《知觉的进化》（塞西莉亚·海斯 & 路德维希·胡贝尔，英文版，“麻省理工”理论生物学教材） 370页

TheEvolutionofCognitionCeciliaHeyesLudwigHuber"Thisimportantcollectionofessaysrepresentsmostmajorcurrentsofpresentthoughtinanimalcognition:fromthemodularityofthemindtoculturalevolution,fromthesearchforepisodicmemoryinanimalstothepropertiesofcausalreasoninginhumans,fromhoneybeestoravens.Aug2000AcrucialreferenceinthisdynamicandrapidlyISBN0262082861evolvingfield."--AlexKacelnik,ProfessorofBehavioural396pp.Ecology,DepartmentofZoology,OxfordUniversity32illus.$52.00((hardback)Inthelastdecade,"evolutionarypsychology"hascometorefer$41.60((hardback)exclusivelytoresearchonhumanmentalityandbehavior,motivatedbyanativistinterpretationofhowevolutionoperates.Thisbookencompassesthebehaviorandmentalityofnonhumanaswellashumananimalsandafullrangeofevolutionaryapproaches.Ratherthanacollectionbyandforthelike-minded,itisadebateabouthowevolutionaryprocesseshaveshapedcognition.Thedebateisdividedintofivesections:Orientations,onthephylogenetic,ecological,andpsychological/comparativeapproachestotheevolutionofcognition;Categorization,onhowvariousanimalsparsetheirenvironments,howtheyrepresentobjectsandeventsandtherelationsamongthem;Causality,onwhetherandinwhatwaysnonhumananimalsrepresentcauseandeffectrelationships;Consciousness,onwhetheritmakessensetotalkabouttheevolutionofconsciousnessandwhetherthephenomenoncanbeinvestigatedempiricallyinnonhumananimals;andCulture,onthecognitiverequirementsfornongenetictransmissionofinformationandtheevolutionaryconsequencesofsuchculturalexchange.TABLEOFCONTENTSSERIESFOREWORDPREFACEIORIENTATIONS1EVOLUTIONARYPSYCHOLOGYINTHEROUNDBYCECILIAHEYES2PSYCHOPHYLOGENESIS:INNOVATIONSANDLIMITATIONSINTHEEVOLUTIONOFCOGNITIONBYLUDWIGHUBER3MODULARITYANDTHEEVOLUTIONOFCOGNITIONBYSARASHETTLEWORTH4COGNITIVEEVOLUTION:APSYCHOLOGICALPERSPECTIVEBYM.E.BITTERMANIICATEGORIZATION 5WHATMUSTBEKNOWNINORDERTOUNDERSTANDIMPRINTING?BYPATRICKBATESON6STIMULUSEQUIVALENCIESTHROUGHDISCRIMINATIONREVERSALSBYJUAND.DELIUS,MASAKOJITSUMORIANDMARTINASIEMANN7ABSTRACTIONANDDISCRIMINATIONBYNICHOLASJ.MACKINTOSH8PRIMATEWORLDSBYKIMSTERELNYIIICAUSALITY9TWOHYPOTHESESABOUTPRIMATECOGNITIONBYMICHAELTOMASELLO10CAUSALCOGNITIONANDGOAL-DIRECTEDACTIONBYANTHONYDICKINSONANDBERNARDW.BALLEINE11CAUSALREASONING,MENTALREHEARSAL,ANDTHEEVOLUTIONOFPRIMATECOGNITIONBYROBINI.M.DUNBAR12CAUSE-EFFECTREASONINGINHUMANSANDANIMALSBYDUANEM.RUMBAUGH,MICHAELJ.BERANANDWILLIAMA.HILLIXIVCONSCIOUSNESS13THEPRIVATIZATIONOFSENSATIONBYNICHOLASHUMPHREY14THESEARCHFORAMENTALRUBICONBYEUANM.MACPHAIL1515DECLARATIVEANDEPISODIC-LIKEMEMORYINANIMALS:PERSONALMUSINGSOFASCRUBJAYBYNICOLAS.CLAYTON,D.P.GRIFFITHSANDANTHONYDICKINSON1616TESTINGINSIGHTINRAVENSBYBERNDHEINRICHVCULTURE17FEEDINGINNOVATIONSANDTHEIRCULTURALTRANSMISSIONINBIRDPOPULATIONSBYLOUISLEFEBVRE18CLIMATE,CULTURE,ANDTHEEVOLUTIONOFCOGNITIONBYPETERJ.RICHERSONANDROBERTBOYD19GOSSIPANDOTHERASPECTSOFLANGUAGEASGROUP-LEVELADAPTATIONSBYDAVIDSLOANWILSON,CAROLYNWILCZYNSKI,ALEXANDRAWELLSANDLAURAWEISERCONTRIBUTORSSPECIESINDEXAUTHORINDEXSUBJECTINDEX PrefaceTheKonradLorenzInstituteforEvolutionandCognitionResearch(KLI)devotedits1998AltenbergWorkshopinTheoreticalBiologyto"TheEvolutionofCogni•tion"—athemethatliesatthecoreoftheKLI"sscientificaims.Theworkshopwasheldexactly25yearsafterpublicationofLorenz"sbookDieRiickseitedesSpiegels.Thisbookfirstappearedin1973(translationpublishedasBehindtheMirrorin1977,London:Methuen),theyearinwhichLorenzretiredandearnedtheNobelPrize.Itsummarizedtheresultsofhislonginquiryintothe"naturalhistoryofhumanknowledge."Thethemesandstructureofthepresentvolumeareintroducedinchapter1.Thebookarosefromthe1998Altenbergworkshop,andcouldalsobesaidtobeabout"thenaturalhistoryofhumanknowledge."However,thesubjectmatterofthisbookisnowmorecommonlyknownas"evolutionarypsychology."Owingtothedistinc•tion,professionalism,andhardworkofthecontributors,thisbookisbetterinte•gratedandmoreauthoritativethanistypicalofconferencevolumes.Itprovidesanoverviewofcontemporaryresearchonthephylogenetic,ontogenetic,andculturalevolutionofcognition.TheworkshopwasheldinthebeautifulLorenzmansionoutsideVienna,whereKonradLorenzgrewupandspentthelatteryearsofhislife.Hishomeisoneofthecradlesofethologyandevolutionaryepistemology.ThesessionstookplaceinLorenz"sspaciouslibrary,andtheconversationcontinued,throughlunchandwellintotheevenings,inthe"winterdiningroom,""thehall,"andunderthenuttreeinthegardens.ItisapleasuretoacknowledgetheassistanceoftheKLI,inparticularGeneralSecretaryDr.AdolfHeschl,ChairmanProf.GerdMiiller,andtheformerchairmanandfounderoftheKLI,Prof.RupertRiedl.Itwashewhoinvitedustoorganizetheworkshopandtoeditthebook,andwhoprovidedinvaluablesupportandencouragementthroughouttheprocess.WewishtoexpressoursinceregratitudetothesponsorsoftheKLIfortheirgenerousfinancialsupport.WearealsoverygratefulforthehelpoftheKLI"ssecretary,UlrikeKiihn;LH"ssecretary,Mag.MariaNausch,whoassistedintheeditorialwork;andCH"scollabo•rators,FionaCampbellandElizabethRay,whoperfectedtheEnglishinthemanu•scriptsfromGerman-speakingcontributors.OurparticularthanksgotothestudentsoftheDepartmentofTheoreticalBiologyattheUniversityofVienna,whohelpedoutwithsomanypracticaldetailsoftheworkshop;andtotheBauerfamilyofHotelMarienhof,whomadetheparticipantswelcomeintheViennaWoods.Finally,ithasbeenapleasuretoworkwithMichaelRutterofTheMITPress,whoseenthusiasmfortheseriesandthepresentvolumeprovidedcontinuousencouragement. 1EvolutionaryPsychologyintheRoundCeciliaHeyesWhenIfirstencounteredtheterm"evolutionarypsychology,"Ithoughtitreferredtothestudyofhowmindandbehaviorhaveevolved.ButIwasmistaken.Inthelastdecade,evolutionarypsychologyhascometoreferexclusivelytoresearchonhumanmentalityandbehavior,motivatedbyaveryspecific,nativist-adaptationistinter•pretationofhowevolutionoperates(e.g.,CosmidesandTooby,1994;Buss,1999;DalyandWilson,1999).Thisisastrange,anthropocentricusage,akintoidentifyinghumanbiologywith"biology"generally,ordescribinggeographyas"astronomy."Thisbookisaboutevolutionarypsychologymorebroadlyandmoretransparentlyconstrued;aboutevolutionarypsychology"intheround."Itencompassesthebehaviorandmentalityofnonhumanaswellashumananimals,andafullrangeofcontemporaryevolutionaryapproaches.Ratherthanacampaignvolumebyandforthelike-minded,itisadebateamongauthoritativeresearchersaboutthewaysinwhichevolutionaryprocesseshaveshapedcognition.Thedebateispresentedunderfivebroadsectionheadings:Orientations,Catego•rization,Causality,Consciousness,andCulture.Inthefirstofthese,thechaptersbyHuber,Shettleworth,andBittermanprovidegeneralargumentsinfavorofdistinctconceptualandmethodologicalapproachestoinvestigatingtheevolutionofcogni•tion:phylogenetic,ecological,andpsychological/comparative.Ineachoftheremain•ingchapters,theauthor(s)adoptoneoracombinationoftheseapproachesinaddressingaspecificissue,orsetofissues,relatingtotheevolutionofcognition.Thechapterson"Categorization"areconcernedwithhowvariousanimalsparsetheirenvironments,howtheythinkabout,orrepresent,objectsandeventsandtherelationsamongthem.Thosechaptersunder"Causality"focusonaparticularkindofrelationship,thatofcauseandeffect,askingwhichnonhumananimals,ifany,representthiskindofrelation,andhowtheydoit.Thediscussionsof"Conscious•ness"considerwhetheritmakessensetotalkabouttheevolutionofconsciousness,andhow,ifatall,thisphenomenoncanbeinvestigatedinnonhumananimals.Thefinalsection,on"Culture,"examinesthecognitiverequirementsfornongenetictransmis•sionofinformation,andtheevolutionaryconsequencesofsuchculturalexchange.Tohelpreadersfollowthedebate,eachcontributioncross-referencesotherchap•tersandconcludeswithanabstract-likesummary.Inaddition,eachsectionbeginswithashortoverviewidentifyingthemainpointsofagreementanddisagreementamongthecontributorstothatsection,andamongtheirviewsandthoseexpressedelsewhereinthevolume.Thepurposeofthisintroductorychapterisbothtointegrateandtopropose.Itaddressesthreeelementaryquestionsabouttheevolutionofcognition,surveysthe 4CeciliaHeyesvariousanswersofferedbythecontributorstothisvolume,and,attheend,combinestheseinageneralhypothesisabouttheevolutionofcognition.Thequestionsare:Whatiscognition?Whatisthedifferenceamong"different"cognitiveprocesses?Whatmakesresearchoncognition"evolutionary"?WhatIsCognition?Fortunately,aroughandreadydefinitionissufficienttosupportinterestingresearchontheevolutionofcognition,andforthepresentpurposesIneedonlyoutlinewhatmostofthecontributorstothisvolumeassumeaboutthenatureofcognition,andtoidentifypointsofpotentiallyconfusingcontention.Mostoftheauthorswouldprobablyagreethatcognitivestatesandprocessesare(1)theoreticalentities,which(2)provideafunctionalcharacterizationofoperationsofthecentralnervoussystem,(3)mayormaynotbeobjectsofconsciousawareness,(4)receiveinputsfromothercognitivestatesandprocessesandfromperception,and(5)haveoutputstoothercognitivestatesandprocessesandtobehavior.Regardingcognitivestatesandprocessesastheoreticalentitiesthatcausebe•haviordistinguishescognitivepsychologyfrommostvarietiesofbehaviorism,andemphasisesthat,unlikebehaviororneuraltissue,theycannotbeobserveddirectly.Hypothesesaboutcognitioncanbeevaluatedonlybytestingtheirpredictionsregard•ingtheeffectsofvariousenvironmentalmanipulationsonbehavior.Thesetheoreticalentitiesaresaidtoprovideafunctionalcharacterizationofthecentralnervoussystem(CNS)toflagthefactthatmostofcontributorstothisvolumeassumethatthesamecognitiveprocesscouldbeimplementedorinstantiatedinavarietyofdifferentneu-roanatomicalstructuresorneurophysiologicalprocesses.Inotherwords,thischar•acterizationofcognitionismaterialist,butitdoesnotassumeasimpleone-to-onemappingbetweencognitiveandneuralstatesandprocesses.Similarly,althoughtheforegoingcharacterizationofcognitionallowsthatsomecognitiveprocessesmaybeconscious,andthattheirsubjectivestatusmaydependontheirfunctionalrole,itdoesnotidentifycognitiveprocesseswithconsciousprocesses;consciousnessplaysnopartinthedefinitionofwhatisandisnotcognitive.DickinsonandBalleine,Bitterman,andShettlewortharetheonlycontributorstothisbookwhoclearlydepartfromtheusageoutlinedabove.DickinsonandBalleineprefertoreservetheterm"cognitive"forprocessesthatsupportgoal-directedbehavior,excluding,forexample,associativelearningfromthecognitivedomain.BittermanandShettleworth,ontheotherhand,favormoreinclusivedefinitionsofcognition.Bittermanequatesitwith""knowing"intheclassicalsenseoftheterm, EvolutionaryPsychologyintheRound5encompassingperception,learning,andunderstanding,"whereasforShettleworthcognitionsubsumes"allmechanismsthatinvertebratesandvertebrateshavefortakingininformationthroughthesenses,retainingit,andusingittoadjustbehaviortolocalconditions."Thesecharacterizationsareincompatiblewith(4)and(5)above,whichimplythatcognitiveprocessesaredistinctfromperceptualprocessesthataredirectlyinvolvedindealingwithsensoryinputandmotorprocessesresponsibleforpreparingeffectormovements.Thesedissentingvoiceshighlightatruism:itmatterslittlehowwelabelourdistinctions,butwewouldbeunwisetoletthemgetlost.Points1-5abovecircumscribeasetofpropertiesoftheCNSthatarephysicallyrelatedto,butconceptuallydistinctfrom,itsneurobiologicalandbehavioralprop•erties.Whetherwelabeltheseproperties"cognitive,""mental,"or"intelligent"isunimportantinitself,butifweweretolosesightoftheirdistinctiveness,therewouldnotonlybeconfusion,butresearchontheevolutionofcognition/mentality/intelligencewouldmoveoutsidethedomainofcontemporarypsychology.Itwouldbereducedtoanexaminationofthewayinwhichevolutionhasaffectednervoussystemsandmotorphysiology.Similarly,itdoesn"treallymatterwhetherwecallassociativelearningorperceptualprocessing"cognitive"or"noncognitiveinformationprocessing,"butweshouldbealerttothepossibilitiesthattheprocessesoneithersideofthesedividesshowdiffer•entevolutionarypatterns.Forexample,overevolutionarytime,associativelearningmaybemoreconservative,andperceptualprocessingmorelabile,than(other)cog•nitiveprocesses.WhatAretheDifferencesAmong"Different"CognitiveProcesses?CarvingCognitionThefirstquestionaskedhowwedistinguishcognitivefromotherprocesses;thesecondexamineshowwedistinguishonekindofcognitiveprocessfromanother.Thecapacitytodothis,thepossessionofconceptualknivesthatwillcarvecognitionrationallyandreliablyintodistinctpieces,isespeciallyimportantinthecontextofanevolutionaryanalysis.Atthemostgenerallevel,evolutionaryanalysisusessyn•chronouspatternsofsimilarityanddiversitytoinferhistoricalcontinuityandchange.Clearly,thiscannotbeachievedinthecaseofcognitionunlesswecanworkoutwhereonekindofcognitionstopsandanotherbegins.Allofthecontributorstothisvolumecarvecognitionintodifferenttypes,butfewcommentontheknivestheyareusing,onwhattheyconsidertobethedifferencesamongdifferentcognitiveprocesses.Ifwefirstconsiderthepieces,wefindthatthey 