lead to a measurableperceptualdeficit,howeverslightortransientinnature.

However,visualcircuitsthatmaypassmusterwithParkerandNewsome’sguidelinesmayneverthelessfailtomaintainawareness,asweshallexplain.Toguidethesearchfortheneuralcorrelatesofconsciousness(NCC),theminimalneu-ronalmechanismsjointlysufficientforaparticularpercept(Crick & Koch, 1995), some additional standards must beadded.

Thefirstadditionalstandardconcernstheuseofillusionsasthetoolofchoicetotestwhetheraneuronalpopulationorcircuitmaymaintainawareness.Visualillusions,bydefi-nition,dissociatethesubject’sperceptionofastimulusfromitsphysicalreality.Thusvisualillusionsarepowerfuldevicesin the search for the NCC (Myerson, Miezin, & Allman,

1981),astheyallowustodistinguishtheneuralresponsestothephysicalstimulusfromtheneuralresponsesthatcorre-latetoperception.Ourbrainsultimatelyconstructourper-ceptual experience, rather than reconstruct the physicalworld(Macknik&Haglund,1999).Thereforeanawareness-maintainingcircuitshouldexpressactivitythatmatchestheconscious percept, irrespective of whether it matches thephysicalstimulus.Neurons(circuits,brainareas)thatproduceneuralresponsesthatfailtomatchtheperceptprovidethemost useful information because they can be ruled out,unambiguously,aspartoftheNCC.Asaresult,thesearchfortheNCCcanbefocusedtotheremainingneuralcircuits.Conversely,neurons thatdo correlatewithperceptionarenotnecessarilycriticaltoawareness,astheymaysimplyplaya support role (among other possibilities) without causingawarenessthemselves.

The second additional standard derives from a majorcontribution of Crick and Koch’s: the distinction betweenexplicit and implicit representations in the study of visualawareness(Crick&Koch,1998a).Inanexplicitrepresenta-tionofastimulusfeature,thereisasetofneuronsthatrep-resents that feature without substantial further processing.In an implicit representation, the neuronal responses mayaccountforcertainelementsofagivenfeature;however,thefeature itself is not detected at that level. For instance, allvisual information is implicitlyencodedinthephotorecep-tors of the retina.Theorientationof a stimulus, however,isnotexplicitlyencodeduntil areaV1,whereorientation-selective neurons and functional orientation columns arefirstfound.CrickandKochproposethatthereisanexplicitrepresentationofeveryconsciouspercept.

HereweofferthefollowingcorollarytoCrickandKoch’sideaofexplicitrepresentation:Beforeonecantestaneuro-nalpopulationorcircuitforitsroleintheNCC,thespecificneurons (or the population/circuit being tested) must beshown to explicitly process the test stimulus. That is, theneuronsmustrespondtotheteststimulusorshowselectivetuningtosomerangeoffeaturesofthestimulus.Thiscorol-laryconstrainsthedesignofneurophysiologicalexperimentsaimedtotest theparticipationofspecificneurons,circuits,andbrainareasintheNCC.Forinstance,ifonefoundthatretinalresponsesdonotcorrelatewithauditoryawareness,suchadiscoverywouldnotcarrygreatweight.Theneuronsintheeyedonotprocessauditoryinformation,andsoitisnotappropriatetotesttheircorrelationtoauditorypercep-tion. However, this caveat also applies to more nuancedstimuli.WhatifV1activitywastestedforitscorrelationtotheperceptionoffacesversushouses?Facesandhousesarevisualstimuli,butV1hasneverbeenshowntoprocessfacesorhousesexplicitly,despitethefactthatvisualinformationaboutfacesandhousesmustimplicitlyberepresentedinV1.Therefore, one cannot test V1’s role in the NCC usinghousesversus facesandexpect tocometoanymeaningful



conclusion.BecausethatformofinformationisnotexplicitlyprocessedinV1,itwouldnotbeinformativeastoV1’sroleintheNCCifV1neuronsfailedtomodulatetheirresponsewhenthesubjectwaspresentedwithfacesversushouses.Itfollowsthatsomestimuliareincapableoflocalizingaware-nesswithinspecificneuralcircuits,becausenoappropriatecontrolexiststotestfortheirexplicitrepresentation.Forthisreason, binocular rivalry stimuli pose a special problemin localizing the circuits that maintain visual awareness.Binocularrivalry (Wheatstone,1838) isadynamicperceptthatoccurswhentwodisparateimagesthatcannotbefusedstereoscopically are presented dichoptically to the subject(i.e.,each image ispresented independently toeachof thesubject’s eyes). The two images (or perhaps the two eyes)appear tocompetewitheachother,and theobserverper-ceives repetitive undulations of the two images, so thatonlyoneofthemdominatesperceptuallyatanygiventime.(Iftheimagesarelargeenough,thenbinocularrivalrycanoccurinapiecemealfashion,sothatpartsofeachimagearecontemporaneouslyvisible.)

