tion - PDF document

tion—ahorizontalsurface(onwhichtosit)properlyalignedwithaverticalsurface(tosupportone’sback).Tosolvethisproblem,theoriesofhumanobjectrecognitionandalgorithmsfortherecognitionofobjectsbymachinesoftenincludearoleforcategorylearning(e.g.,Duvdevani-Bar&Edelman,1999;Ullman,1996).Theimplicationisthatcategory-relevantdescriptionsofobjectshapearenotapriorideterminantsofobjectcategoriesbutratherthedevelopmental product of forming those categories.Recentdevelopmentaldatasupportthisidea.Duringtheperiodfrom1to3yearsofage,childrenacquirenamesforcommonobjectcategories,andtheyalsoshiftfrommorefeature-basedtomoreglobal-shape-basedobjectrecognition(see,forexample,Jones&Smith,2004;Nelson,1972;Quinn,2004;Rakison&Cohen,1999).Inonestudy(Smith,2003),youngchildrenwerepresentedwithrichlydetailedandtypicalinstancesofcommonobjectcategories,orwithminimalistthree-dimensionalrepresentationsofthesesamethings.Theseminimalistrepresentationsweremadefromthreetofourgeometricvolumesalignedintheproperspatialarrangement.Theresultsshoweddramaticdevelopmentalchangelinkedtochildren’sknowledgeofcommonobjectcategories.Two-year-oldswhoknewrelativelymanyobjectcategoriesfortheiragerecognizedtheminimalshaperepresentationsaswellastheydidtherichlydetailedones.Incontrast,same-agedchildrenwhodidnotknowasmanycategoriesdidnotrecognizetheminimalshapesatall,eventhoughtheyrecognizedtherichlydetailedversionsaswellasthemoreadvancedchildren.Thusthereseemstobealinkbetweenlearningobjectcategoriesandlearningaminimalistdescriptionofobjectshape,thekindofdescriptionthatwouldallowfortheimmediaterecognitionofawidevarietyofin-stancesofacategory.Insupportofthissecondidea,asecondexperiment(Smith,2003)showedthatthemoreadvanced2-year-oldswereabletorecognizeminimalistversionsofevennovelobjects.Childrenthusappeartobelearningnotjustthecategory-relevantshapesofspecificthings,buttheyarealsolearningtodescribetheshapeofanythinginacate-gory-relevantway.1.2.Action is also a good candidate for teaching category-relevantAstheoristsfromavarietyofperspectiveshavenoted,thereisacausallinkbetweenshapeandfunction(e.g.,Gelman&Bloom,2000;J.J.Gibson,1966).Itisbecausechairshavehorizontalandverticalsurfacesthatwecansitonthemwithbacksupport,andthustheseatsandbacksofchairsaremoredefiningof“chair-shape”thanthecurveofthelegsorthepresenceofarms.Aroleforactionisalsosuggestedbygrowingevidencethatourhabitualmannerofactingonanobjectispartofthememoryforthatobject,suchthatthevisualrecognitionofathingautomaticallyactivatestheactionsassociatedwithit(seeCreem&Proffitt,2001;Ellis&Tucker,2000;Glenberg,Robertson,Kaschak,&Malter,2003).Otherevidenceindicatesthatobject-relatedactionsandobjectidentificationrelyoncommonbrainregions(e.g.,Faillenot,Toni,Decety,Gregoire,&Jeannerod,1997;Grezes&Decety,2001).Mightnotaction,then,playaformativeroleintherepresentationofanobject’sshapeandperhapsalsoincreatingtheverydimensionsofshapeacrosswhichobjectsarecategorizedandcompared?L. B. Smith/Cognitive Science29(2005) 2.1.2.StimuliFig.1showstheexemplarandtwoofthetestobjects.Theexemplarhadabasediameterof6cmandaheightof6cm.Threeverticallyextendedtestobjectshadthesamebasebutdifferingheightsof8cm(V1),12cm(V2),or16cm(V3).Threehorizontallyextendedtestobjectshadheightsof6cm;thebaseofthesetestobjectsvariedprogressivelyfromacircletoellipsewithmajordiametersof8cm(H1),12cm(H2),and16cm(H3).TheseobjectswerecarvedfromwhiteStyrofoamwithblackprotrusionsonthesidesforaddedinterest.Ameter-widebymeter-tallwoodenbackdropwithgrasspaintedalongthebottomandatreeontherightsidewasusedtoguidetheactions.Forwarm-up,theexperimenterusedatoyflower,bunny,car,and2.1.3.Forced-choice