6CeciliaHeyescomeinarangeofshapesandsizes.Atoneextreme,Shettleworthmentionsatleast14typesofcognition:spatialmemory,circadiantiming,intervaltiming,deadreck•oning,landmarkuse,imprinting,songlearning,motorimitation,associativelearning(andcomponentsthereof),socialintelligence,theoryofmind(andcomponentsthereof),language,reasoningaboutsocialobligations,andconsciousness.Attheotherextreme,severalauthorscarvethecognitivepieintoconsciousandunconsciousprocesses(Claytonetal.,Heinrich,Humphrey,Macphail),and/orintotwoorthreepieces,oneofwhichcorrespondslooselywithwhatShettleworthcallsassociativelearning.Thus,Bitterman,Delius,andHuberdistinguishassociativelearningfromconceptualthought(and,inthecaseofHuber,fromlanguage);DickinsonandBalleine,Dunbar,Macphail,Rumbaugh,andTomasellocontrastitwiththerepre•sentationofintentionalrelations(and,inthelattercase,withrepresentationofrela•tionsmoregenerally);Mackintoshleavesoutrepresentationofintentionalrelationsandcontrastsassociativelearning,basedoninvariantfeaturedetection,withrepre•sentationofrelationsgenerally;andBatesonopposes"learninginvolvingexternalreward"withrecognitionlearning.DichotomiesthatapparentlyhavenothingtodowithassociativelearningarementionedbyClaytonetal.(rememberingfactsvs.personalexperiences)andSterelny(representingmentalstatesvs.behavior);listsofcognitiveprocesses,similarto,butshorterthan,Shettleworth"scanbefoundinthechaptersbyLefebvre(spatialmemory,imprinting,songlearning,[associative?]learning,andimitation,butnotsociallearning)andbyRichersonandBoyd(decisionmaking,[associative?]learning,sociallearning,imitation,andlanguage).Thisrangeofmethodsofcarvingcognitionisrepresentativeofcontemporaryresearchontheevolutionofcognition,andsomeofthediversityisalmostcertainlyduetosubstantivedisagreementsaboutevidence.Fromascientificperspective,thesearetheinterestingdisagreements,theunresolvedempiricalquestions,buttheyaredifficulttoisolatefromthediversityduetotheuseofdifferentknives,differentprin•ciplesofclassification.Forexample,circadiantiming,intervaltiming,deadreckon•ing,andlandmarkusefallwithinShettleworth"sinclusivedefinitionofcognition,butothersmayregardthemasnoncognitive,perceptualprocesses.Inthisexample,itisthecutbetweencognitiveandnoncognitiveprocessesthatisatissue,butsomeoftheremainingvariationmaybeduetotheuseofdifferentknivestodistinguishamongprocessesthatareagreedtobecognitive.What/WhenandHowRulesExtrapolatingfromtheshapesandsizesoftheirpiecesofcognition,andfromocca•sionalcommentsaboutprinciplesofclassification,itseemsthatthecontributorstothisvolume,andevolutionarypsychologistsmoregenerally,areusingtwosortsof EvolutionaryPsychologyintheRound7bladestocarvecognition:"What/when"and"how"rules.What/whenrulesdistin•guishtypesofcognitionaccordingtotheircontent(dennedbyinputstooroutputsfromcognitiveprocesses)and/orthetimeinontogenywhentheytypicallyoperate.Howrules,bycontrast,distinguishtypesofcognitionintermsoftheabstractprin•ciples(moreorlessformallyspecified)thatcharacterizethewayinwhichcontentinformationisprocessed.Whenbirdsonglearningisidentifiedasadistinctvarietyofcognition,itispri•marilyonthebasisofawhat/whenrulesomethinglike,"Conspecificsong,firstseason"(e.g.,CatchpoleandSlater,1995).Itspecifiesthecontentofwhatisusuallylearned(species-typicalsong)andwheninontogenythislearningusuallytakesplace.Describingacognitiveprocessas"domainspecific"(e.g.,Shettleworth,thisvolume),impliesthatitisdistinctatleastintermsofitswhat/whenrules.AnexampleofaformallyspecifiedhowruleistheRescorla-Wagnerequation(BushandMosteller,1951;RescorlaandWagner,1972;Bitterman,thisvolume).HowrulesarewhatSherryandSchacter(1987)describeas"rulesofoperation,"Bitterman(thisvolume)as"equations,"andBateson(thisvolume)as"designrules."ExamplesCloserexaminationoftwoexamplesofbehaviorwill,Ihope,clarifythedistinctionbetweenwhat/whenandhowrules.Ifthefirst,snakefearlearninginrhesusmonkeys(e.g.,CookandMineka,1990),isbasedonadistinctivetypeofcognition,thenitisdistinctiveintermsofitswhat/whenrules.Thesecond,same-differentcategorisationbyAlextheparrot(Pepperberg,1987),seemstobebasedonacognitiveprocesswithdistinctivehowrules.SnakeFearCookandMineka(1987,1989,1990)haveshownthatrhesusmonkeysacquirefearofsnakesmorereadilythanfearofflowersthroughexposuretoacon-specificbehavingfearfullyinthepresenceofthetargetstimuli,thatis,snakesorflowers.Tocheckwhetherthiseffectisduetothecontentoridentityofthestimuli(snakesvs.flowers),ratherthantothedifferentialsalienceofthesnakeandflowerstimuliemployed(e.g.,differencesincolororbrightness),theydidanexperimentinwhichsnakeandflowerstimuliwerepairedwithfoodratherthanfear(CookandMineka,1990,experiment3).Theypredictedthatifthefeareffectwasduetodiffer•entialsalience,themonkeyswouldstilllearnaboutthesnakesmorereadilythanabouttheflowers,butthatifitwasduetotheidentityofthestimuli,theywould,ifanything,learnmoreslowlythatthesnakessignaledapositiveevent.Ineachtrialinthisexperiment,amonkeywasshownoneoffourpairsofstimulionavideoscreen.Ifitreachedtowardoneofthestimuli(+)itwasrewardedwith 8CeciliaHeyesfood;reachingtowardtheotherwasnotrewarded(-).Thus,themonkeyshadfourdiscriminationproblems:coiledsnake+/redsquare-(snakepositive);longsnake-/redsquare+(snakenegative);chrysanthemums+/bluediamond-(flowerpositive);silkflowers-/bluediamond+(flowernegative).Theresultsindicatedthatthemonkeyssolvedtheflowerproblemsatleastasfastasthesnakeproblems,andinsomecasestheflowerproblemsweresolvedmorequickly.Therearethreethingstonoteaboutthisexample.First,itdoesnotshowthatthecognitiveprocessunderlyingsnakefearinmonkeysdiffersintermsofitshowrulesfromthekindofassociativelearningmediating(other)Pavlovianconditioning.Indeed,snakefearacquisitionissubjecttotheovershadowingandlatentinhibitioneffects(MinekaandCook,1986;CookandMineka,1987;Heyes1994)thatarecharacteristicofPavlovianconditioningmoregenerally,andwhichhaveplayedakeyroleintheformulationofitshowrules.Second,snakefearacquisitionisadistinctformofassociativelearningintermsofitswhat/whenrulesonlyifitisthecontent,notthesalience,ofthesnakestimulithatisresponsibleforfasterlearning.Thus,notallvariationsinlearningthataretypicallydescribedas"quantitative"ratherthan"qualitative,"orasbeingduetochangesin"constants"ratherthan"equations"(Bitterman,thisvolume),provideevidenceofcognitionwithdistinctivewhat/whenrules.Moregenerally,thewhat/whenvs.howdistinctionisnotequivalenttothequantitativevs.qualitativedistinction.Finally,thesnakefearexampleillustrateshowdifficultitcanbetofindoutwhetherdifferentratesoflearningaboutstimuliareduetotheidentity,ratherthanthesalienceofthestimuli.EventheresultsofCookandMineka"s(1990)subtleexperimentdonotshowthisconclusivelybecause(1)thesnakeandflowerstimuliinthisstudywerenotidenticaltothoseusedtotestobservationalconditioningoffear,and(2)themonkeyswererequiredtodiscriminatesnakesandflowersfromdifferent,arbitrarystimuli.Itmayhavebeenmoredifficulttodiscriminatetheredsquarefromthesnakestimulithantodiscriminatethebluediamondfromtheflowerstimuliand,ifthiswasthecase,themonkeysmighthavesolvedthesnakeproblemsmoreslowlyevenifthesnakestimuliweremoresalientthantheflowerstimuli.AlextheParrotPepperberg"sAfricanGreyparrot,Alex,appearstobeabletorep•resentrelationsamongobjectsinwhatcouldbedescribed,forthewantofabetterword,asan"abstract"way(Pepperberg,1987;Mackintosh,thisvolume).Inthetrainingphaseoftherelevantexperiment,Alexwasshownpairsofobjectsvaryingincolor,material,and/orshape(e.g.,aredwoodentriangleandabluewoodenoval),andasked,inspokenEnglish,"Whatsame?"or"Whatdifferent?"Intheformercase EvolutionaryPsychologyintheRound9hewasrewardedfornamingadimensiononwhichthestimulihadacommonattri•bute(material),andinthelatterfornamingadimensiononwhichtheyhaddifferentattributes(e.g.,colororshape).Alexnotonlysucceededinlearningthesediscrim•inations,butwasaccurateon82percentoftransfertrialsinvolvingnovelobjects(e.g.,agraywoodencubeandagraywoollenball).TheprocessformingAlex"srepresentationsofsimilarityanddifference(butnotnecessarilythatlinkingtheserepresentationswiththeexperimenter"squestionsandwithreward)seemstohavehowrulesdistinctfromthosethatcharacterizePavlovianconditioning.Theseequations(seeDickinson,1980forasurvey)cannotfabricatefromabsolutevaluesofstimuli(e.g.,redness)arepresentationthatdoesnotfunc•tionaccordingtoitsabsolutevalues.Thus,whenAlexisshowntworedobjects,Pavlovianprocesseswouldallowhimtorepresentdouble-redness,butnotinawaythathespontaneously,withoutfurthertraining,treatsasequivalenttodouble-greenness.However,thereisnoreasontosupposethattheprocessesmediatingAlex"srepresentationofrelationsaredistinctiveintermsoftheirwhat/whenrules,noevidencethatheisabletorepresentrelationsamongonlyasubsetofthestimulihecanperceive,orthathehasusedthiscapacitymoreduringacertainphaseofhisontogenythanatothertimes.TheforegoingexperimentinvolvedarbitrarystimuliandwasconductedwhenAlexwasalreadymature.Inbothoftheseexamplesandinmanyothers,associativelearningofthekindthatmediatesPavlovianconditioningfeaturesasakindofbenchmark,asthatfromwhichpotentiallydistinctcognitiveprocessesareshowntodiffer.Thereareprobablytwosoundreasonsforthis,aswellasathirdthatismorecontentious.Thefirstreasonispragmatic:thehowrulesofassociativelearningarerelativelywellspecified,andthereforeinmanycasestheonlyevidencecurrentlyavailablethatdifferenthowrulesunderlieabehavior,isthatthebehaviorinquestioncannotbeexplainedwithreferencetoassociativelearning.Thealternativeruleshavenotbeenformulated.Second,thereisevidencetosuggestthatthecapacityforPavlovianassociativelearningispresentinaverybroadrangeofvertebrateandinvertebratespecies(e.g.,Bitterman,thisvolume;Mackintosh,thisvolume;Macphail,thisvolume),andthereforeitisanaturalcontendertoexplainabroadrangeofbehaviors.Third,itissometimesclaimedthatitismore"parsimonious"toattributebehaviortoassociativelearningthantoanalternativecognitiveprocess(e.g.,Macphail,1985).Thisreasonisproblematicif,goingbeyondthesecond,itassumesthatassociativemechanismsarenecessarilysimplerthannonassociativeprocesses,orthatevolutionissoconservativethatanybehavioraladaptationthatcanbeachievedbyassociativelearningwillbesoachieved(Sober,1998). 