Binocular rivalry has been used as a tool to assess theNCC,butithasgeneratedcontroversybecauseofconflictingresults (Macknik & Martinez-Conde, 2004a; Tse et al.,2005).SomehumanfMRIstudiesreportthatBOLDactivityinV1correlateswithawarenessofbinocularrivalrypercepts(Lee, Blake, & Heeger, 2005; Polonsky, Blake, Braun, &Heeger, 2000; Tong & Engel, 2001). In contrast, otherhuman fMRI studies (Lumer, Friston, & Rees, 1998), aswell as neuronal recording studies in nonhuman primates(Leopold&Logothetis,1996),reportthatactivityinareaV1doesnotcorrelatewithvisualawarenessofbinocularrivalrypercepts. One possible reason for this discrepancy is thatnoneofthesestudiesdeterminedthatthevisualareastestedcontained the interocular suppression circuits necessary tomediatebinocularrivalry.Thatis,sincebinocularrivalryisa process of interocular suppression, the neural circuitsunderlying the perception of binocular rivalry must beshowntoproduceinterocularsuppression—explicitly.Oth-erwise,itcannotbedemonstratedthatbinocularrivalryisavalid stimulus for testing the NCC in those areas. Thusawareness studies using binocular rivalry are valid only inareasthathavebeenshowntomaintaininterocularsuppres-sion.Ifbinocularrivalryfails tomodulateactivitywithinavisual area, one cannot know, by using binocular rivalryalone,iftheperceptualmodulationfailedbecauseawarenessisnotmaintainedinthatareaorbecausetheareadoesnothavecircuitsthatdriveinterocularsuppression.Thisismorethan just a theoretical possibility: as we will describe, wehave shown in the human and the macaque monkey thattheinitialbinocularneuronsoftheearlyvisualsystem(areasV1andV2)arebinocularforexcitationbutmonocularforinhibition.Thatis,theyfailtoprocessinterocularsuppres-sion explicitly (Macknik & Martinez-Conde, 2004a; Tse

V1 V2d V2v V3d V3v V3A/B V4v

-10

0

10

20DichopticMonoptic

% B

OL

D D

iffe

ren

ce (

MO

/SW

I)

A

B

Figure81.7 (A) Summarystatisticsofmonopticversusdichopticmasking responses in the LGN and area V1 of the macaquemonkey.Monoptic(blackbars)anddichoptic(whitebars)maskingmagnitude as a function of cell type: LGN, V1 monocular, V1binocular(nonresponsivetodichopticmasking),andV1binocular(responsive todichopticmasking)neurons. Inset shows the linearregressionofdichopticmaskingmagnitudeinV1binocularneuronsas a function of their degree of binocularity (all neurons plottedweresignificantlybinocularasmeasuredbytheirrelativeresponsestomonoculartargetspresentedtothetwoeyessequentially):BIof0 indicates that thecellsweremonocular,whileaBIof1meansboth eyes were equally dominant. (Reprinted from Macknik &Martinez-Conde, 2004a.) (B) Monoptic and dichoptic maskingmagnitude as a function of occipital retinotopic brain area inthe human. Negative values indicate decreased visual masking(increased targetvisibility),whereasvalues≥0 indicate increasedmasking (decreased target visibility). (Reprinted from Tse,Martinez-Conde,Schlegel,&Macknik,2005.)


1174 consciousness

etal.,2005)(figure81.7).Thereisnocontrolconditionwithwhichtoestablishwhetherbinocularrivalryfailsbecauseofa(mundane)lackofinterocularsuppressionor(moreinter-estingly) fails because of a lack of awareness maintainingcircuits. One could address this issue by using binocularrivalry in tandem with a different stimulus to test for theexplicitrepresentationandstrengthof interocularsuppres-sion, such as visual masking stimuli. In visual masking, amonoptic formof the illusion is available, and soone candistinguish failuresof interocular suppression from failuresofvisualawareness.Butifoneweretousemaskingstimulitoassessinterocularsuppressioninagivenvisualarea,thentheroleofsuchareainmaintainingvisibilityandawarenesswouldhavealsobeenestablished(intheabsenceofbinocu-lar rivalry stimulation), thus obviating the need for subse-quent testing with binocular rivalry stimuli. Because onemustrelyonnon-binocularly-rivalrousstimulitodeterminetheexplicit representationandstrengthof interocular sup-pressioninagivenarea,itisnotpossibletounambiguouslyinterprettheneuralcorrelatesofperceptualstateusingbin-ocular rivalry alone in any visual area, irrespective of thestrengthofthebinocularrivalryresponse.