procedureTheforced-choiceprocedurebeganwithawarm-uptasktohelpchildrenunderstandthetesttrials.Thechildwaspresentedwithtwowell-knownobjects,forexample,theflowerandbunny,placedonthetableapproximately20cmapart.Thechildwasasked,“Whereisthebunny?Showmethebunny?”Childrenwerehelpedtomakethisfirstresponseiftheydidnotdosoimmediatelyandwarmfeedbackwasprovidedforallcorrectresponses.Thiswasrepeatedwiththeexperimentervaryingtherequestedobjectuntilthechildrespondedcorrectlyon three consecutive trials. All children did so within eight warm-up trials.Immediatelyafterthewarm-uptrials,themainexperimentbeganwiththeprimingevent.Theexperimenterandchildstoodononesideofatablefacingthebackdrop.TheexperimenterL. B. Smith/Cognitive Science29(2005) Fig. 1.The exemplar (A) and two of the test objects (B) used in Experiment 1. introducedtheexemplarandtoldthechild“Thisisawug.”Inthetwoactionconditions,theexperimenterthentooktheexemplarandmoveditthefulllengthalongthegrass(inthehorizontalcondition)oralongthetree(intheverticalcondition).Thechildwasthenaskedtotakethe“wug”andmoveitinthesamewayastheexperimenter.Inbothactionconditions,thechildheldtheobjectinthesameway,withthepalminaverticalorientationwiththefingerswrappedaroundtheexemplar.Theexperimenterhelpedthechildifnecessaryandmadesurethechildmovedtheobjectthefulllengthbackandforthacrossthepaththreetimes(i.e.,sixlengths).Theexperimenterthenremovedtheexemplarandplacedtwotestobjectsonthetable,askingthe child, “Where is the wug? Show me the wug.” No feedback was provided on test trials.Theproceduresintheno-actionhorizontalandverticalconditionswerethesame,excepttheexperimenterneveraskedthechildtomovetheobject.Aftertheinitialdemonstrationoftheaction,theexperimenteraskedthechildtowatchastheexperimentermovedtheexemplarbackandforththreemoretimes.Intheno-action–no-movementcondition,theexperimentermerelyintroduced the exemplar and set it on the table, naming it several times.Theexemplarwasremovedfromsightaftertheseprimingevents,andaseriesof12testtrials(presentedinoneoftworandomorders)followed.Theseincluded4eachofthreecontrasts:H1versusV1,H2versusV2,andH3versusV3.Halfwaythroughtesting,theexperimenterre-peatedtheprimingeventappropriatetotheconditionandthenremovedtheexemplarbefore2.1.4.Yes–no procedureTheyes–noprocedurealsobeganwithwarm-uptrials.Thechildwaspresentedwithonewarm-upobjectatatimeandthenaskedifitcouldbecalledbyacertainname.Thesetrialswerearrangedsothattherequiredanswercouldbea“yes”ora“no.”Forexample,theexperi-mentermightpresentthechildwiththeappleandask,“Isthisanapple?”(thechildshouldsay“yes”),andonthenexttrial,theexperimentermightpresentthechildwiththebunnyandask,“Isthisacar?”(thechildshouldsay“no”).Childrenweregivenfeedback.Afterachildansweredthreeconsecutivequestionscorrectly,themainexperimentbeganwiththeprimingevent.Theprocedurewasidenticaltothatoftheforced-choiceversionexceptduringthetesttrials.Underthisprocedure,childrenwerequeriedabouteachofthesixtestobjectsindividually,withtheexperimenteraskingofeachtestobject,“Isthisawug?”Nofeedbackwasprovided.Eachtestobjectwasqueriedthreetimesfor18totaltrials.Childrenwererequiredtopassthewarm-upprocedurein10trialsandtoalsoanswer“no”atleasttwiceduringthetestingtrials. Four