10CeciliaHeyesInferringHowfromWhat/WhenMostofthevarietiesofcognitiondiscussedinthisbookseem,onthesurface,tobecircumscribedbytheirwhat/whenrules.Thenamestheyaregivenidentifyacategoryofenvironmentalinput(e.g.,spatialmemory,circadiantiming,intervaltiming,landmarkuse,songlearning,sociallearning,socialintelligence,reasoningaboutsocialobligations,language),ofcognitiveproduct(e.g.,representationsofrelations,intentionalrelations,behavior,andmentalstates;rememberingfactsandpersonalexperiences;decisionmaking;recognitionlearning),orofbehavioralproduct(e.g.,imprinting,imitation,language).Possibleexceptionsareassociativelearning,con•ceptualthought,theoryofmind,andconsciousness,butevensomeofthesecanbeconstruedascharacterizingtypesofinformationprocessedratherthanprocessingoperations.Thiswayoflabelingtypesofcognitionmaygivetheimpressionthatwhat/whenrulesareconsideredmoreimportantthanhowrules,butthereverseistrue.Thevastmajorityofcontributorstothisvolume,andtoresearchontheevolutionofcognitiongenerally,usedifferencesinwhat/whenrulesasmarkersforputativedifferencesinhowrules,andconsiderdistinctionsofthelatterkindtobeofprimarysignificanceinevolutionaryanalysis.Shettleworthmakesthispriorityveryclear:"Whendistinctclassesofinput(domains)arecomputedonindistinctwaysasinferredfrombehavior,wehaveadistinctmentalmoduleormemorysystem.Computationaldis•tinctivenessistheprimarycriterionforcognitivemodularity."Othercontributorsarelessexplicit,anddonotusethelanguageofmodularity,buttheyseemtohaveacommonpurpose.Forexample,itisunlikelythatTomasellowoulddistinguish"intentional/causalcognition"fromunderstandingtherelationshipbetweenone"sownactionsandtheiroutcomesifhebelievedthatthesetwodifferonlyintermsofwhatisunderstood,andnotwithrespecttohowthatunderstandingisachieved.Inadditiontobeingusedasmarkersfordistinctivehowrules,what/whenrulesaresometimesusedtoinfertheexistenceofdistinctivehowrules.Awhat/whenruleactsasasimplemarkerif,ofthemanyslicesintowhichcognitioncouldbecutbychar•acterizingitsinputsorproducts,aresearcherdelineatesonlythosetypesthatheorshebelieves,onthebasisofindependentevidence,alsotobedistinctiveintermsoftheirhowrules.Bycontrast,adifferenceinwhat/whenrulesisusedtoinferadiffer•enceinhowruleswhentheformerisitselftreatedasevidenceofthelatter.Thechaptersinthisbooksuggestthatdifferencesinhowrulescansometimesbeinferredfromdifferencesinthekindofwhat/whenrulesthatcircumscribecognitiveproducts(e.g.,understandingcausality,representingrelations,representinginten•tionalrelations),butrarelyifeverfromwhat/whenrulesthatcharacterizeenvi- EvolutionaryPsychologyintheRound11ronmentalinputs(e.g.,sociallearning)orbehavioralproducts(e.g.,imprinting).Intheformercase,itrequirespainstakingempiricalwork(e.g.,Delius,thisvolume;DickinsonandBalleine,thisvolume;Mackintosh,thisvolume)toestablishthatthecognitiveproductisreallyofthespecifiedkind—forexample,thattheanimalsinquestionarereallyrepresentingcausality,relations,orintentionalrelations—but,ifthisinformationcanbesecured,inconjunctionwithknowledgeofinputs,itmayprovidethebasisforinferringthepresenceofdistinctivehowrules.Thisispossiblebecauseinprincipleonecanworkoutwhichhowrulescould,andwhichcouldnot,generatespecifiedoutputsfromspecifiedinputs.Incontrast,atleastthreechaptersinthisvolume(Bateson,Claytonetal.,andLefebvre)makeitclearthat,astheyarecurrentlycircumscribed,howrulesdonotcoincidewithbehavioralproductsandenvironmentalinputsinawaythatallowstheformertobeinferredreliablyfromthelatter.Bateson"smodelofimprintingimpliesthatthistypeofcognitiondiffersfrom(other)recognitionlearningintermsofitswhat/whenrules,butnotintermsofitshowrules.TheexperimentsreportedbyClaytonetal.raisetheintriguingpossibilitythatspatialmemoryinfood-storingbirds,althoughitisdistinctiveintermsofwhatisremembered(cachelocations)andwhenencodingtakesplace(duringseasonalgluts),occursviathesamehowrulesasepisodicmemoryinhumans.Similarly,Lefebvre"sdemonstrationthatsociallearningabilitycovarieswith(other)learningabilitysuggeststhat,atleastinaviantaxa,thesetwomaydifferintermsofwhethertheirenvironmentalinputsareorarenotfromsocialinteractants,butnotintheirhowrules.Covariationbetweenneuralsubstratesandbehavioralproductorenvironmentalinputwhat/whenrulesiscommonlyinterpretedassupportforthesuggestionthattheformercharacterizeimportantdifferencesamongcognitiveprocesses.Forexample,Lefebvre(thisvolume)arguesthattheassociationofdistinctiveneuralsubstrateswithspatialmemoryinfood-storingbirds(hippocampus),parentalimprinting(leftintermediatemedialhyperstriatumventrale),andbirdsonglearning(highvocalcenter)contributestomakingthesewhat/whentypesofcognitionmoredistinctivethansociallearning.Thisisundoubtedlytrue,butnotbecausethediscoveryofadistinctiveneuralmechanismnecessarilyindicatesthatwhat/whenrulesarecoinci•dentwithhowrules.Thiswouldbethecaseonlyifweassumedthatdifferentneuro•anatomiesstructuresnecessarilyimplementdifferentcognitiverulesofoperation,andthatdifferencesinwhat/whenrulesalonecouldnotbeassociatedwithdistinctiveneuralsubstrates;thisassumptiondoesnotappeartobewarranted.Forexample,food-storingbirdsmayhavelargerhippocampithanrelated,nonstoringspecies(e.g.,ClaytonandKrebs,1994;Clayton,1996),notbecausethespatialmemoryofstoringspeciesoperatesaccordingtodistinctivehowrules,butbecauseinthesebirdsa 12CeciliaHeyessystemusinghowrulescommontostoringandnonstoringspeciesprocessesalargevolumeofinformationaboutthelocationsoffoodcaches.Covariationbetweenwhat/whenrulesandneuralmechanismsissignificant,notasashortcuttodiscov•eringdiversityinhowrules,butinitsownright;asanindicatoroftheeffectsofevolutiononwhat/whenrulesofcognition.Addressingthequestion"Whatisdifferentabout"different"cognitiveprocesses?"Ihavesuggestedthatthecontributorstothisvolumeuseacombinationofwhat/whenandhowrulestodistinguishtypesofcognition,andthatmostorallofthemconsiderdistinctionsbasedonhowrules,onthewayinwhichinformationisprocessed,tobeprimary,andusedifferencesinwhat/whenrulesasmarkersforthesedistinctions,orasabasisforinferringhowrulediversity.Thus,atroot,theauthorsthatpostulatedifferentrangesofcognitiveprocessesdisagreeaboutthevarietyofhowrulesfoundintheanimalkingdom,anditisunlikelythatthesedisagreementscanberesolvedexceptbygeneratinghypotheticalhowrulesforvariouscategoriesofbehaviorandtestingthemempiricallyagainstothersuchrules.Atpresent,researchofthiskindtypicallyexamineswhetherbehaviorcanorcannotbeexplainedintermsofhowrulesofPavlovianassociativelearning,butthereisnoreasoninprinciplewhyotherhowrulesshouldnotbeformulatedandtested.Thecurrentpreoccupationwithdifferentiatingcognitiveprocessesaccordingtotheirhowrulesisunderstandableinhistoricalcontext.Itmaybeahealthyreactiontothemanyyearsinwhichthe"generalprocess"traditiondeniedtheexistenceofanysuchdiversity,andthesubsequentperiodinwhich"biologicalboundaries"or"constraintsonlearning"approachesfoughtfortherecognitionofwhat/whenvari•ation(e.g.,Johnston,1981).However,thereisariskofover-compensation,ofexag•geratingtheextentofhowrulediversity,andofunderestimatingthepotentialcontributiontobehavioraladaptationofwhat/whenvariationalone.WhatMakesResearchonCognition"Evolutionary"?AllofthecontributorstothisvolumesubscribetoabroadlyDarwinianaccountofevolution.Withinthis,however,atleastfourevolutionaryapproachestothestudyofcognitionarediscernable.Forconvenience,Iwillcallthemtheecological,phyloge-netic,comparative,andselectiontheoreticapproaches.EcologicalandPhylogeneticTheecologicalandphylogeneticapproachesare,toasignificantdegree,comple•mentary.Eachhashistoricalrootsinethology,andproceedsfromaknownfact EvolutionaryPsychologyintheRound13aboutevolution.Theecologicalapproachfocusesonthefactthatevolutiontendstoproduceadaptations,phenotypiccharacteristicsthatenableorganismstosurviveandreproduceintheiruniqueenvironments;thephylogeneticapproachemphasizesdescentratherthanadaptation.Itconcentratesonthefactthatthephenotypeofagiventaxondependsnotonlyontheselectionpressurestowhichthoseanimalsorplantshavebeensubjected,butalsoonthegeneticvariantsthattheyinheritedfromtheirancestralspecies.Asaconsequenceofthisdifferenceinemphasis,thetwoapproachesseekevidenceofdifferentevolutionaryfootprintsoncognition.Theecologicalapproachanticipatesthataspecies"cognitivecapabilitieswillbecorrelatedwiththedemandsofitsnaturalenvironment,andinvestigatesthecharacterandspecificityofthiscorrelation—whichcognitivecharacteristicsaretunedtoenvironmentaldemand,andwithwhatdegreeofprecision.Thephylogeneticapproach,ontheotherhand,aimstochartthewayinwhichcognitivecapabilitiesvarywithphylogeneticrelatedness—toidentifywhereinevolutionarylineagesmajorcognitivechangehasoccurred,andtospecifythenatureofthesechanges.Noneofthecontributorstothisvolumewoulddenythattherearelikelytobebothecologicalandphylogenetictrendsintheevolutionofcognition.Bothpatternsaresoclearlyapparentintheevolutionofmorphological,anatomical,andphysiologicalcharacteristics,that,whenweturntocognitionandbehavior,thechallengeisnottodiscoverwhethertheyarebothpresent,buttouncovertheirrelativecontributionstoparticularcognitivecharacteristicsinparticularregionsofthephylogenetictree.Consequently,noneofthechaptersinthisvolumerepresenttheecologicalapproach,orthephylogeneticapproach,inpureform.However,Huberconcentratesonthephylogeneticapproach,examiningitshistoricalrootsintheworkofKonradLorenzandilluminatingsomeofthemethodologicalproblemsthatmakeitdifficulttopursue.Shettleworth,attheeditors"request,makesastrongcaseinfavorofmodu•larity,acontemporaryvariantoftheecologicalapproach.Themodularityapproachischaracteristicofwhatiscurrentlyknownas"evolu•tionarypsychology."Incommonwithotherecologicalapproaches,itisconcernedwithbehavioraladaptations,butthemodularapproachisdistinctiveinthatitattributesthemtopsychologicalmechanismswithspecificproperties.Thesepsycho•logicalmechanismsor"modules"arethoughttobe"domain-specific,"tohavedis•tinctivewhat/whenrules,andtohavedistinctivehowrules.FollowingFodor(1983),itisoftenalsoassumedthatmodulesareinnate,havedistinctiveneuralsubstrates,areautomaticallyactivatedbyinputfromtherelevantdomain,andare"informa-tionallyencapsulated."Thismeans,roughly,thatmodulesarerelativelyimperme•abletoinformationfromcentralormoregeneralcognitiveprocesses. 14CeciliaHeyesThebroaderecologicalapproachisrepresentedinthechaptersbyBateson,Clay•tonetal.,Dunbar,Heinrich,Lefebvre,Sterelny,andTomasello.Bateson"stopic,imprinting,isdennedbyitsroleinbehavioraladaptation,andClaytonetal.weremotivatedtoseekevidenceofepisodicmemoryinscrubjaysbyreflectiononthedemandsoftheirnaturalecologyasfood-storingbirds.Dunbar,Heinrich,Sterelny,andTomaselloallexpresssomesupportforthe"socialintelligence"or"socialfunc•tionofintellect"(Humphrey,1976)hypothesis,theideathatcomplexsocialenvi•ronmentsareapowerfulstimulusfortheevolutionofcomplexcognitivecapacities,andLefebvreexaminestherelationshipbetweenanecologicalvariable,distributionoffoodresources,andlearningandinnovationinbirds.However,likeDunbar"sandTomasello"schapters,Lefebvre"sdiscussionalsosubsumesthephylogeneticapproach,andintegratesitwithanecologicalanalysis.DunbarandTomaselloareconcernedwithcognitivetransitionsintheprimatelineage;Lefebvreexaminesaviantaxa.ComparativeThecomparativeapproachtothestudyofcognitionisrepresentedinthisvolumebyBitterman,DickinsonandBalleine,Delius,Mackintosh,andMacphail.Practitionersofthisapproachfocusintensivelyonthehowrulesofcognition(e.g.,associativelearning,conceptualization,goal-directedness),studythemwithahighlevelofmeth•odologicalrigorinafewnonhumantaxa(includingrats,pigeons,rhesusmonkeys,honeybees,ravens,goldfish,andchimpanzees),andcomparetheresults,implicitlyorexplicitly,witheachotherandwithwhatisknownabouthumancognition.Theoutcomeofthiscomparisonprocess,unlikethoseoftheecologicalandphylogeneticapproaches,ismorecommonlythediscoveryofsimilaritiesthanofdifferences.But,whilethecomparativeapproachemphasizesevolutionarycontinuityoverevolution•arydiversity,andtheroleofontogenyratherthanofphylogenyinbehavioraladap•tation,itisnoless"evolutionary"thantheotherapproaches.SelectionTheoreticTheselectiontheoreticapproach,representedinthisvolumebyLefebvre,RichersonandBoyd,andWilsonetal.,hasmuchincommonwiththeecologicalapproach.Likethelatter,itispreoccupiedwithadaptation,withthefitbetweenanimals"behavioralandcognitivetraitsandthedemandsoftheirnaturalenvironments.However,theselectiontheoreticapproacharguesthatthisfitarisesnotfromjustoneevolutionaryprocess,naturalsectionoperatingongeneticvariation,butfromseveral(e.g.,Campbell,1974;PlotkinandOdling-Smee,1981;Campbelletal.,1997;SoberandWilson,1998;Wilsonetal.,thisvolume).Anevolutionaryselectionprocess,a EvolutionaryPsychologyintheRound15processinvolvingvariationandselectiveretention,operatesnotonlyongeneticvariation,thatis,atthephylogeneticlevel,butalsoatontogeneticandculturallevels.Intheontogeneticcase,cognitivevariants—contentsorprocessesofthought—arisefrominteractionbetweenanindividualanimalanditsenvironment,andareselec•tivelylost/forgottenorretained/rememberedaccordingtotheconsequencesoftheirbehavioralexpressionforthatindividual.Attheculturallevel,thecognitivevariantsarecontentsorprocessesofthoughtcharacteristicofgroupsofindividuals,andnor•mallyacquiredbyindividualsthroughsocialinteraction.Theyaregeneratedwithinthegroup,orviathegroup"sinteractionwithitsenvironment,andselectedaccordingtotheirsuccessinbeingtransmittedtonewindividualsorgroups,whichmayormaynotbeorcontainbiologicaldescendantsofthepreviousculturalgeneration.Thus,culturalselectionmayormaynotconstitutegroupselection(SoberandWilson,1998),andvariantsthatarerelativelysuccessfulinculturalselectionmayormaynoten•hancethereproductivefitnessoftheindividualorgroup(BoydandRicherson,1985).Accordingtotheselectiontheoreticapproach,theseontogeneticandculturalpro•cessesareevolutionaryinthreesenses:(1)Phylogeneticevolution,naturalselectionoperatingongenes,hasshapedthecognitiveprocessesthatmakethempossible.