Our visual masking studies have shown that binocularneurons inareasV1 (thefirst stage in thevisualhierarchywhereinformationfromthetwoeyesiscombined)andV2ofhumansandmonkeyscanintegrateexcitatoryresponsesfromthetwoeyes(Macknik&Martinez-Conde,2004a;Tseetal.,2005)(figure81.7).However,thesesameneuronsdonotexpressinterocularsuppressionbetweentheeyes.Thatis,binocularneuronsinV1arelargelybinocularforexcita-tion while nevertheless being monocular for suppression(thatis,inputfromoneeyewillnotsuppressthefiringrateofaV1orV2cellthatisprimarilytunedforinputfromtheopposite eye). Because most early binocular cells do notexplicitly process interocular suppression, these neuronscannotexplicitlyprocessbinocularrivalry.Further,binocu-larrivalrycannotdistinguishbetweentheroleofinterocularsuppression and the role of awareness at any level of thevisual system. Therefore no conclusions can be reachedabout the localization of the NCC to specific parts of thevisual systembasedonbinocularrivalrystudiesalone.Ifagivenvisualareadoesnotcorrelate tobinocular rivalry, itmaysimplymeanthatinterocularsuppressionisnotatplayinthatarea,ratherthanthatareaisnotmaintainingaware-ness.However,thesefindingsalsobegthequestionofwhysome studies have found binocular rivalry modulation inlow-levelvisualareas(Haynes,Driver,&Rees,2005;Leeetal.,2005;Polonskyetal.,2000;Tong&Engel,2001;Wun-derlich,Schneider,&Kastner,2005).Onepossible reasonforthisparadoxisthatthesestudiesfailedtocontrolfortheeffects of attentional feedback, thus confounding apparentmodulationtointerocularsuppressionwithattentionalmod-ulation. Because the subjects in these studies needed to

attendtothebinocularrivalrystimuli,attentionitself,ratherthanbinocularrivalry,mayhaveproducedtheretinotopicactivationfound.

Monopticallyanddichopticallypresentedvisual-maskingillusions (such as forward and backward masking and thestanding wave) can differentiate between interocular sup-pressionandawareness,andthustheyareimmunetotheseshortcomings.Thereforevisualmasking isan ideal illusionto isolate theNCC.Further,visual-masking illusionsallowus to examine the brain’s response to the same physicaltargetundervarying levelsofvisibility (unlike inbinocularrivalry, where one only measures which of two rivalrousperceptswasdominantatanygiventime,withoutconsider-ationofhowvisiblethatperceptwas).Thusbyquantifyingthe perceptual and physiological effects of visible versusinvisible(masked)targetswewilldeterminemany,ifnotall,oftheconditionsthatcausevisibility.

Weproposethat,totestforexplicitprocessinginneuronalpopulationsorcircuits,oneshoulduseavisualillusion,suchas visual masking, that can be presented in at least twomodes of operation: one mode to ensure that the neuralcircuit inquestion isable toprocess the stimulusexplicitlyandanothermode to test the correlation toawareness. Invisual masking, the monoptic mode establishes that theneural circuit in consideration explicitly processes visual-maskingstimuli,andthenthedichopticmodecanbeusedtoprobetheNCC.

Choosing an appropriate stimulus that is processed atmultiplelevelsofthevisualsystemiskeytolocalizingaware-ness. However, one must take care to control for otherpotential experimental confounds. Lamme and colleaguesusedvisual-maskingstimulitoexaminetheNCCandcon-cluded that stimulus-derived late responses (i.e., after-discharges)areduetofeedbackfromhigherareas(Lamme,Zipser,&Spekreijse,2002)andthatthisfeedbackiscriticalto maintaining awareness (figure 81.8). But if the lateresponsesaredue to feedbackandnot to feedforwardcir-cuits,theirtimingshouldbestablewithrespecttostimulusduration.Thatis,iflateresponsesareduetofeedback,targetduration should not affect their latency, because the feed-backwouldbedrivenbythetarget’sonset-responseasitrisesthroughthevisualhierarchy(figure81.9A).Onthecontrary,if the lateresponsesarecausedby the target’s terminationina feedforward fashion, then targetdurationwouldcriti-callyaffecttheir latency(figure81.9B).Figure81.10showsthat,asthestimulusdurationincreases,sodoesthelatencyoftheafterdischarge,againstthepredictionsfromLamme’sfeedbackmodel(Macknik&Martinez-Conde,2004b).Themodelisfurtherruledoutonpsychophysicalgrounds,astheperceptual strengthofmaskingvarieswith targetduration(Macknik&Livingstone,1998).

Despitethesearguments,Lamme’sgrouphasmaintainedthat late responses are due to feedback. In a recent study


C

A B

Figure81.8 Alternativemodelofvisualbackwardmaskingandawarenessthatrequirerecurrentfeedback.(A)Populationresponsestrength inawakemonkeyV1toafigure-present (thick line)andground (no figure-present, thin line) stimulus. (B) Responses tofigureandgroundconditionsinwhichthefigurewasseen(left)andnotseen(right).Noticethatinthenot-seentrialsthelateresponse

doesnotdifferforfigureandgroundconditions.(ReprintedfromSuper,Spekreijse,&Lamme,2001.)(C )Modelsuggestingthatthevisibility-correlatedlateresponseinpanelsAandBisduetorecur-rent feedback from higher level cortices (activated by the feed-forwardonsetresponse).(ReprintedfromLamme,2003.)

firin

g ra

te

time

stimulusduration

firin

g ra

te

time

stimulusduration

firin

g ra

te

time

stimulusduration

firin

g ra

te

time

stimulusduration

firin

g ra

te

time

stimulusduration

firin

g ra

te

time

stimulusduration

A B

Figure81.9 Predictedtemporaldynamicsofneuronalresponses (asafunctionofstimulusduration) if theafterdischargeis (A)duetorecurrentfeedbackdrivenbythestimulusonsetor(B)drivenbytheterminationofthestimulusinafeedforwardmanner.