children failed to meet these criteria and were replaced.Each version of the experiment took approximately 30 min.2.2.Results and discussionFig.2showstheproportionoftimeschildrenchosethehorizontallyextendedtestobjectovertheverticallyextendedoneintheforced-choicetask.Inthetwoactionconditions,childrenchosethehorizontallyextendedtestobjectwhentheyhadmovedtheexemplaralongahorizontalpathbutchosetheverticallyextendedonewhentheyhadmovedtheexemplaralongaverticalpath,asindicatedbyasignificantmaineffectofdirection,(1,22)=53.01,.001.Therewasalsoaninteractionbetweendirectionandtestcontrast,(2,44)=3.44,L. B. Smith/Cognitive Science29(2005) .05.Notsurprisingly,perhaps,inthisforced-choiceprocedurechildrenhadsomedifficultychoosingbetweentwoalternativesthatdifferedlittleintheirhorizontalandverticalextentfromeachotherandthatalsodifferedlittlefromtheexemplar.Still,theeffectofactionisreliableforallcontrasts(Tukeyhonestlysignificantdifferencetest,.05).Children’schoicesinthethreeno-actionconditionsdidnotdifferfromchanceorfromoneanother.Thus,inthetwoconditionsinwhichchildrenwatchedtheexperimentermovetheexemplarhorizontallyorvertically,therewerenosystematicdirectionaleffects.Manyindividualchildrenchoseinadirectionallyconsistentway(i.e.,preferringthehorizontallyorverticallyextendedobjects),buttheirpreferenceswerenotrelatedtocondition.Theforced-choiceprocedureforceschildrentochoosebetweentwoalternatives—bothofwhichorneitherofwhich—mightbeseenasgoodcategorymembers.Thetaskthusprovidesameasureoftherelativeimportanceofverticalversushorizontalextension,andtheresultsshowthattherelativeimportancedependsonthechild’sactions.Theyes–noprocedure,incontrast,assessestheperceivedsimilarityofeachtestobjecttotheexemplarsingly,andthusthechildisnotforcedbytaskstructuretoincludeorexcludeanytestobjectfromthecategory.Fig.3showstheproportionof“yes”responses(thetestobjectisawug)inthistask.Again,onlyintheactionconditionswasthereadirectionalef-fect.Morespecifically,intheactionconditionstherewasaninteractionbetweenthedirec-tionoftheactionandmagnitudeofvertical–horizontalextent,(2,44)=9.48,.001,be-tweenthehorizontal–verticalextensionofthetestobjectandmagnitudeofthatextent,44)=8,72,.001,andbetweendirection,horizontal–verticalextension,andmagnitudeofthatextent,(2,44)=10.32,.001.Thatis,childreninthehorizontalactionconditionsaid“yes”moretothehorizontallythantotheverticallyextendedtestobjects,andchildrenintheverticalconditionsaid“yes”moretotheverticallyextendedthantothehorizontallyextendedtestobjects,andtheseeffectsweregreaterforthetestobjectsofintermediatesimi-laritytotheexemplar(Tukeyhonestlysignificantdifferencetest,.05).Inbrief,childrenaremorelikelytoincludetestobjectsthatdifferbysmallerthanbygreateramountsfromtheexemplarinthecategory,butdirectionalactionbroadensthemagnitudeofallowablewithin-categorydifferencesthatarecongruentwiththedirectionofaction.Intheno-actionconditions,therewasonlyamaineffectofmagnitudeofvertical–horizontalextent,(2,66)=8.97,.001.Thetwotaskproceduresimposeverydifferenttaskdemands(choosingthe“better”oftwoalternativesversusmakingindividualdecisionsabouttheinclusionornoninclusionofanobject).Yetbothindicatethatactionstronglyinfluencestherangeofshapestakenassimilar.Theseeffectsobtainonlyintheconditionsinwhichchildrenactonobjectsandnotintheconditionsinwhichtheywatchtheexperimenterperformthesameactions.Thisfactconstrainspossible explanations as will be considered in the