(2)Theymakeautonomouscontributionstocognitiveandbehavioraladaptation.Thatis,phylogeneticevolutionalonecouldnotachievethesamedegreeoffitbetweencognitivesystemsandtheirenvironments,andontogeneticandculturalprocessessometimesperpetratecharacteristicsthatarenot"goodforthegenes,"thatwouldnotbeselectedatthephylogeneticlevel.(3)Theontogeneticandculturalprocesseseachpromoteadaptationthroughvariationandselectiveretention,thefundamentalDarwinianevolutionaryalgorithm.Thesefourapproachestostudyingtheevolutionofcognitionarecomplementaryratherthanantagonistic.Asfigure1.1suggests,selectiontheoreticanalysiscanbeseenasasubsetoftheecologicalapproach,thecomparativeperspectiveasasubsetofthephylogeneticapproach,andthecombinationoftheecologicalandphylo•geneticapproaches,broadlyconstrued,ascomprisingtheuniversalsetofcurrentevolutionaryresearchoncognition.Thispictureis,ofcourse,averysimplerepre•sentationofcomplexconceptualgeography.Eachofthefourmaincategoriescouldbesubdividedmanytimes.Thoseadoptingdifferentperspectivesoftenfindthem•selvesinhealthydispute,andmanyresearchersproductivelycombineapproaches.Forexample,inthisvolumeTomasello,Dunbar,andLefebvrecombineecologicalandphylogeneticperspectives,BatesonandClaytonetal.combineecologicalwithcomparativeanalysis,andintheirtheoreticalworkRichersonandBoydusetheproductsofallfourapproaches.However,inspiteofitssimplicity,figure1.1pro•videssomeindicationofwhatevolutionarypsychologylookslike"intheround." 16CeciliaHeyesFigure1.1Schematicrepresentationofrelationshipsamongcomponentsofevolutionarypsychologyintheround.EvolutionaryPsychologyintheRoundAtthebeginning,Ipointedoutoneprominentdifferencebetweentheresearchenterprisecurrentlyknownas"evolutionarypsychology"(e.g.,Buss,1999),andthatwhichisrepresentedbythisbook,evolutionarypsychologyintheround:theformerisconcernedalmostexclusivelywithhumancognitionandbehavior;thelatterinves•tigatesthesephenotypiccharacteristicsthroughouttheanimalkingdom.NowthatIhavesurveyedthecontributors"viewsonthenatureofcognition,onvariationamongcognitiveprocesses,andonthecharacteristicsofevolutionaryanalysis,furtherdif•ferencesbetweenthetworesearchenterpriseshavebecomeapparent.Thesecanbesummarized:(1)Evolutionarypsychologyintheroundisconcernedwiththephylo-genetichistory,aswellastheadaptivecharacteristics,ofbehaviorandcognition,descentaswellasselection.(2)Itrecognizesthatbehavioraladaptationcanbeachievedviamodificationofperceptualandmotorprocesses,insteadoforinaddi•tiontomodificationofcognitiveprocesses.(3)Itinvestigatesthecontributionstocognitiveadaptationofontogeneticandculturalprocesses,aswellasthatofnaturalselectionoperatingongenes.(4)Evolutionarypsychologyintheroundseeksinde- EvolutionaryPsychologyintheRound17pendentevidencethatvariationinwhat/whenrulesofcognitioniscorrelatedwithvariationinhowrules,resistingeasyinferencesfromdomainspecificitytomodu•larity.Twoexamples,takenfromresearchonspatialmemoryinfood-storingbirdsandmotorimitation,willservetoillustratethesecontrasts.Certainbirds,suchasClark"snutcrackersandMarshtits,whichexperiencesea•sonalvariationintheavailabilityoffood,scatterhoardforwintersurvival.Cacheretrievalhasbeendemonstratedtoinvolvememory(e.g.,ClaytonandKrebs,1994),andthesebirdshavebeenreportedtoperformbetterontestsofspatialmemorythanrelatedspeciesthatcachelessassiduously(e.g.,Kamiletal.,1994;Olsonetal.,1995).Furthermore,lesionsofthehippocampusimpairmemoryforcachesites(e.g.,SherryandVaccarino,1989),andbirdsthatdependheavilyonfoodstoringforwintersur•vivalhavelargerhippocampithanrelatedspecies(e.g.,Krebsetal.,1989).Oneinterpretationofthesedataisasfollows.Food-storingbirdshaveaninnate,spatialmemorymodulelodgedinthehippocampus.Thatis,naturalselectionoper•atingatthegeneticlevelhasgivenrisetoacognitiveprocess,basedinthehippo•campus,whichhasdistinctwhat/whenrulesanddistincthowrules;itprocessesspatialinformation,anddoessoinawaythatdiffersfromtheprocessingofnon-spatialinformation.Thiskindofinterpretationischaracteristicof"evolutionarypsychology."Byitsnature,evolutionarypsychologyintheroundrecognizesarangeofplausibleinterpretationsofthesespatialmemorydata.Itdoesnotunderestimateinferentialcomplexity,evenwherethetopichasbeenresearchedsoextensivelyandsoelegantlyasspatialmemory.Onealternativeaccount,whichisconsistentwithrecentevidencethattheactofcacheretrievalstimulateshippocampalgrowth(e.g.,ClaytonandKrebs,1994;Leeetal.,1998),andofepisodic-likememoryinscrubjays(Claytonetal.,thisvolume)isasfollows.Inresponsetoselectionpressurefromseasonalvariationsinfoodsupply,phylogeneticevolution(naturalselectionoperatingatthegeneticlevel)hasfurnishedcertainbirdswithaspecializedbehavior,namely,scatterhoarding.Theremaybewhat/whendistinctiveperceptualandmotorprocessescon•trollingthisbehavior,buttheactionofcachingdoesnotitselfinvolveasignificantcognitivecomponent.Fromthepointinontogenywhenhoardingbehaviorbegins,higherthanaveragedemandsaremadeonmemory.Hoardingcreatesforthehoarderanenvironmentinwhichfoodresourcesarescattered,andtherebyexercisesthememorysystem.Thissystem,locatedinthehippocampus,growsasitdealswithahighervolumeofinformation,butneitheritspotentialforgrowth,northehowrulesthatitimplements,havebeenchangedbynaturalselectionrelativetothoseusedbyotherspeciestorememberspatialstimuli.Thus,accordingtothisinterpretation,thespatialmemoryperformanceoffood-storingbirdsisaproductofphylogeneticand 18CeciliaHeyesontogeneticspecialization.Phylogeneticevolution,naturalselection,hasyieldedabehavioral,butnotacognitiveadaptation,and,provokedbythisbehavioraladap•tation,acognitivesystemwithspecializedwhat/whenrules,butnotdistinctivehowrules,emergesinthecourseofontogeny.Researchonmotorimitationismorediffuse,andhasprovidedmuchlessreliableinformationthanthatonspatialmemoryinfood-storingbirds.However,itillustratesthecontrastingattitudestowardcultureofevolutionarypsychologyandevolutionarypsychologyintheround.Motorimitation,thecapacitytolearnanovelbodymovementbyseeingitdone,hasbeenclearlydemonstratedonlyamonghumans.Manyresearchersbelievethatitoccursinotherapes(e.g.,TomaselloandCall,1997),but,likeallotherputativeevi•denceofimitationinnonhumananimals,thebasesoftheseclaimshavebeenorcouldbechallenged(e.g.,AkinsandZentall,1996;Moore,1996;Lefebvreetal.,1997;Campbelletal.,1999).Furthermore,itisnotclearhowhumans,oranyotherani•mals,couldimitatecertain"perceptuallyopaque"actions(HeyesandRay,2000).Forexample,imitationofnovelfacialexpressions,whichareseeninothersbutfeltbyoneself,wouldseemtorequiresomekindofcross-modaltransformationofinformation,anditisnotclearwhatsortofcognitivehowrulescouldachievethistransformation.Thisproblemnotwithstanding,ithasbeennotedrepeatedlythatimitationlearningcouldbethemeansbywhichmanyculture-specificbehaviorsaretransmitted.Evolutionarypsychologyhasdeducedfromobservationsofthiskindthatimitationlearningisaphylogeneticallyspecializedcognitivemodule,whichselectivelypro•cessessensoryinputfromothers"bodymovementsaccordingtodistinctiverulesofoperation,isfoundonlyinhumansandpossiblyotherapes,andsupportsculturaltransmissionofinformation.Aninterpretationthatisatleastequallyconsistentwithcurrentevidence,andmoreinthespiritofevolutionarypsychologyintheround,suggeststhatthecapacitytoimitatearisesfromontogeneticspecialization,usingculturalinput,ofphylogeneticallygeneral,associativelearningprocesses(HeyesandRay,2000).AccordingtothisAssociativeSequenceLearning(ASL)hypothesis,childrenacquirethecapacitytoreproduceactionunits(fragmentsofwhatwouldnormallybedelineatedas"anaction")throughcontiguousexperienceofseeinganddoingeachunit,therebyformingassociationsbetweensensoryandmotorrepresen•tationsofeachunit.Theselinksestablishanimitationrepertoire.Thatis,whenunitsintherepertoireareobservedinanovelsequence,thesequencecanbelearnedbyobservationalone,andthemodeledmovementcanbereproduced.Themostimpor•tantsourcesofcontiguousexperienceofseeinganddoingactionunits,andtherefore EvolutionaryPsychologyintheRound19themostsignificantinfluencesonthedevelopmentofanimitationrepertoire,areopticalmirrorsandtheadulttendencytoimitateinfants;both,broadlyspeaking,areculturalphenomena.Thus,ontogeneticspecializationyieldsacognitiveprocessthatisdistinctiveinitswhat/whenrules(itreproducesbodymovements)andthatbothsupportsandissupportedbyculture.Forbothoftheseexamples—spatialmemoryandimitation—moreevidenceisneededtoestablishtherelativemeritsofthealternativeinterpretationsIhaveout•lined.Forexample,weneedtoknowwhetherthehippocampiofstoringandnon-storingspeciesgrowatcomparablerateswhentheyaregivenspatialmemorytasks,and,toevaluatetheASLtheoryofimitation,whetherpriorexperienceofseeinganddoingactionunitsfacilitatessubsequentimitationofsequencesoftheseunitsinhumansandotheranimals.Whatevertheresultsofthesefurtherexperiments,thespatialmemoryandimitationexamplesillustratethenatureofthebiasesinherentinmuchevolutionarypsychology,and,moregenerally,emphasizethatevolutionarypsychologyintheroundgeneratesandtestsmorecomplexhypotheses.Manyofthecontributorstothisvolume,myselfincluded,wouldarguethatsuchcomplexityisnecessarytocaptureeventhebasicsoftheevolutionofcognition,butitcertainlydoesn"tmakelifeeasy.Itismucheasiertoattributealladaptivevariationamongcognitiveprocessestophylogeneticallyevolvedmodulesthantoinvestigatethenature(what/whenand/orhowrules)andevolutionarysources(therelativecontributionsofphylogenetic,ontogenetic,and/orculturalprocesses)ofsuchvariation.Perhapsthisdifficult(butrewarding)processofenquirycanbeassistedbyboldhypothesesabouttheevolutionofcognitionthattakeaccountofthedifferenttypesandsourcesofvariation.Inthisspirit,Ioffer,asapartingshot,myownboldhypotheses:What/whenrulesofcognitionaremorelabile,moreresponsivetochangeinecologicaldemand,thanhowrules,andadaptivespecializationofcognitivepro•cessesoccursmorereadilyattheontogeneticthanthephylogeneticlevel.Therefore,wewillfindacrosstheanimalkingdomthatmostevolutionaryvariationsincogni•tionarisethroughontogeneticspecializationofwhat/whenrules;thatontogeneticspecializationofhowrulesandphylogeneticspecializationofwhat/whenrulesoccurlessoften;andthatphylogeneticspecializationofhowrulesisveryrareindeed.SummaryThischapterintroducesthecontentsofthisvolumebyaddressingthreefundamentalquestionsabouttheevolutionofcognition:Whatiscognition?Whatarethediffer•encesamong"different"cognitiveprocesses?Whatmakesresearchoncognition 