1176 consciousness

damage,includingsubstantialdamagetothecorticalvascu-lar systems as well as fibers of passage and nearby neuralstructuressuchastheopticradiations.Thereforeitisunclearexactlywhatprocessesmayormaynothavebeenaffectedbysuchadrasticablation.

Conclusions

Wehavereviewedtheliteratureontheanatomyandphysi-ologyof feedback in thevisual systemandconcluded thatfeedbackconnectionsmaybethesourceofattentionalfacili-tation and suppression, and that other proposed roles forfeedback are not as clearly supported. We have alsoproposed that the large ratio of feedback to feedforwardconnectionsdoesnotnecessarilyindicateasignificantphysi-ologicalroleforfeedback,butmayinsteadbearequirementof any feedback pathway operating within a hierarchicalneuralsystem,suchasthevisualhierarchy.Thisstatementwouldbetrueeveniffeedbacksubservesonlyasinglerole,suchastop-downattentionalmodulation.

Finally,wehavediscussedthestrengthsofvisualmaskinginthestudyofvisualawareness,ascomparedtobinocularrivalry,andhaveconcludedthatvisualmaskingisasoundparadigm in awareness studies, whereas binocular rivalryhas serious shortcomings as a tool to localize the NCC.Usingvisualmaskingasa tool,wehavedevelopedseveralnew standards that must be met to determine the role ofneural circuits, neurons, and brain areas in maintainingconsciousness.

Wehaveemphasizedtheneedtocontrolfortheeffectsofattentionasanimportantstrategyindesigningexperimentsthatlocalizeawareness.Attentioncanenhanceorsuppressthemagnitudeofneuralresponsestoagivenstimulus(Chenet al., 2008; Desimone & Duncan, 1995; McAdams &Maunsell, 1999; Moran & Desimone, 1985; Spitzer,Desimone, & Moran, 1988; Williford & Maunsell, 2006),andthus itmayfacilitateorsuppress itsperceptualaware-ness.However,attentionisadistinctprocessfromawarenessitself (Koch & Tsuchiya, 2007; Merikle, 1980; Merikle &Joordens, 1997; Merikle, Smilek, & Eastwood, 2001). Forinstance,low-levelbottom-uphighlysalientstimuli(suchasflickering lights or loud noises) can lead to awareness anddrawattention,evenwhenthesubjectisactivelyattendingtosomeothertask,ornotattendingtoanything(i.e.,whenthe subject isasleep). It follows thatexperiments to isolatetheNCCshouldcontrolfortheeffectsofattention.

Therefore,weaddthefollowingthreestandardsfortestinga neural circuit’s contribution to awareness to Parker andNewsome’slist:

8.Thecandidateneuronsshouldbetestedwithanillusionthatallowsone todissociate thephysical stimulus from itsperception. If the candidate set of neurons is capable of

Target = 184ms

Target = 33ms

Target = 84ms

Target = 150ms

Target = 50ms

Target = 117ms

Target = 167ms

Target = 217ms

Target = 334ms

Target = 17ms

80 s

pike

s/se

c100ms

Figure81.10 TypicalresponsesfromasingleneuroninmonkeyareaV1toatargetofvariousdurations.Thelatencyandmagni-tude of the afterdischarge grow as the target duration increases.(ReprintedfromMacknik&Martinez-Conde,2004a.)

theysurgicallyremovedtheentireextrastriatevisualcortexof a monkey (V3, V3A, V4, MT, MST, DP, LOP, LIPd,and7a), aprocedurewhich led toa reductionofareaV1late responses (Super&Lamme,2007).However, surgicalablationsare irreversiblebydefinition:onecannot reversethe procedure to show that reinstating the ablated tissuecancelsouttheeffect.Moreover,thesurgicalremovaloftheextrastriatecortex involves theresectionofa largeportionofthecerebralcortex,thuscausingmassivetraumaticbrain



maintainingawareness, theneuralresponsesshouldmatchthe subjective percept, rather than the objective physicalstimulus.

9.Thecandidateneuronsmustexplicitlyprocessthetypeofinformationorstimulususedtotestthem.

10.Theresponsesof theneurons,andof theperceivingsubject,shouldbemeasuredwithexperimentalcontrolsfortheeffectofattention.

acknowledgments WethankMonaStewart,HectorRieiro,andJorgeOtero-Millan for their technical assistance.This studywasfundedbytheBarrowNeurologicalFoundation,NationalScienceFoundation,ArizonaBiomedicalResearch Institute,andScienceFoundationArizona.