General Discussion.3.Experiment 2Experiment2demonstratesthegeneralityofthephenomenonbyexaminingthecaseofsymmetry.L. B. Smith/Cognitive Science29(2005) 3.1.Method3.1.1.ParticipantsChildren(60girls,60boys)betweentheagesof24and34months,(meanage30months)participated,halfinaforced-choiceversionandhalfinayes–noversion.NonehadparticipatedinExperiment1.Threechildrenfailedtomeetthecriteriaforinclusionintheyes–no3.1.2.Stimuli and procedureAllaspectsofthedesignandprocedurewereidenticaltoExperiment1,withtheexceptionofthestimuliandtheactions.Fig.4showstheexemplarandtestobjects.Theexemplarwas16cmlongatitsbaseand3cmdeepwithtwo“bumps”oneachend,one6.5cmhigh,theother4.5cmhigh,andaheightof2.5cmatthemidpointbetweenthem.Thetestobjectswerearrangedinpairsorderedbytheiroverallsimilaritytotheexemplar:A1versusS1,A2versusS2,andA3versusS3.TheAmemberwaslesssymmetric,andthememberwasmoresymmetricthantheexemplar. All stimuli were carved from Styrofoam and painted purple.Fortheasymmetricaction,theexemplarwasheldinonehandbytheendwiththesmallerbumpandmovedinalinearmotionforwardandbackward.Forthesymmetricaction,theex-emplarwasheldbytwohands,oneoneachbumpandrotatedbackandforth.Intheno-actionstationarycondition,theexperimenterintroducedtheexemplarbyholdingitflatonthepalmL. B. Smith/Cognitive Science29(2005) Fig. 4.The exemplar and the 6 test objects used in Experiment 2. ofonehandandthensettingitonthetable.Allotheraspectsoftheprocedureanddesignwere3.2.Results and discussionFig.5showstheresultsfromtheforced-choiceprocedure;again,performinganactionbutnotwatchinganactionbeingperformedalteredchildren’sjudgmentsofshape.Intheactionconditions,childrenfavoredthesymmetricobjectwhentheyhadmovedtheexemplarsymmetricallyaboutitscenterbutnotwhentheyhadactedontheexemplarinawaythattreatedthetwosidesdifferently,(1,22)=38.30,.001.Thepreferencefortheshapecongruentwiththeactionalsoincreasedwiththemagnitudeofstimulusdifferencebetweenthetestobjectsandtheexemplar,(2,44)=3.48,.04.Therewerenodifferencesbetweenthethreeno-actionconditions;inallthreeconditions,therewasapreferenceforthemoresymmetrictestobjectineachpairandthispreferenceincreasedwiththegreatersymmetryandasymmetryofthetestobjectstotheexemplar,(2,44)=5.45,.01.Thisresultmayindicateageneralbaselinepreferenceforsymmetricaloverasymmetricalforms(seealsoRock,1973).Fig.6showstheresultsfromtheyes–noprocedure;againthefindingisthatactionalterschildren’scategorizationsofshape.Intheactionconditions,childrensaid“yes”thenameap-pliesmoretothesymmetrictestobjectswhentheyhadmovedtheexemplarsymmetricallywithtwohands,butsaid“yes”thenameappliesmoretotheasymmetrictestobjectswhentheyhadheldonepartoftheexemplarandmoveditwithonehand,(1,22)=25.93,.001.Childrenalsosaid“yes”tothetestobjectslessdifferentinoverallshapetotheexemplar,44)=13.70,.001.Theno-actionconditionsdidnotdifferfromoneanother;inallthreecon-ditions,childrensaid“yes”mosttoobjectsthatdifferedlessfromtheexemplarinoverallshapethantothosethatdifferedmore,(2,66)=19.78,.001,andtosymmetricmorethanasym-Again,underbothtaskprocedures,actionchangestherangeofshapeschildrenjudgetobemembersofthesamecategory,butwatchingsomeoneelseactonandmovetheobjectsinthesame way does not.4.General