20CeciliaHeyes"evolutionary"?Inanswertothefirstofthese,cognitivestatesandprocessesarelooselydefinedastheoreticalentitiesprovidingafunctionalcharacterizationoftheoperationsofthecentralnervoussystem,whichmayormaynotbeobjectsofconsciousawareness,andthataredistinctfromperceptualandmotorprocesses.Indiscussingthesecondquestion,itissuggestedthatcontemporaryresearchersdifferentiatecognitiveprocessesusingwhat/whenrules,whichspecifyenvironmen•talinputsand/orcognitiveproducts,andusinghowrules,whichspecifyprocessingoperations.Inferringhowrules(whichareofprimaryconcerntocontemporaryinvestigators)fromwhat/whenrulesisdifficultunderanycircumstances,andmaybeimpossiblewhenthelattercircumscribeenvironmentalinputsratherthancognitiveproducts.Addressingthethirdquestion,itisarguedthattherearefourprincipalevolutionaryapproachestothestudyofcognition—ecological,phylogenetic,com•parative,andselectiontheoretic—andthattogethertheycompriseevolutionarypsy•chology"intheround".Bycontrastwithwhatiscurrentlyknownasevolutionarypsychology,thisresearchenterpriseinvestigatesphylogenetic,ontogenetic,andcul•turalcontributionstobehavioraladaptationinhumanandnonhumananimals.Alsobycontrastwiththeassumptionsofevolutionarypsychology,itissuggestedincon•clusionthatmostevolutionaryvariationsincognitionarisethroughontogeneticspecializationofwhat/whenrules;thatontogeneticspecializationofhowrulesandphylogeneticspecializationofwhat/whenrulesoccurlessoften;andthatphyloge•neticspecializationofhowrulesisexceptionallyrare.ReferencesAkinsCK,ZentallTR(1996)ImitativelearninginmaleJapanesequailusingthetwo-actionmethod.JournalofComparativePsychology,110:316-320.BoydR,RichersonPJ(1985)Cultureandtheevolutionaryprocess.Chicago:ChicagoUniversityPress.BushRR,MostellerF(1951)Amathematicalmodelforsimplelearning.PsychologicalReview,58:313-323.BussDM(1999)Evolutionarypsychology.Boston:AllynandBacon.CampbellDT(1974)Evolutionaryepistemology.In:ThephilosophyofKarlRPopper(SchlippPA,ed),pp413-463.LaSalle,IL:OpenCourt.CampbellDT,HeyesCM,FrankelB(1997)Fromevolutionaryepistemologyviaselectiontheorytoasociologyofscientificvalidity.EvolutionandCognition,3:5-38.CampbellFM,HeyesCM,GoldsmithA(1999)StimuluslearningandresponselearningbyobservationintheEuropeanstarlingusingatwo-object/two-actiontest.AnimalBehaviour,58:151-158.CatchpoleCK,SlaterPJB(1995)Birdsong:Biologicalthemesandvariations.Cambridge:CambridgeUniversityPress.ClaytonNS(1996)Developmentoffoodstoringandthehippocampusinjuvenilemarshtits.BehavioralBrainResearch,74:153-159. EvolutionaryPsychologyintheRound21ClaytonNS,KrebsJR(1994)Hippocampalgrowthandattritioninbirdsaffectedbyexperience.Proceed•ingsoftheNationalAcademyofSciencesUSA,91:7410-7414.CookM,MinekaS(1987)Second-orderconditioningandovershadowingintheobservationalcondition•ingoffearinmonkeys.BehaviorResearchandTherapy,25:349-364.CookM,MinekaS(1989)Observationalconditioningoffeartofear-relevantversusfear-irrelevantstimuliinrhesusmonkeys.JournalofAbnormalPsychology,98:448-459.CookM,MinekaS(1990)Selectiveassociationsintheobservationalconditioningoffearinrhesusmon•keys.JournalofExperimentalPsychology:AnimalBehaviorProcesses,16:372-389.CosmidesL,ToobyJ(1994)Originsofdomainspecificity:Theevolutionoffunctionalorganization.In:Mappingthemind(HirschfeldLA,GelmanSA,ed),pp85-116.Cambridge:CambridgeUniversityPress.DalyM,WilsonMI(1999)Humanevolutionarypsychologyandanimalbehaviour.AnimalBehaviour,57:509-519.DickinsonA(1980)Contemporaryanimallearningtheory.Cambridge:CambridgeUniversityPress.FodorJA(1983)Themodularityofmind.Cambridge,MA:MITPress.HeyesCM(1994)Sociallearninginanimals:Categoriesandmechanisms.BiologicalReview,69:207-231.HeyesCM,RayED(2000)Whatisthesignificanceofimitationinanimals?AdvancesintheStudyofBehavior,29:215-245.HumphreyNK(1976)Thesocialfunctionofintellect.In:Growingpointsinethology(BatesonPPG,HindeRA,ed),pp303-317.Cambridge:CambridgeUniversityPress.JohnstonTD(1981)Contrastingapproachestoatheoryoflearning.BehavioralandBrainSciences,4:125-173.KamilAC,BaldaRP,OlsonDJ(1994)Performanceoffourseed-cachingcorvidspeciesintheradial-armmazeanalog.JournalofComparativePsychology,108:385-393.KrebsJR,SherryDF,HealySD,PerryVH,VaccarinoAL(1989)Hippocampalspecializationinfood-storingbirds.ProceedingsoftheNationalAcademyofSciencesUSA,86:1388-1392.LeeDW,MiyasatoLE,ClaytonNS(1998)Neurobiologicalbasesofspatiallearninginthenaturalenvi•ronment:Neurogenesisandgrowthintheavianandmammalianhippocampus.Neuroreport,9:15-27.LefebvreL,TempletonJ,BrownK,KoelleM(1997)Caribgracklesimitateconspecificandzenaidadovetutors.Behaviour,134:1003-1017.MacphailEM(1985)Vertebrateintelligence:Thenullhypothesis.PhilosophicalTransactionsoftheRoyalSocietyB,308:37-51.MinekaS,CookM(1986)Immunizationagainsttheobservationalconditioningofsnakefearinrhesusmonkeys.JournalofAbnormalPsychology,95:307-318.MooreBR(1996)Theevolutionofimitativelearning.In:Sociallearninginanimals:Therootsofculture(HeyesCM,GalefJrBG,ed),pp245-265.SanDiego:AcademicPress.OlsonDJ,KamilAC,BaldaRP,NimsPJ(1995)Performanceoffourseed-cachingcorvidspeciesinoperanttestsofnonspatialandspatialmemory.JournalofComparativePsychology,109:173-181.PepperbergIM(1987)Acquisitionofthesame/differentconceptbyanAfricanGreyparrot(Psittacuseri-thacus):Learningwithrespecttocategoriesofcolor,shape,andmaterial.AnimalLearningandBehavior,15:423-432.PlotkinHC,Odling-SmeeFJ(1981)Amultiple-levelmodelofevolutionanditsimplicationsforsocio-biology.BehavioralandBrainSciences,4:225-268.RescorlaRA,WagnerAR(1972)AtheoryofPavlovianconditioning:Variationsintheeffectivenessofreinforcementandnonreinforcement.In:ClassicalconditioningII:CurrentResearchandTheory(BlackAH,ProkasyWF,ed),pp64-99.NewYork:Appleton-Century-Crofts. 22CeciliaHeyesSherryDF,SchacterDL(1987)Theevolutionofmultiplememorysystems.PsychologicalReview,94:439-454.SherryD,VaccarinoAL(1989)Hippocampusandmemoryforfoodcachesinblack-cappedchickadees.BehavioralNeurosciences,103:308-318.SoberE(1998)Morgan"scanon.In:Theevolutionofmind(CumminsDD,AllenC,ed),pp224-242.NewYork,Oxford:OxfordUniversityPress.SoberE,WilsonDS(1998)Untoothers:Theevolutionandpsychologyofunselfishbehavior.Cambridge,MA:HarvardUniversityPress.TomaselloM,CallJ(1997)Primatecognition.Oxford:OxfordUniversityPress. Psychophylogenesis:InnovationsandLimitationsintheEvolutionof2CognitionLudwigHuberMindcanbeunderstoodonlybyshowinghowmindisevolved.—HerbertSpencer,1855AhundredfortyyearsafterSpencer"sassertion,evolutionarypsychologistslikePlotkin(1997)againclaimthat"thelightofevolution"willilluminatepsychology.Nowatthebrinkofanewmillennium,wheredoesourunderstandingofthemindstand?Somebelievethatthetwenty-firstcenturycouldbecometheCenturyoftheMind,changingtheunderstandingofourselvesinwaysthatweareonlybeginningtoimagine.Otherswarnagainstbeingtoooptimistic,becausewearefarfromclosingthegapbetweenpsychologyandbiology,aprerequisiteforacompleteunderstandingofcognition.HereIwanttoadvocatetheethologicalapproachexaminedbythelateKonradLorenz.TheattractionofthisendeavorstemsfromLorenz"sattempttoexpandthebasicbehaviorsystemsapproachbyemphasizinggreaterrolesforlearninganddevelopment,stimulusprocessingandintegration,responseorganizationandcoor•dination,andtheco-regulationofhierarchicalandinteractingmotivationalstates.Drawingoncontemporaryevolutionarytheoryandsystemstheory,especiallythenotionsofemergentmacroevolutionarytransitionsandhierarchicallevelsofselec•tion,hesoughttoembodycognitioninbiology.Ifweignore,forthemoment,theartificialaspectsofcognitionshownbycom•puters,wemayagreethatcognitionisanaturalabilityofbiologicalsystems.Itdependson—andisconstrainedby—acertainorganizationoftheeucaryoticcell.Moreover,asabio-function,itisapartofthebehavioralequipmentofanyorganismandthussubjectedtothesamerulesasdootheraspectsofitsphenotype.IagreewithPlotkin"s(1997,p.1)formulationofDobzhansky"sfamousclaimthat"nothinginbiologymakescompletesenseexceptinthelightofevolution."Amazingadvanceshavebeenmadeinourunderstandingofgeneticprinciples,geneticcodeshavebeendeciphered,thecompletegenomeofsome(invertebrate)specieshavebeenencoded,andourunderstandingofhowgeneticinformationistransformedintoproteinsduringthecourseofontogeneticdevelopmenthasim•proved.Unfortunately,thepathfromDNAtocognitionisoftenpoorlyunderstood.Aone-to-onecorrespondencebetweengeneandcognitiondoesnotexist.Instead,theevidencesuggestsanetworkofgeneticinteractionsthatactsonandsetsupanetworkofneuronalcircuits.These,inturn,interacttogiverisetocognitivebehavior.Thequestionofhowgenessetupbehavioraffordsananswerintermsofhighlyinteractivesystemsactingatmanylevels(figure2.1). 24LudwigHuberFigure2.1Levelsandloopsofcognitionandtherespectivescientificareas.(AfterPlotkinandOdling-Smee,1981;Greenspanetal.,1994)Behaviorisanemergentpropertyofwholeorganismsthatresultfromthecomplexhierarchicalorganizationofunderlyingmolecular,biochemical,developmental,andphysiologicallevels...Theselevelsoforganization,however,donotactonlyinonedirection:eachonefeedsforwardandbackwardontotheothersthroughvariousshort-andlong-termmechanisms.Thismeansthattheultimateinfluenceofgenesonbehaviorisseldomdirectorsimpletotrace.(Greenspanetal.,1994,p.75,78)Ifadirectreconstructionofthecognitiveabilitiesofdifferentspeciesintermsoftheirgeneticconstitutionisunrealistic,whythenstudytheevolutionofcognition?Orwhyisanyoneinterestedinthephylogenyofhumancognition?Althoughnobody Psychophylogenesis25woulddenythatcognitionhasevolved,somedenythatwewillbeabletotraceitsexactevolutionarypath(Lewontin,1998).Ofcourse,thefactthatevolutionishis•toricalposesproblemsnottotheidentificationofevolutionaryinnovations,buttotheidentificationoftheirsequenceandconsequence.Aretherewaystoescapethetrapofmerestorytellinginseekingevolutionaryexplanationsofhumancognition?DifferentApproachestotheEvolutionofCognitionDespitesomecourageousexceptions(e.g.,Rensch,1967;Hodos,1970;Thomas,1980;Povinelli,1993;Byrne,1994;Moore,1996),itremainsafactthathephyloge•neticinvestigationofthemajortransitionsincognition—whichIwillbrieflycall"psychophylogenesis"—isunderrepresentedrelativetoecological,microevolutionary,orprotoevolutionaryconcerns(Timberlake,1993).TheDarwinianparadigmforstudyinghowbiologicaltraitshavecomeintoexistencehasbeenusedinalmostallbranchesofbiology,especiallyinthecomparativedisciplines.Infact,thecomparativestancehasalsobeenclaimedinthebeginningsofmodernpsychology(seeBitterman,thisvolume).Inretrospect,however,theproductivityofthisenterprise,measuredintermsofitsimpactontextbooksofevolution,wasquiteweak.Thereasonmightbethatinthecognitivedomainthefocuswas,atbest,onevolutionintheirbroadcon•cernandontheassessmentof(functional)similarity,ratherthanofhowandwhentheseabilitiescameintoexistence(HodosandCampbell,1969).Incontrasttotheresearchagendadevotedtothereconstructionoftheconcretecourseofevolution,therelianceonabstractrelationssuchasscalesofintelligenceanduniversallawsappearedtoremainprotoevolutionary,thatis,doesnotmeetthecomplexitiesofevolutionaryprocesses.Inethology,individualproblem-solvingabilitiesbecameacentralobjectofinves•tigationquitelateinthiscentury.Atfirstsight,itwouldappearthattheecologicalapproachhasachievedconsiderableprogresstowarddescribingtheevolutionofcog•nition.Forexample,inordertotracethepathsfromspecializedadaptationstoenvi•ronmentaldemands,ecologicalcognitivistshaveconcentrateduponcloselyrelatedspeciesthatfilldivergentniches,anddistantrelativesthatfillconvergentniches(seeShettleworth,thisvolume).However,thereisatleastonereasontodoubtthatthisapproachissufficienttoreconstructmajorevolutionarypaths.Itiscommonlysup•posedthatcognitivefaculties(e.g.,recoveringfood)aretheresultofselectiveforcesexertedbyspecificenvironmentalcircumstances(e.g.,meagerperiods).However,evolutionisseldomasstraightforwardasthis.Thebestmetaphoristoviewevolutionasa"tinkerer,"buildinginnovationsbymodifyingexistingstructures,ratherthanbydesigningthemfromscratch.Also,thereconstructionofthesequenceofcognitive 26LudwigHuberbreakthroughsliesoutsidetheresearchprogramoftraditionalecologicallearningtheoristsorbehavioralecologists.Quiteclosetothisresearchparadigmofexplaininghowanimalssolvethesortsofproblemstheyfaceintheirnativehabitatsisafurther,recentlyemerged,discipline.Cognitiveethology(Ristau,1991)grewfromasolidethologicalbackgroundinordertounderstandmental-stateattributionsacrossspeciesboundaries.Therangeofphe•nomenaincludesfeelingsandconsciousness,andthesourcesofinformationinvolveanecdotes,aswellaslaboratoryandfieldstudies.However,itisworthmentioningthattheideaofextendingtheethologicaltraditiontoincludecognitivefunctionsandmentalstatesisnotabrandnewone(Menzel,1986).Amongothers,itwasLorenzwhosuggestedthatthebestwaytounderstandthehumanmindwouldbetofocusonthose"ratiomorphic,"ratherthanrational,aspectsthathaveevolvedforthesakeofsurvival.Thecommonstanceofevolutionaryepistomologistsisa"naturalized"epis-temologythataimstoexplain"man"sstatusasaproductofbiologicalandsocialevolution"(Campbell,1974,p.413).Similartotheaboveapproachestowardunderstandinginwhatrespectsthecog•nitivebehaviorofspeciesdifferfromeachother—andespeciallyfromHomosapiens—isachallengeforneurosciencetoimproveinferencesaboutthehumanbrainfromthecomparativestudyofthebrainsofanimals.Thisparadigmshiftstandsinsharpcontrasttothelongtraditioninneurosciencetofocusstudyona"model"or"repre•sentative"species.Asincomparativepsychology,itwassupposedthatallmammalspossessvariantsofthesamebrain,andthatthisjustifiedextrapolatingfromonegrouptoanother.Thecomparativeapproachisguidedbythefactthatbrainsdifferconsiderablyandthatourunderstandingofevolutionaryrelationshipscanpromotefundamentalinsightsonfunction(Hodos,1970).TheBiologyofLearningEthologists,likeLorenz,approachcognitionfromabehavioralpointofview,emphasizingthatlearningisonlyoneofseveralwaystocopewithanunpredictableenvironment.Ifbehaviorfunctionsasthemeansofadjustingtheorganismtoexter•nalorinternalenvironmentalchanges,inordertoremaincoupledwithexistingrele•vantpatternswithinthemselvesandtheirenvironments,itmustbesufficientlyplasticinthefaceofunexpectedchange.Thisrenderstheorganism"stasktwofold.Itmustoperateefficientlyunderknownconditions,anditmustmaintainadaptationundernewconditionswithaminimumofdisruption.Understandingthemainclassesofhistoricalprocesses—constraintsandinnovations—andtheirinterrelationshiprequiresabroaderframeworkthanthatusedintraditionallearningtheory. Psychophylogenesis27StimulatedbySchrodinger"sfamousbookWhatIsLife?