REFERENCES

Allman,J.,Miezin,F.,&McGuinness,E.(1985).Stimulusspecificresponsesfrombeyondtheclassicalreceptivefield:Neurophysi-ological mechanisms for local-global comparisons in visualneurons.Annu. Rev. Neurosci., 8,407–430.

Angelucci,A.,&Bressloff,P.C. (2006).Contributionof feed-forward,lateralandfeedbackconnectionstotheclassicalrecep-tive field center and extra-classical receptive field surround ofprimateV1neurons.Prog. Brain Res., 154,93–120.

Angelucci, A., Levitt, J. B., & Lund, J. S. (2002). AnatomicaloriginsoftheclassicalreceptivefieldandmodulatorysurroundfieldofsingleneuronsinmacaquevisualcorticalareaV1.Prog. Brain Res., 136,373–388.

Barone, P., Batardiere, A., Knoblauch, K., & Kennedy, H.(2000). Laminar distribution of neurons in extrastriate areasprojectingtovisualareasV1andV4correlateswiththehierar-chical rank and indicates the operation of a distance rule.J. Neurosci., 20(9),3263–3281.

Blasdel,G.G.,&Lund,J.S.(1983).Terminationofafferentaxonsinmacaquestriatecortex.J. Neurosci., 3,1389–1413.

Boussaoud,D.,Ungerleider,L.G.,&Desimone,R.(1990).Path-ways for motion analysis: Cortical connections of the medialsuperior temporal and fundus of the superior temporal visualareasinthemacaque.J. Comp. Neurol., 296,462–495.

Boyapati,J.,&Henry,G.(1984).Corticofugalaxonsinthelateralgeniculatenucleusofthecat.Exp. Brain Res., 53,335–340.

Bullier,J.,Hupe,J.M.,James,A.,&Girard,P.(1996).FunctionalinteractionsbetweenareasV1andV2inthemonkey.J. Physiol. Paris, 90(3–4),217–220.

Bullier,J.,McCourt,M.E.,&Henry,G.H.(1988).Physiologi-calstudiesonthefeedbackconnectiontothestriatecortexfromcorticalareas18and19ofthecat.Exp. Brain Res., 70,90–98.

Chalmers,D.J.(Ed.).(2000).What is a neural correlate of consciousness?Cambridge,MA:MITPress.

Chen,Y.,Martinez-Conde,S.,Macknik,S.L.,Bereshpolova,Y., Swadlow, H. A., & Alonso, J. M. (2008). Task difficultymodulatestheactivityofspecificneuronalpopulationsinprimaryvisualcortex.Nat. Neurosci., 11(8),974–982.

Crick,F.,&Koch,C.(1995).Whyneurosciencemaybeabletoexplainconsciousness.Sci. Am., 273(6),84–85.

Crick,F.,&Koch,C.(1998a).Consciousnessandneuroscience.Cereb. Cortex, 8(2),97–107.

Crick, F., & Koch, C. (1998b). Constraints on cortical andthalamic projections—The no-strong-loops hypothesis. Nature, 391(6664),245–250.

Desimone,R.,&Duncan,J.(1995).Neuralmechanismsofselectivevisualattention.Annu. Rev. Neurosci., 18,193–222.

Desimone, R., Schein, S. J., Moran, J., & Ungerleider, L. G.(1985). Contour, color and shape analysis beyond the striatecortex.Vision Res., 25,441–452.

Doty, R. W. (1983). Nongeniculate afferents to striate cortex inmacaques.J. Comp. Neurol., 218(2),159–173.

Erisir,A.,VanHorn,S.C.,&Sherman,S.M.(1997).Relativenumbersofcorticalandbrainsteminputstothelateralgeniculatenucleus.Proc. Natl. Acad. Sci. USA, 94(4),1517–1520.

Felleman,D.J.,&VanEssen,D.C.(1991).Distributedhierarchalprocessing in the primate cerebral cortex. Cereb. Cortex, 1(1),1–47.

Fitzpatrick,D.,Usrey,W.M.,Schofield,B.R.,&Einstein,G.(1994). The sublaminar organization of corticogeniculateneuronsinlayer6ofmacaquestriatecortex.Vis. Neurosci., 11(2),307–315.

Fries, W. (1990). Pontine projection from striate and prestriatevisualcortexinthemacaquemonkey:Ananterogradestudy.Vis. Neurosci., 4,205–216.

Fries,W.,&Distel,H.(1983).LargelayerVIneuronsofmonkeystriate cortex (Meynert cells) project to the superior colliculus.Proc. R. Soc. Lond. B Biol. Sci., 219(1214),53–59.

Girard,P.,Hupe, J.M.,&Bullier, J. (2001).FeedforwardandfeedbackconnectionsbetweenareasV1andV2ofthemonkeyhave similar rapid conduction velocities. J. Neurophysiol., 85(3),1328–1331.

Guillery,R.W.(1969).Aquantitativestudyofsynapticintercon-nections in the dorsal lateral geniculate nucleus of the cat. Z. Zellforsch, 96,39–48.