DiscussionTheoriesofobjectrecognition(bothstructuralandview-based)areforthemostparttheoriesofstaticobjectrecognition(seeLui&Cooper,2003).Yethowweactonobjectsisintimatelyrelatedtotheirshapes,andperhapsevendefiningofthem.Attheveryleast,theseresultsshowthatactionhasastronginfluenceontherangeofshapes2-year-oldstakeasbeingsimilarandappearstodosobydefiningaxesofelongationandsymmetry.Becauseelongationandsymmetryarecrucialtoshapeprocessingmoregenerally,thisraisesthepossibilitythatactionmoldsprocessesfundamentaltoobjectrecognition.Thecriticalopenquestionisthemechanism that underlies the observed effects. Four possibilities are outlined.L. B. Smith/Cognitive Science29(2005) L. B. Smith/Cognitive Science29(2005) Fig.6.Meanproportion“yes”responsesasafunctionofthesimilarityofthetestobjecttotheexemplarinthetwo directionaswell.Emergingevidencesuggestsexuberantmultimodalconnectionsinyoungbrainsthatareprunedwithdevelopment(Huttenlocher,2002;Johnson&Vecera,1996;Maurer&Mondloch,inpress;Thelen&Smith,1994).Ifthisisso,actionmightwellactivatevisualbrain regions in young children and play a role in the development of those regions.Clearlymoreexperimentsareneededtodisentangleandelucidatepossiblemechanisms.Theimportanceoftheseresults—andthecandidatemechanisms—concernstheirpotentialdevelopmentalconsequencesinthecontextofchildren’severydayandrepeatedinteractionswithobjects.Youngchildrenhabituallyactoncategoriesofobjectsinsystematicways.Theserepeatedactionsmaybeexpectedtochangehowaspectsofshapeareprocessedandremembered.Theimplicationofthisisclear:Ifwedidhabituallyusechairstotamelionsratherthanto sit on, we would represent their shapes differently than we do now.AcknowledgmentsThisresearchwassupportedbyR01HD28675–10fromtheNationalInstitutesofHealth.IthankLorindaCagle,ReginaHildebrand,andEmilyUnikelforthecreationofthestimuliandReferencesBiederman,I.(1987).Recognition-by-components:Atheoryofhumanimageunderstanding.PsychologicalRe-view, 94,Bucher,N.M.,&Palmer,S.E.(1985).Effectsofmotiononperceivedpointingofambiguoustriangles.Perception& Psychophysics, 38,Creem,S.H.,&Proffitt,D.R.(2001).Graspingobjectsbytheirhandles:AnecessaryinteractionbetweencognitionJournal of Experimental Psychology: Human Perception and Performance, 27,Duvdevani-Bar,S.,&Edelman,S.(1999).VisualrecognitionandcategorizationonthebasisofsimilaritiestomulInternational Journal of Computer Vision,33,Edelman,S.,&Intrator,N.(2003).Betterlimitedsystematicityinhandthanstructuraldescriptionsinthebush:ACognitive Science, 27Ellis,R.,&Tucker,M.(2000).Micro-affordance:Thepotentiationofcomponentsofactionbyseenobjects.Journal of Psychology, 91,Faillenot,I.,Toni,I.,Decety,J.,Gregoire,M.,&Jeannerod,M.(1997).Visualpathwaysforobject-orientedactionand object recognition: Functional anatomy with PET.Cerebral Cortex, 7,Gelman,S.A.,&Bloom,P.(2000).YoungchildrenaresensitivetohowanobjectwascreatedwhendecidingwhatCognition, 76,Principles of perceptual learning and development.New York: Academic.The senses considered as perceptual systems.Oxford, England: Houghton Mifflin.Glenberg,A.M.,Robertson,D.A.,Kaschak,M.P.,&Malter,A.J.(2003).EmbodiedmeaningandnegativeprimBehavioral Brain Sciences, 26,Grezes,J.,&Decety,J.(2001).Functionalanatomyofexecution,mentalsimulation,observation,andverbgeneraHuman Brain Mapping, 12,Hummel,J.E.(2001).Complementarysolutionstothebindingprobleminvision:Implicationsforshapeperception and object recognition.Neural binding of space and time.Visual Cognition, 8,Huttenlocher,P.R.(2002).Neuralplasticity:Theeffectsofenvironmentonthedevelopmentofthecerebralcortex.Cambridge, MA: Harvard University Press.L. B. Smith/Cognitive Science29(2005)