(1944),biologicalsystemstheorygrewinordertodiscoverandformulategeneralconceptsandlogicalrelationscharacteristicoflivingascontrastedwithinorganicsystems.Generally,themainte•nanceofcouplingwiththeenvironmentisthemostfundamentalcharacteristicofallself-reproducingsystems.Riedl(1995)distinguishestwoimportantprinciples:(1)correspondencewiththeconditionsexternaltothesystem(theorganism),and(2)coherencewiththeinternalconditionsofthesystem.Theprocessescontributingtothelattertaskaremostlyobscuretotheexternalobserver,andhavebeenunder•estimatedorevenignoredbybehavioristsfocusingontheprocessesthatleadtoobservablechangesinbehavior.Consequently,theprocessesinvolvedinsupportingadaptationtoexternalcircumstancesareconsideredtobeonesided,requiringtheanimalto"track"environmentalchanges.Thisimpliesthattheorganismismerelyreactingtoitsenvironmentandthustakesaratherpassiverole.Informationis,accordingtotheinformationprocessingparadigm,fedintotheorganismfromoutside,andthereafterstoredeitherbygeneticprocessesduringphylogenyorbymemoryprocessesduringontogeny.Systemicapproacheshavechallengedthisviewbysuggestingthatexternalinflu•encesareonlyakindof"trigger"or"selector"ofinternalmodifications,andareseverelyboundbytherestrictedpotentialsoftheorganism.Growingfromthisnotion,asmallbutquitevigorousmovement,calledconstructivism,emphasizesthatcognitionisnottherepresentationofapregivenworldbyapregivenmind;itisrathertheconstructionofaworldandamindonthebasisofthehistoryofactionsthatabeingintheworldperforms(MaturanaandVarela,1979).AccordingtoLorenz,learningisaprocessthatoccursinaccordancewiththoseprocessesthatoperateascontrolmechanismstokeepthesystemincoherence.Theseconservativeprocessesaretheresultofanearlierprocessofnaturalselectionthathasledtoahighdegreeofstructuralcouplingbetweentheorganismanditsenvi•ronment.Thishighdegreeofadaptationisresponsibleforthefactthatonlyaverylimitedamount,orspecifickind,ofinformationisacquiredduringtheprocessofontogeneticlearning,andfedintowhatErnstMayr(1974)callsopenprograms.Theterm"program"referstothefactthatthesesetsofinstructionsareplansforactionsthatappearasintegratedandcoordinatedstreamsofbehavior(Gass,1985).Fur•thermore,itimpliesthattheactionsoftheorganisminthefaceofenvironmentalchangeshaveastrongconservativefunction,"defending"theorganismagainstbeingperturbedrandomly.Inordertoachievethisdoublecontrol,thatis,trackingtherelevantaspectsoftheenvironmentandkeepingtheorganisminequilibrium,acon•siderableamountofgeneticinformationisrequired.Lorenzcalledthispredispositionforacquiringandstoringrelevantinformationthe"teachingmechanism."Green•spanetal.(1994,p.79)offeramodernformulation: 28LudwigHuberFinally,wemustbemindfulofthefactthat[these]geneticexperimentsservetopromotethenotionthatbehavior,includingthatofhumans,isboundtobegeneticallydetermined.Whileitmaybecorrecttoviewindividualgeneproductsasdeterministicelements,continuedstudyoftheprocessesunderlyingbehaviorwilleventuallyleadustotherealizationthatgeneproductsarebutaminutefractionofthetotalnumberofbehavioraldeterminants.Asecond,smallfractionwillbeidentifiableasrelativelystraightforwardenvironmentalfactors.Mostimpor•tantly,however,thevastmajorityofdeterministicfactorswillresideinthemultidudeofasyetunpredictableinteractionsbetweengeneticandenvironmentalfactors.Itisexactlythismodernviewofthecoalescenceofcausesaroundahighlycomplexanddynamicdevelopmentalprocessthattriestosolvethe"nature-nurture"debate(Plotkin,1997).AcceptingthatthesetwoprocessesarenotmutuallyexclusiveallowsustounderstandwhyLorenzemphasizedthat"alllearningabilityisbasedonopenprogramswhichpresupposethepresence,notofless,butofmoreinformationinthegenomethandoso-calledinnatebehaviorpatterns"(1977,p.65).Thedifferencebetweenclosedandopenbehaviorprogramsisthereforetheiramountofflexibility,orthedegreetowhichtheexactformofthebehaviorsequencestheygenerateiscontingentuponcurrentconditions.Variationisalsofoundintheirtaxonomicdis•tributionandintheirecologicalappropriateness.Instincts,reflexes,andmostlocomotorypatternsarerigidlyprogrammed,notonlyinloweranimals,butalsoinhumans.Thesamemayalsobetrueofcommunicativeactions,becausetheirinterpretationbyconspecificshasconsiderablesurvivalvalue.Closedprogramsareobservedinsolitaryandshort-livedspecies,whereasinsocialandlong-livedspecies,inwhichindividualshaveabundantopportunitytolearnfromtheirownandothers"experience,openprogramsdominate(Gass,1985).Further•more,inearlydevelopment,openprogramsareimportantformediatingthestorageofspecificrelevantinformation,forexample,the"object-to-be-followed"inthegosling"sbrain.InMayr"sscheme,anotherkindofbrainareamaythenberespon•sibleforprocessingandstoringofallsortsofinformation,althoughitsrelativeimportancewithinandacrossspeciescannotbedeterminedatthepresenttime.Effectorsystemsalsoshowvariationasafunctionofthelevelofoperation.Acloseanalysisofthefinestructureofbehaviororactionpatternssuggeststhatinheritanceisintact,whereasatthelevelofgrossmovementsequences,variabilityincreasesdramatically.Recentadvancesinourunderstandingofthedevelopmentofsimplemovementsindicatestheimportanceofunderstandinglearningas"achiefcharac•teristicoflivingthings,endowingtheorganismwithameanstoescapeitslimitedbuilt-inbehavioralrepertoire"(Kelso,1995,p.159).Changingbehavioralrepertoiresinthefaceofanewproblemisnotaccomplishedbychangingtheweightsofsomesynapsesinordertostrengthenorweakencertainassociations,butoccursaschange Psychophylogenesis29totheentiresystem.Itwasdemonstratedthatsensorystimulitriggerabruptchangesofneuroactivitypatterns,whichthenleadtotheemergenceofendogenousactivitypatternsinthecortex(Freeman,1995).LookingBehindtheMirrorAccordingtoMenzel(1986),philosophical,psychometrical,anthropocentrical,ortechnologicalquestionsdeterminetraditionalapproachestocognition.Thesetradi•tionsareconcernedwith(1)whetheranimalscomprehendwhattheyaredoing,(2)howthevariousspeciescanbesortedaccordingtotheirperformanceonstandardintelligencetests,(3)ofwhetherortowhatdegreeanimalsdowhatweourselvesdo,and(4)ofwhetherornotitwouldbepossibleto"shape"intelligentperformanceinanimalsortocreateartificialcognitivesystems.Incontrast,inBehindtheMirror(1977)Lorenzstartedfromthefundamentalquestionsofhowweasscientistscometounderstandnatureandhowweasorgan•ismsescapetheprisonofourbrains,whichareimpenetrablyseparatedfromtheouterreality.OnlybyaddingtheevolutionarydimensioncanwesolvetheKantianproblemthattheindividualhasnomeansofcheckingthecorrespondencebetweentheworldanditsself-constructedrepresentations.Eventhoughthesingleindividualisstillaprisonerinhisconstructedworld,thesystemassuchwillslowly,duringmillionsofgenerations,improveitsconstructs(SjSlander,1993).Inanutshell,thisistheadaptionistparadigm,holdingthattheresultofnaturalselectionisthegenerationofadaptiveorganicstructuressustainedbystructuralcouplingwithexternalcir•cumstances.Accordingtothisview,notonlydothegenomeandthebrainstorein-formation,butalsoorgansthemselvesareakindofin-corporatedknowledgereflectingthisfit.Thebiologicalversionofknowledgeisbasedonthewiderperspectiveofsystemstheory.Ifoneacceptstheexistenceofmanydifferentlevelsofintegrationandthenotionthatprocessesthatarecapableofacquiringandstoringrelevantinformationoccuronmanydifferentlevelsandareinterlinkedatmanypoints,onecanspeakofahierarchyof"knowledge-gainingprocesses,"withknowledgeasaseriesofnestedproductsofsuchahierarchy(PlotkinandOdling-Smee,1981).Transcendingthestaticviewof"strataofexistence"byconsideringthenatureofevolutionarychangeasthemainconcernforsystemicanalysis,Lorenzinventedthehistoricaldimension.Theevolutionaryargumentistheonlywaytoescapethe"Platonic"viewofnaturalvariation,aviewthatismaintainedeveninmoderncognitivescience(Menzel,1986).Thereisnotonebrain,onecognition,oneproblem,andonesolution.Biological 30LudwigHuberentities,betheystructuresorfunctions,havebeengeneratedinadynamicway,astheresultofphylogeneticandontogeneticprocesses.Theinescapabledifficultythenbecomestotracktheprocessofinnovationandtosortphenomenaamongwhichtherearenoclear-cutbordersbutallsortsofsmoothtransitions.Awayoutofthisdilemmamaybefoundbyapplyingfuzzylogicandchaosmodels.Thesecanshowthatthesuddenemergenceofnewsystemcharacteristicshasnothingtodowithmiracles,thoughourscientifictoolstoexplaintheprocessesatthecriticaltransitionpointsarestillinsufficient.ThreedecadesagoitwouldhavebeenunreasonabletodisagreewithLorenzthatthehistoricaluniquenessofphylogeneticinventionsleavebehindunexplainableresidues.AlthoughLorenzrejectedvitalistexplanations,hebelieved"thatnosystemonahigherlevelofintegrationcanbededucedfromalowersystem,howeverfullyonemayunderstandthislowersystem"(1977,p.35).Heemphasizedthreecrucialfactors:(1)theindependencyandmain•tenanceofsurvivalcapacityoftheformer"simpler"systems,(2)thecontinuityofimportanceandfunctioningoftheconstituentpartsinthe"higher"system,and(3)theimpossibilityofdetectinganytraceofthosesystemcharacteristicsthatcomeintoexistenceonlyatahigherlevelofintegration.Theclassicexampleinbiologyistheemergenceofconsciousnessfromtheinte•grativeactivityofalargenumberofneurons.Thesingleneuronfiresonthebasisofan"all-or-nothing"rule.Itdoesnotshowtheslightestsignofconsciousbehavior.Theneuronisalsoagoodexampleoftheprincipleofsimplificationbyspecialization.TheunicellularParameciumisabletobehaveasanorganismperformingallthefunctionsthatservesurvival,buttheganglioncellhaslostmostoftheseproperties.Theneuronexamplealsodemonstratesthat"higher"brainfunctionsdonotemergefromtheaccumulationofsimpleelements,butfromasophisticatedorganizationthat,inturn,allowssimplificationatthelowerlevels.Themutualrelationshipbetweenascent(improvedorganization)anddescent(simplification)maybecon•ceivedasakeycharacteristicofevolutionaryprocesses.AHierarchyofCognitiveCapacitiesThekindofprogressionthatLorenztoucheduponinhiswritingswasaresultofhisattempttocomparespeciesaccordingtodifferentmeansofprocessinginformationaboutsomerelevantexternalcircumstancesinordertobehaveappropriately.Thearrangementofthesecapacitiesorproclivitiesfollowsahypotheticalphylogeneticsequenceofinnovationsthatresultfrommodificationsofwhathasalreadybeenavailableatacertainpointinevolution.Therefore,theproposalisbasedonwhat Psychophylogenesis31Lorenzhypothesizedfromhisenormouscollectionofanimalbehaviorsandfromevolutionaryfacts,ratherthanfromanengineeringpointofview.Iwillexplicatethispointlater.Lorenzremainedlargelyreservedinsuperimposingthesequenceofcog•nitivemechanismsupontheanimalsystem.InBehindtheMirror(1977),thesequenceofknowledge-gainingprocessesissortedintosixchapters.AccordingtoLorenz"sscheme,thisclassificationofcognitivephe•nomenareflectsqualitative,ratherthanquantitative,differences.Thisisbecauseeachstepinvolvestheemergenceofnewsystemcharacteristics,thoughtheprocessesinvolvedinearlierstagesremainfullyfunctionalinthosespeciesthathavereachedahigherstage.Short-TermInformationGainInadditiontothegenomeandallprocessesinvolvingindividuallearning,thereareanumberof"closedprograms"thatacquireandoperateupon,butdonotstore,informationaboutinstantaneouslyarisingconditionsintheenvironment.Thebasicfunctionoftheseprogramsisnotmodificationoradaptation,butoperationonpre-adaptedstructures.Theyoccurinallspeciesatalltaxonomiclevels,andrepresentthefoundationofallexperience.Homeostasis,orthefeedbackcycle,isthemostbasicmeansbywhichanorganismmaintainsequilibrium.Togetherwiththeequallybasicfunctionofirritability,itallowsmobileorganismstotrackbeneficialconditions.Theamoeboid,kinetic,phobic,andtaxisresponsesareallexamplesofthisprinciple.Fur•thermore,innatereleasingmechanisms,fixedmotorpatterns,andappetitivebehaviorsareexamplesofevenmorecomplexsystemsthatexploitshort-terminformation.AdaptiveModificationsofBehavior(ExcludingConditioningbyReinforcement)Thefundamentaldifferencebetweenthemechanismsliesintheirpotentialtomodifythesensoryorneuronal"machinery"oftheanimal,butonlyinamannerthatimprovestheirsurvivalvalue.Theseopenprogramsarethemostbasicmeansbywhichrelevantinformationisacquiredandstored.Theycanbecalledlearningmechanismsinthesensethat"alllearningisanadaptive,teleonomicmodificationofphysiologicalmechanismswhoseoperationconstitutesbehaviour"(Lorenz,1977,p.81).Onlythemostelementaryofthesemechanisms,facilitationofmotorresponsesandsensitizationofreceptors,areconsideredtobenonassociative.Thefirsttrulyassociativemechanisms,distinguishedonthebasisoftheirbehavioralconsequences,involvehabituation,habitformation,escaperesponsesfollowingtraumaticexperi•ences,andimprinting,Lorenzremainedreluctanttospecifythephylogeneticorigin,acautiousattitudethatseemsjustifiedinviewofevidenceforconditioninginpara-mecia(Hennessyetal.,1979). 