Gutierrez, C., & Cusick, C. G. (1997). Area V1 in macaquemonkeys projects to multiple histochemically defined subdivi-sions of the inferior pulvinar complex. Brain Res., 765(2),349–356.

Haynes, J. D., Driver, J., & Rees, G. (2005). Visibility reflectsdynamicchangesofeffectiveconnectivitybetweenV1andfusi-formcortex.Neuron,46(5),811–821.

He,S.,Cavanagh,P.,&Intriligator,J.(1996).Attentionalreso-lutionandthelocusofvisualawareness[seecomments].Nature, 383(6598),334–337.

Hendry,S.H.,&Yoshioka,T.(1994).Aneurochemicallydistinctthirdchannelinthemacaquedorsallateralgeniculatenucleus.Science, 264(5158),575–577.

Hilgetag,C.C.,O’Neill,M.A.,&Young,M.P.(1996a).Inde-terminate organization of the visual system. Science, 271(5250),776–777.

Hilgetag,C.C.,O’Neill,M.A.,&Young,M.P. (1996b).Onhierarchies.Science, 271(5250),777b.

Hupe,J.M.,James,A.C.,Payne,B.R.,Lomber,S.G.,Girard,P.,&Bullier,J.(1998).Corticalfeedbackimprovesdiscrimina-tionbetweenfigureandbackgroundbyV1,V2andV3neurons.Nature, 394(6695),784–787.

Kastner,S.,Nothdurft,H.C.,&Pigarev,I.N.(1999).Neuronalresponsestoorientationandmotioncontrastincatstriatecortex.Vis. Neurosci., 16(3),587–600.

Knierim,J.J.,&VanEssen,D.C.(1991).NeuronalresponsestostatictexturepatternsinareaV1ofthealertmacaquemonkey.J. Neurophysiol., 67,961–980.

Koch, C., & Tsuchiya, N. (2007). Attention and conscious-ness: Two distinct brain processes. Trends Cogn. Sci., 11(1),16–22.

Lachica,E.A.,&Casagrande,V.A.(1992).DirectW-likegenicu-late projections to the cytochrome oxidase (CO) blobs in


1178 consciousness

primatevisualcortex:Axonmorphology.J. Comp. Neurol., 319(1),141–158.

Lamme,V.A.(2003).Whyvisualattentionandawarenessaredif-ferent.Trends Cogn. Sci., 7(1),12–18.

Lamme,V.A.,Zipser,K.,&Spekreijse,H.(2002).Maskinginter-rupts figure-ground signals in V1. J. Cogn. Neurosci., 14(7),1044–1053.

Lee,S.H.,Blake,R.,&Heeger,D.J.(2005).Travelingwavesofactivity in primary visual cortex during binocular rivalry. Nat. Neurosci., 8(1),22–23.

Leopold,D.A.,&Logothetis,N.K.(1996).Activitychangesinearly visual cortex reflect monkeys’ percepts during binocularrivalry[seecomments].Nature, 379(6565),549–553.

Livingstone,M.S.,&Hubel,D.H.(1987).Connectionsbetweenlayer 4B of area 17 and the thick cytochrome oxidase stripesof area 18 in the squirrel monkey. J. Neurosci., 7(11), 3371–3377.

Lumer,E.D.,Friston,K.J.,&Rees,G.(1998).Neuralcorrelatesof perceptual rivalry in the human brain. Science, 280(5371),1930–1934.

Lund, J. S., Lund, R. D., Hendrickson, A. E., Bunt, A. H., &Fuchs,A.F. (1975).Theoriginof efferentpathways from theprimaryvisualcortex,area17,ofthemacaquemonkeyasshownby retrograde transport of horseradish peroxidase. J. Comp. Neurol., 164,287–303.

Macknik,S.L.(2006).Visualmaskingapproachestovisualaware-ness.Prog. Brain Res., 155,179–217.

Macknik, S. L., & Haglund, M. M. (1999). Optical images ofvisibleandinvisibleperceptsintheprimaryvisualcortexofpri-mates.Proc. Natl. Acad. Science USA, 96(26),15208–15210.

Macknik,S.L.,&Livingstone,M.S.(1998).Neuronalcorrelatesof visibility and invisibility in the primate visual system. Nat. Neurosci., 1(2),144–149.

Macknik,S.L.,&Martinez-Conde,S.(2004a).Dichopticvisualmasking reveals that early binocular neurons exhibit weakinterocular suppression: Implications for binocular vision andvisualawareness.J. Cogn. Neurosci., 16(6),1–11.

Macknik,S.L.,&Martinez-Conde,S.(2004b).Thespatialandtemporal effects of lateral inhibitory networks and their rele-vance to the visibility of spatiotemporal edges. Neurocomputing, 58–60C,775–782.

Macknik,S.,&Martinez-Conde,S.(2007).Theroleoffeedbackin visual masking and visual processing. Advances in Cognitive Psychology, 3(1–2),125–152.