32LudwigHuberConditioningbyReinforcementThe"feedbackloopthroughwhichthefinalsuccessorfailureofthechainofpro•cessesisabletohaveamodifyingeffectonitsinitiatinglinks"(Lorenz,1977,p.84)representsafurther"fulgurative"improvement.Lorenzassumedthehigherfacultyofconditioningbyreinforcementdevelopedindependentlyinfivedifferentgroupsofanimals:cephalopods,crustaceans,arachnids,insects,andvertebrates.Heproposedthreepreconditionsforitsevolution.First,thebehaviorprogramthatisadaptivelymodifiedmustbeopen,orrequirealargevolumeofgeneticinformation.Second,thegeneticmemoryofthebehaviorprogrammustbeaccompaniedbyanassociativememorythatstorestherelevantcontext.Finally,therecordofreinforcementthatisfedbackintothesystemmustbesufficientlyreliable,andcharacterizedbyareduc•tioninmotivationaltension.Associationsareformedbetweenaprespecifiedclassofexternalinformationandbehavioralpredispositions,asexemplifiedbythedemon•strationofflavoraversionlearninginrats.Pathfmdingandmotorlearningalsofitintothisscheme.ThisbehavioralconceptionofconditioningdeviatesconsiderablyfromPavlov"sphysiologicalor"reflexological"terminology.Lorenzdistinguishedseveraldifferentformsofadaptivebehaviormodificationoccurringthroughconditioning,amongwhichheregarded"operantconditioning"tobeaspecialcase.Onlyhighlyexplorativeorganisms,suchastheratandsomeothermammals,plusafewbirds,areabletoassociate"significant"actions(e.g.,defendingmovements)withreinforcingsignalsfromcompletelydifferentbehaviorsystems(e.g.,foodreinforcement).Itismucheasiertoconditionmultipurposeresponses,suchasthoseinvolvedinthelocomotorrepertoire.Butitisvirtuallyimpossible,forexample,totrainarattomakecopulatoryresponsesinordertogainaccesstofoodpellets(S+)ortoavoidelectricshock(S-).Inordertounderstandwhatroledifferentformsoflearningplayinnatureandwhytheyleadtoadaptivemodificationsofbehavior,Lorenzsuggestedthreeheuristicrules.First,itisnecessarytodeterminewhichsubsystemofacomplex,modifiablebehaviorpatterncontainstheinnateinformationthatensurestheanimallearnsthebehaviorpatternsrequiredforsurvival.Second,itisnotpossibletounderstandanylearningprocesswithoutunderstandingthewholesystemthatthisprocessmodifies.Finally,thephysiologicalnatureofthereinforcementprocesshastobestudiedinde•pendentlyineachindividualcaseoflearning,becausethereisnouniversalreinforce•mentbutonlyreinforcementdependantuponthebehaviortobemodified.Howdoesthisconceptionfitintowhatiscurrentlyknownaboutassociativelearning,perhapsthemostgeneralandmostpowerfullearningmechanisminnature?Andhowhasthismechanismevolved?Fortunately,insightintothecellularmecha- Psychophylogenesis33nismunderlyingclassicalconditioningininvertebrates,especiallythemarinesnailAplysia,hasenormouslyimprovedourknowledgeinthisrespect.Acell-biologicalalphabetforsimpleformsoflearningexistsanditispossibletoexplaincertainhigherformsoflearninggenerallyassociatedwithcognitionincellular-connectionisttermsbycombinationsofafewrelativelysimpletypesofneuronalprocesses(HawkinsandKandel,1984).Particularly,itseemsasifthecellularmechanismofconditioningisanelaborationofthemechanismunderlyingsensitizationandhabituation.Evenhigher-orderfeaturesofconditioning,includinggeneralization,second-orderconditioning,blocking,andtheeffectofcontingency,canbeaccountedforbycombinationsofthealmostsimplermechanismofsensitization.TheRootsofConceptualThoughtOfcourse,thecellularmodelmaynotaccountformanyothercognitivephenomena.Lorenzsuggestedeightfacultiesasbeingimportantforourunderstandingofhumanconceptualthought.Allofthesefacultiesarefoundinanimalsandallhavetheirownsurvivalvalue.AlthoughitwouldbeextremelyvaluabletoexaminethevalidityofLorenz"shypothesesusingthedatafromrecentexperiments,thiswouldtranscendthescopeofthischapter.Instead,Ipointthereader"sattentiontotherespectivechaptersinthepresentvolume:(1)abstractionandGestaltperception(Deliusetal.;Mackintosh),(2)orientationandthecentralrepresentationofspace(Shettleworth),(3)insightfullearning(Heinrich),(4)voluntarymovement(DickinsonandBalleine),(5)objectandself-exploration(Heinrich;Macphail),(6)imitation(Heyes;Tomasello),and(7)tradition(Lefebvre).Iwouldonlyaddthewell-studiedcapacitiesofcategorizationandtheimportantcognitivebreakthroughofintermodality.Ampleevidencefortheformercapacityisavailablefrompigeons(Huber,1995).Inthevertebrateline,multimodalrepresenta•tionshaveoccurredabovethereptilianlevel.Asnakehuntingapreyissequentiallyfollowinga"visualrabbit,"an"olfactorialrabbit,"andfinallya"tactilerabbit"inordertoswallowit(Sjolander,1993).Thedog"shunting,bycontrast,isguidedbyamultimodalrepresentationofarabbitthatremainsactiveinachangingcontext.Also,thedogisabletodream,butthesnakeisnot.TheHumanMindandBeyondLorenzregardeditparadoxicalthatsomewhorefusedtolookatallthethingsthatmanandanimalshaveincommonunderestimatedthedifferencesbetweenthem.InordertodisputeDarwin"s(1871)notionthatthehumanminddiffersfromthemindofanimals"onlybydegreebutnotofkind,"onehastotaketwofactsintoaccount.First,withrespecttomostofthefacultiesmentionedsofar,humansoutperform 34LudwigHuberanimals.Second,thetrulyfundamentalstepforwardtowardthedevelopmentofthehumanmindwasonlypossibleafterarapidaccumulationoffigurationscausedbytheintegrationofseveralpre-existingcognitivefunctions.Thefacultiesofthehumanmindarecharacterizedbynewproperties,suchasverballanguageandcontext-freethought,thatcouldn"teasilybereducedtotheirprerequisites.However,itisofindispensableheuristicvaluetokeepinmindallthose"ratiomorphic"elementsofthehumanmindthatindicatethe"survival"ofoldercompetencies.Althoughsimplereductionismfails,itisneverthelessouronlychancetounderstandthehumanmindwithallits"deficienciesofadaptation"(Riedl,1995)byempiricallystudyingthecognitivecomponentslistedabove.EvolutionasInnovationDiversityisanessentialfeatureofthelivingworld.Thegeneralmethodofcopingwithdiversityistomakecomparisons.ComparisonsamongspecieshavebeenmadesinceAristotle,andwerealsoDarwin"sfavoredscientifictool.Lorenzespeciallywasadmiredforhiswide-rangingandextensiveexperienceofanimalbehavior,whichallowedhimtouseanactualexampleforvirtuallyanyphenomenonunderdiscus•sion.Manyoftheearlyethologists,likeWhitman,Heinroth,andJulianHuxleyalsoconsideredspeciescomparisonsastheessenceofbiologicalscience.Lorenzpro•posedthreeindispensablestagesforthedevelopmentofanyinductivenaturalscience,proceedingfrompurelyobservationalrecordinganddescribingoffacttotheorderlyarrangementofthesefactsinasystemand,finally,tothequestfortherulespre•vailinginthesystem.Therecentincreaseintheuseofthecomparativemethodincognitiveethologyandbehavioralecologyisduetoarevivedinterestinthefunctionofbehavioraltraits.Thekeyconceptfordescribingandexplainingbehavioralorcognitivediversityisadaptation,whichresultsfromadialecticalrelationshipbetweenananimalanditsenvironment.Intryingtouncoverthereasonsforevolutionarychange,comparisonsaremadeamonggroupsofanimalsthatshareasimilarwayoflife,inhabitasimilarecologicalniche,orfacesimilarbehavioraldifficulties.Convergentandparallelevo•lutionarychangecanleadtophenotypicresemblanceevenamongthosemembersofa"guild"thatarenotcloselyrelated.AdaptationandconvergencewerealsoimportantconceptsinLorenz"sthinking.Hestudiedanimalbehaviorinordertoelucidatetheamazingfactsofadaptedness,andconsidered"analogyasasourceofknowledge"(1974,p.229).However,com•parativethinkingmustnotstophere;thehistoricalaspectoflifeneedstobeincluded.In1898(p.328;butseeAtz,1970,p.67)Whitmanwrote,"Instinctandstructure Psychophylogenesis35aretobestudiedfromthecommonstandpointofphyleticdescent,andthatnotthelessbecausewemayseldom,ifever,beabletotracethewholedevelopmentofaninstinct."Thissentencemarksthebirthofcomparativeethology.LikeWhitman,Lorenzwasdeeplyconvincedthatacompleteunderstandingofbehaviorfailsifitdoesn"tincludeitsoriginandcontinuityinphylogeny.Amuch-citedexampleforthevalidityofevolutionaryreconstructiononthebasisofpurelybehavioraldataisLorenz"s(1941)classificationof20speciesofducksbytheirsimilaritieswithrespectto48behavioraltraits.Nevertheless,despitesupportfromleadingevolutionistssuchasSimpson,Dobzhansky,anddeBeer(seeAtz,1970),thephylogeneticapproachtothestudyofbehaviorfellintodecline.Theobviousweaknessofthisapproachisdueto:(1)thelackofafossilrecord,(2)thetenuousconnectionbetweenbehaviorandthenervoussystem,and(3)thediffi•cultiesassociatedwithfindingpropercriteriafordeterminingbehavioralhomologies(Atz,1970).Furthermore,homologyisessentiallyamorphologicalconceptthatenablescomparativeanatomiststodetectphylogenetictracesfarbeyondthefamilylevel.Theonlyreliablehomologizingofbehaviorhasbeenconfinedtocloselyrelatedforms.Ingeneral,thecontinuoustransmissionofgeneticinformationinphylogenyisnotnecessarilyaccompaniedbytheresemblanceofphenotypictraits.Thisdivergenceincreaseswithtime,asdoesthedifficultyofdetectinghomologouscharacteristics.Nevertheless,byapplyingindirectmethodsthistypeofevolutionaryanalysisremainsfeasible.Iftheevolutionofbehaviorproceedsinthesamewayastheevolutionofmolecularorstructuralcharacteristics,thenitmustatleastinparthaveageneticbasis.Fur•thermore,thisbasismustbesomewhatvariableifitsuppliesthematerialonwhichnaturalselectionacts.Onlyrecently,wehavewitnessedarevivedinterestinunder•standingthephylogenyofbehavior.Thisisdueinparttoadvancesinthequalityofphylogeneticreconstruction(HarveyandPagel,1991;Martins,1996).Modernsta•tisticalmethodologiesmayhaveledtothisdevelopment,andjustifythenotionthat"wemustlearntotreatcomparativedatawiththesamerespectaswewouldtreatexperimentalresults"(MaynardSmithandHolliday,1979,p.vii).Forexample,comparingintraspecificandinterspecificnoncommunicativebehaviorsallowsustomakeplausibleinferencesaboutthenatureofthebehaviorprogram(closedoropen),itsroleinmacroevolution,andtheoriginofmajorevolutionaryinventions(Mayr,1974).Suchanalysisrevealsthatnoncommunicativebehaviorsarebyfarthemostimportantfactorinmacroevolution.Theinvasionofnovelhabitatsandtheexploi•tationofnovelfoodsourcesisfacilitatedbyopenprogramsandtheuseofcognitiveandmemorycapacities.Thesecognitivestrategiessetupdiverseselectionpressuresthat,inturn,arethemselvesmoldedbyevolution. 