Macknik, S. L., & Martinez-Conde, S. (2009). Consciousness:Neurophysiologyandvisualawarenessin.InL.R.Squire(Ed.),Encyclopedia of neuroscience (Vol. 3, pp. 105–116). Oxford, UK:Elsevier.

Macknik,S.L.,Martinez-Conde,S.,&Haglund,M.M.(2000).Theroleofspatiotemporaledgesinvisibilityandvisualmasking.Proc. Natl. Acad. Sci. USA, 97(13),7556–7560.

Martinez-Conde,S.,Cudeiro,J.,Grieve,K.L.,Rodriguez,R.,Rivadulla,C.,&Acuna,C.(1999).Effectsoffeedbackprojec-tions fromarea18 layers2/3 toarea17 layers2/3 in thecatvisualcortex.J Neurophysiol., 82(5),2667–2675.

Martinez-Conde, S., Macknik, S. L., & Hubel, D. H. (2000).Microsaccadic eye movements and firing of single cells in thestriate cortex of macaque monkeys. Nat. Neurosci., 3(3),251–258.

Martinez-Conde,S.,Macknik,S.L.,&Hubel,D.H.(2002).Thefunctionofburstsof spikesduringvisualfixation in theawakeprimate lateral geniculate nucleus and primary visual cortex.Proc. Natl. Acad. Sci. USA, 99(21),13920–13925.

Maunsell,J.H.,&VanEssen,D.C.(1983).Theconnectionsofthemiddletemporalvisualarea(MT)andtheirrelationshiptoa cortical hierarchy in the macaque monkey. J. Neurosci., 3,2563–2586.

McAdams,C.J.,&Maunsell,J.H.R.(1999).Effectsofattentionon orientation-tuning functions of single neurons in macaquecorticalareaV4.J. Neurosci., 19(1),431–441.

Merikle, P. M. (1980). Selective metacontrast. Can. J. Psychol., 34(2),196–199.

Merikle,P.M.,&Joordens,S.(1997).Parallelsbetweenpercep-tionwithoutattentionandperceptionwithoutawareness.Con-scious. Cogn., 6(2–3),219–236.

Merikle,P.M.,Smilek,D.,&Eastwood,J.D.(2001).Perceptionwithout awareness: Perspectives from cognitive psychology.Cognition, 79(1–2),115–134.

Mignard,M.,&Malpeli,J.G.(1991).Pathsofinformationflowthroughvisualcortex.Science, 251,1249–1251.

Montero,V.M.(1991).Aquantitativestudyofsynapticcontactson interneurons and relay cells of the cat lateral geniculatenucleus.Exp. Brain Res., 86(2),257–270.

Moran,J.,&Desimone,R.(1985).Selectiveattentiongatesvisualprocessingintheextrastriatecortex.Science, 229,782–784.

Murphy,P.C.,&Sillito,A.M.(1996).Functionalmorphologyof the feedbackpathway fromarea17of thecatvisual cortexto the lateral geniculate nucleus. J. Neurosci., 16(3), 1180–1192.

Myerson,J.,Miezin,F.,&Allman,J.M.(1981).Binocularrivalryinmacaquemonkeysandhumans:Acomparativestudyinper-ception.Behav. Anal. Lett., 1,149–159.

Ogren,M.,&Hendrickson,A.(1976).PathwaysbetweenstriatecortexandsubcorticalregionsinMacaca mulattaandSaimiri sciu-reus:Evidenceforareciprocalpulvinarconnection.Exp. Neurol., 53(3),780–800.

Parker, A. J., & Newsome, W. T. (1998). Sense and the singleneuron:Probingthephysiologyofperception.Annu. Rev. Neuro-sci., 21,227–277.

Perkel, D. J., Bullier, J., & Kennedy, H. (1986). Topographyoftheafferentconnectivityofarea17inthemacaquemonkey:Adouble-labellingstudy.J. Comp. Neurol., 253,374–402.

Peters,A.,Payne,B.R.,&Budd,J.(1994).Anumericalanalysisof the geniculocortical input to striate cortex in the monkey.Cereb. Cortex, 4(3),215–229.

Polonsky, A., Blake, R., Braun, J., & Heeger, D. J. (2000).Neuronal activity in human primary visual cortex correlateswith perception during binocular rivalry. Nat. Neurosci., 3(11),1153–1159.

Rezak,M.,&Benevento,L.A.(1979).Acomparisonoftheorga-nizationoftheprojectionsofthedorsallateralgeniculatenucleus,the inferior pulvinar and adjacent lateral pulvinar to primaryvisualcortex(area17)inthemacaquemonkey.Brain Res., 167(1),19–40.

Rockland, K. S. (1996). Two types of corticopulvinar termina-tions: Round (type 2) and elongate (type 1). J. Comp. Neurol., 368(1),57–87.

Rockland,K.S.,Saleem,K.S.,&Tanaka,K.(1994).DivergentfeedbackconnectionsfromareasV4andTEOinthemacaque.Vis. Neurosci., 11,579–600.

Salin, P. A., Girard, P., Kennedy, H., & Bullier, J. (1992).Visuotopic organization of corticocortical connections in thevisualsystemofthecat.J. Comp. Neurol., 320,415–434.

Salin,P.A.,Kennedy,H.,&Bullier,J.(1995).Spatialreiprocityofconnectionsbetweenareas17and18inthecat.Can. J. Physiol. Pharmacol., 73(9),1339–1347.



Sandell,J.H.,&Schiller,P.H.(1982).Effectofcoolingarea18onstriatecortexcellsinthesquirrelmonkey.J. Neurophysiol., 48,38–48.

Sherman, S. M., & Guillery, R. W. (1998). On the actionsthat one nerve cell can have on another: Distinguishing“drivers” from“modulators.”Proc. Natl. Acad. Sci. USA, 95(12),7121–7126.

Sherman,S.M.,&Guillery,R.W.(2002).Theroleofthethala-musintheflowofinformationtothecortex.Philos. Trans. R. Soc. Lond. B Biol. Sci., 357(1428),1695–1708.

Shipp, S., & Zeki, S. (1989). The organization of connectionsbetweenareasV5andV1inmacaquemonkeyvisualcortex.Eur. J. Neurosci., 1(4),309–332.

Sillito,A.M.,&Jones,H.E.(1996).Context-dependentinterac-tions and visual processing in V1. J. Physiol. Paris, 90(3–4),205–209.

Spitzer,H.,Desimone,R.,&Moran,J. (1988).Increasedatten-tion enhances both behavioral and neuronal performance.Science, 240,338–340.

Super,H.,&Lamme,V.A.(2007).Alteredfigure-groundpercep-tion in monkeys with an extrastriate lesion. Neuropsychologia, 45(14),3329–3334.

Super,H.,Spekreijse,H.,&Lamme,V.A. (2001).Twodistinctmodesofsensoryprocessingobservedinmonkeyprimaryvisualcortex(V1).Nat. Neurosci., 4(3),304–310.

Suzuki,W.,Saleem,K.S.,&Tanaka,K.(2000).Divergentback-ward projections from the anterior part of the inferotemporalcortex (area TE) in the macaque. J. Comp. Neurol., 422(2),206–228.

Tong,F.,&Engel,S.A.(2001).Interocularrivalryrevealedinthehuman cortical blind-spot representation. Nature, 411(6834),195–199.

Troncoso,X.G.,Macknik,S.L.,&Martinez-Conde,S.(2005).NovelvisualillusionsrelatedtoVasarely’s“nestedsquares”showthat corner salience varies with corner angle. Perception, 34(4),409–420.

Troncoso,X.G.,Tse,P.U.,Macknik,S.L.,Caplovitz,G.P.,Hsieh, P. J., Schlegel, A. A., et al. (2007). BOLD activationvariesparametricallywithcorneranglethroughouthumanreti-notopiccortex.Perception, 36(6),808–820.

Tse,P.U.,Martinez-Conde,S.,Schlegel,A.A.,&Macknik,S.L. (2005). Visibility, visual awareness, and visual masking ofsimpleunattendedtargetsareconfinedtoareasintheoccipitalcortexbeyondhumanV1/V2.Proc. Natl. Acad. Sci. USA, 102(47),17178–17183.

Ungerleider,L.G.,&Desimone,R.(1986a).Corticalconnectionsof visual area MT in the macaque. J. Comp. Neurol., 248,190–222.

Ungerleider,L.G.,&Desimone,R. (1986b).Projections to thesuperior temporal sulcus from the central and peripheral fieldrepresentationsofV1andV2.J. Comp. Neurol., 248,147–163.

VanHorn,S.C.,Erisir,A.,&Sherman,S.M.(2000).RelativedistributionofsynapsesintheA-laminaeofthelateralgeniculatenucleusofthecat.J. Comp. Neurol., 416(4),509–520.

Vidyasagar,T.R.,&Urbas,J.V.(1982).OrientationsensitivityofcatLGNneuroneswithandwithoutinputsfromvisualcorti-calareas17and18.Exp. Brain Res., 46,157–169.

Wheatstone,C.(1838).Onsomeremarkable,andhithertounob-served,phenomenaofbinocularvision.Philosophical Transactions, 128,371–394.

Williford,T.,&Maunsell,J.H.(2006).Effectsofspatialatten-tion on contrast response functions in macaque area V4. J. Neurophysiol., 96(1),40–54.

Wunderlich,K.,Schneider,K.A.,&Kastner,S.(2005).Neuralcorrelates of binocular rivalry in the human lateral geniculatenucleus.Nat. Neurosci., 8(11),1595–1602.

Zeki,S.M.(1978a).Functionalspecialisationinthevisualcortexoftherhesusmonkey.Nature, 274(5670),423–428.

Zeki, S. M. (1978b). Uniformity and diversity of structure andfunction in rhesus monkey prestriate visual cortex. J. Physiol., 277,273–290.


81 the role of feedback in visual attention and...

Documents