36LudwigHuberWemustdistinguishbetweenattemptstousebehavioralcharacteristicsasataxo-nomictoolforthestudyofphylogenyandsystematics,andattemptstotracetheevolutionofparticularbehavioralpatternsbymakingindirectinferencesfromcom•parativestudies.Intheabsenceoffossilrecords,comparativestudiesofseveraldif•ferentspeciesoftenrevealaspectrumofpatternsandvariation,suggestingthatsomespecieshave"primitive"characteristics.Giventhatbehavioralpatternsmaysimplifyduringevolution,theassumptionaboutwhatisancestralandwhatisderivedisnotbasedupontheanalysisofasinglespecies,butuponthesystematiccomparisonofmanyrelatedforms.A"strongcomparativeinference"ispossibleonlyifasequenceofintermediateformsofthespectrumofextantspeciesisavailable.Furthermore,onecanmake"weakinferences"ontheassumptionthattheelementsofabehaviorpat•ternsharedbyanumberofspeciesrepresentancestraltraits.Usingthismethod,Tinbergen(1959)compileda"progressreport"ofthecommunicativebehaviorofgulls.Onlyrecently,Sober(1998)developedastrongcaseforthecomparativemethod.Inordertodeterminewhetheragivenhighercapacityisoutsideaspecies"repertoire,itisnecessarytorelyuponobservationofthenonoccurrenceofthatbehaviorinthespecies.This,inturn,requiresthatmanyindividualsbeobservedinappropriate,ornatural,circumstances.Afterfindingphenotypicvariationamongtaxa,andpro•ducingseveraladaptiveexplanationsforthisvariation,theseexplanationscanbetestedbypredictingtheenvironmentalorconstitutionalcorrelatesofthevariation,andcomparingancestralandderivedcharacterstateswhereverpossible(HarveyandPagel,1991).Asaconsequence,ithasbeenrepeatedlydemonstratedthatbe•havioralcharacteristicshavefunctionsthatdifferconsiderablyfromthoseforwhichtheyevolved.Evolutionary,functional,anddevelopmentalconstraintsuponpheno•typicevolutionposefundamentalproblemsforthevalidityofpureadaptionistexplanations.Therefore,adaptationdoesnotserveasaproperguidetounderstandingthebigtrendsinevolution.Thesyntheticviewofevolution,whichisbasedupontheprin•ciplesofvariationandselection,cannotexplainmacroevolutionaryphenomena.Systemstheory(Riedl,1978),ontheotherhand,takesphylogeneticconstraintsintoaccountandconsidersthelimitationsofevolution.Itstatesthatevolutioncannotbereducedtogeneticevolution.Anygivensetofphenotypescontainsonlyasmallsubsetofthecharactercombinationsthatcouldbeselectediftheywereproduced.Successfuladaptabilityattheexpenseofflexibilityisthereasonwhytherearesofewsystemictypes.Theonlypossiblewayinwhichtheenvironmentcaninfluencepopu•lationdynamicsisbyselectionactingonthephenotype.Theinherentsystemiccon- Psychophylogenesis37ditionsoftheorganismrestrictthenumberofpossibletransformationsamongpop•ulationstates,eveninthosecaseswherethesetransformationscorrespondtoselectivepressuresandgeneticpossibilities.Thecrucialpointhereisthatmanygenotypesfailtogenerateaviableandfertilephenotypeindependentlyoftheenvironment.Therefore,eveninthosecasesinwhichitispossibletodescribehistoricalenvironmentalcon•ditions,itisinsufficienttodescribethepathsofphenotypicevolution.Whatweneed,then,is,complementarytotheexternalfactors,acloserunder•standingoftheinner,systemicfactorsinfluencingevolution.Itcanbedescribedasafeedbackloopthatoperatesattheinterfacebetweenthegenotypeandthephenotype,andwhichmaintainsorganization,ratherthanadaptation.Thefeedbackloopleadstocoherencewithintheorganism,andthiscomplementstheorganism"scorrespon•dencewithitsenvironment.Asaresult,structures,functions,andpathwaysbeyond,andofteninoppositionto,newfunctionalrequirementsbecamefixatedorcanalized.Thisisthereasonwhyacephalopod"feedsthroughitsbrain,"andwhyadolphinwillneverbecomeafishandabatwillneverbecomeabird,eventhoughselectivepressuresarisingfromtheenvironmentmayactinthesedirections.Fromanadap-tionistorstrictlyengineeringpointofview,itisparadoxicalthatoureyesareinverted(largeganglioncellssitinfrontofthecellsthatreceivetheincominglight),andthatdolphinsandgiraffeshaveexactlysevencervicalvertebra(likeallothermammals)whenarespectivedecreaseandincreaseinnumberwouldbeafunctionaladvantage.Moregenerally,thoseaspectsofmorphologythatvaryamongtaxacanremainconservativewithinataxonevenwhenmembersofthetaxonliveindistinctlydif•ferentenvironments.Theaboveexampleswereselectedbecausetheyillustratethatfunctionally"better"solutionsarestillpossibleinnature.Cephalopodshave"cor•rect"eyes;fishesareabletodispensewithcervicalvertebra,whereasdinosaurspos•sessedhundreds.Thephylogeneticconsequencesofselectionforsystemiccoherencearethecanalizationofadaptivetraits,theexistenceofpathwaysinphylogeny(markedbyhomologies),andtheestablishmentofhighertaxonomicclassessharingthesamebasiccharacteristics("Bauplan").Therefore,acompleteunderstandingofthestructuralorfunctionalphenotypespresentduringanygivenphylogeneticstagecannotbeachievedbymappingassumedenvironmentalconditionsontogenotypes.Wecanapplytheseevolutionaryprinciples,derivedfrommorphology,tocognitivemechanismsintwoways.First,thespaceinwhichcognitivemechanismsevolveislimitedandstructuredbyconstraintsactingontheorganism.Therefore,onecanunderstandtheevolutionofcognitivemechanisms,atleastinpart,byunderstandingtheconstraintsactingonthem.Thisideaisclearlyevidentatthemolecularlevel.Cognitionisafunctionofthebrain,whichisacollectionofanenormousnumberof 38LudwigHubercells.Thebrainisacombinatorialsystem.Itnotonlyopensapotentialuniverseofbehavioralpossibilitiesforanorganism,butalsoimposessignificantconstraints.Forexample,specificfeaturesofregulatorypatternscannotbealteredwithoutalteringothersbydefault.Second,athigherlevelsoforganization,thenumberofwaysinwhichcognitiveevolutioncaninteractwiththeenvironmentislimited.Consequently,thereareonlyafewpathwaysalongwhichtheevolutionofacognitivefacultycanproceed.Thesepathwaysaretheresultofthesuccessivemodificationofthestructuresresponsibleforbehavior.Anygivencognitiveinventionistheresultofthemodificationofpre•existingstructures(prerequisites),whicharethenabletoopenecologicalnichesinwhichanewspeciescanexploittheadaptivedeficienciesofotherspecies.Meta•phoricallyspeaking,wearedealingherewiththeadventureoflife,whichtakesriskandactivelyseeksnew,unknownconditions.Thisprocessplaysanimportantroleinhigherdevelopment.Itdoesnotexplainhigherdevelopmentfully,butdoesplayadecisiverole.Manycognitivephenomenacannotbeinterpretedinadaptivetermsbecausetheymaynotactuallybeadaptive.Furthermore,manycognitivemechanismsappeartobeadaptiveeventhoughtheyevolvedunderfundamentallydifferentcircumstancestothoseinwhichtheynowoperate."Homoiology"(analogyonahomologousbasis;seeRiedl,1978)isanimportantprincipleincognitiveevolution.However,constraintsresultingfromthepreservationoffunctionalmodulescanpromotetheevolutionofadaptivecomplexitybyenablingtheestablishmentofcomplexhierarchicallyorga•nizedsystems.Thefixationorconservationofbrainstructuresorcognitivemodulesdoesnotnecessarilymeanahaltinevolution.Extraordinaryqualitiesmayemergefromthecombinationofoldcomponents(astheaforementionedcellalphabetforsimpleformsoflearning),ortheirgradualmodificationmay"create"newproblemspaces.Owingtothelimitedamountofspaceavailable,Icanonlydescribeoneexampleoftheevolutionofacognitivefacultytoillustratethispoint.Moore(1996)proposedtwodifferentphylogeneticpathwaysleadingtotheevolutionofimitativelearning.Oneisfoundinbirds,whereitoriginatesinsonglearning.Fromsonglearningthispathwaymovestovocalmimicry,percussivemimicry,visualmovementimitation,andfinallycross-modalmatching.Theotherpathway,foundinprimates,hasitsoriginsinskilllearning,fromwhichitmovestovisualmovementimitationandfinallycross-modalmatching.Theheuristicvalueofthisevolutionaryscenarioisthatitcanbeempiricallytested.Forexample,inourlaboratoryinViennawearecurrentlyexaminingwhetherkeas(Nestornotabilis)areabletoimitatethebodymovementsofskilledgroupmembers.Keasareknownfortheirextraordinary Psychophylogenesis39manipulativeabilitiesandfortheirrichsociality.However,theyhaveneverbeenshowntoimitatesounds,andtheydidnotevolvefromanyknownvocalmimics.ConclusionCognitionisoneofthemostchallengingtopicsforthenaturalsciences.Inhumans,cognitiveprocessesaretheresultofanenormouslyintricatebrainsystemorganizedatmanydifferentlevels,andconsistingofapproximatelyonemillionbillioncon•nectionsinthecorticalsheetalone.Thesheerintricacyandsizeofthebrainmeritsmorethanonekindofscientificinquiry.Furthermore,giventhatcognitionreferstotheabilityoflivingcreaturestoadaptivelymodifytheirbehaviorinordertodecidewhattodonext,anevolutionarytheoryofhowwecametohavemindsmustbeanessentialpartofthecognitivesciences.Weasked,howdidcognitiveprocessesevolve?Manydifferentkindsofbehaviorscanbeadaptivelymodifiedbyindividualexperi•ence(e.g.,foodfinding,matechoice,homeostasis).Therefore,weexpecttofinddiversityintheinformationthatcontrolsbehavior,regardlessofwhetheritcomesfromtheinternalortheexternalenvironment,orfromthebehavioritself.Inotherwords,therearelikelytobedifferencesinthe"what/whenrules"ofcognition(Heyes,thisvolume).AccordingtoLorenz,notonlylearnedbehavior,butalsotheabilitytolearn(e.g.,avoidancelearninginparamecia,"reasoning"inbonobos)emergedinevolution.Itseemscompletelyimplausibletosuggestthatthe"howrules"(Heyes,thisvolume)underlyingtheseabilitiesareallthesame.Nevertheless,fromasystemic,macro-evolutionarypointofview,whichtakesintoaccountwhatweknowaboutthemolecularandcellularbasisoflearning,majorconstraintsariseatthelevelofsingleneurons.Theseconstraintsmayhaveincreasedprogressivelyforhigherlevelsofanalysis(e.g.,synapticconnections,entireneuronalnetworks).Becauseprocessingoperationsareestablishedatthelevelofneuronalcircuitryfromwhichcomplexadaptivebehavioremerges,howrulesshouldhavebeeninventedonlyrarelyinevo•lution,oratleastmuchlessfrequentlythanwhat/whenrules.Adefinitiveanswertothequestionofhowcognitiveprocessesevolvedisnotcur•rentlyavailable.Toanswerthisquestionfromapurelyfunctionalistpointofviewwouldbeimpossible.Anunderstandingoftheevolutionofcognitioncanonlybegainedbyreconstructingthesequenceofconstraints,prerequisites,andinnovations,andbytakingintoaccountthebehavioralcontextofabroadspectrumofspecies,asLorenzdid.Unfortunately,suchendeavorshavereceivedonlyscantattentionincognitivesciencesofar. 40LudwigHuberAcknowledgmentsIamgratefultoRupertRiedl,CeciliaHeyes,SverreSjolander,WolfgangSchleidt,andGerhardMedicusfortheircommentsonearlierversionsofthemanuscript.IalsothankMonikaKickertforallowingmetorummageinLorenz"slibrary.ReferencesAtzJW(1970)Theapplicationoftheideaofhomologytoanimalbehavior.In:Developmentandevolu•tionofbehavior:EssaysinhonorofTCSchneirla(AronsonLR,TobachE,LehrmanDS,RosenblattJS,ed),pp53-74.SanFrancisco:Freeman.ByrneRW(1994)Theevolutionofintelligence.In:Behaviourandevolution(SlaterPJB,HallidayTR,ed),pp223-264.Cambridge:CambridgeUniversityPress.CampbellDT(1974)Evolutionaryepistemology.In:ThephilosophyofKarlR.Popper(SchlippPA,ed),pp413-463.LaSalle,IL:OpenCourt.DarwinC(1871)Thedescentofmanandselectioninrelationtosex.London:Murray.FreemanWJ(1995)Societiesofbrains.Hillsdale,NJ:Erlbaum.GassCL(1985)Behavioralfoundationsofadaptation.In:Perspectivesinethology,vol.6(BatesonPPG,KlopferPH,ed),pp63-107.NewYork:PlenumPress.GreenspanRJ,TullyR,Alvarez-BuyllaA,BenjaminPR,BorstA,FischbachK-F,HallJC,KupfermannI,RothG,SokolowskiMB,StraussR,TrumanJW(1994)Groupreport:Howdogenessetupbehaviors?In:Flexibilityandconstraintinbehavioralsystems(GreenspanRJ,KyriacouCP,ed),pp65-80.Chichester:Wiley.HarveyPH,PagelMD(1991)Thecomparativemethodinevolutionarybiology.Oxford:OxfordUniver•sityPress.HawkinsRD,KandelER(1984)Isthereacell-biologicalalphabetforsimpleformsoflearning?Psycho•logicalReview,91:375-391.HennesseyTM,RuckerWB,McDiarmidCG(1979)Classicalconditioninginparamecia.AnimalLearn•ingandBehavior,7:417-423.HodosW(1970)Evolutionaryinterpretationofneuralandbehavioralstudiesoflivingvertebrates.In:Theneurosciences(SchmidtFO,ed),pp26-39.NewYork:RockefellerUniversityPress.HodosW,CampbellCBG(1969)Whythereisnotheoryincomparativepsychology.PsychologicalReview,76:337-350.HuberL(1995)Onthebiologyofperceptualcategorization.EvolutionandCognition,1:121-138.KelsoJAS(1995)Dynamicpatterns.Cambridge,MA:MITPress.LewontinRC(1998)Theevolutionofcognition:Questionswewillneveranswer.In:Aninvitationtocognitivescience,vol.4:Methods,models,andconceptualissues(ScarboroughD,SternbergS,ed),pp107-132.Cambridge,MA:MITPress.LorenzK(1941)VergleichendeBewegungsstudienanAnatiden.JournalfurOrnithologie,89:194-293.LorenzK(1974)Analogyasasourceofknowledge.Science,185:229-234.LorenzK(1977)Behindthemirror.London:Methuen.MartinsEP,ed(1996)Phylogeniesandthecomparativemethodinanimalbehavior.NewYork:OxfordUniversityPress.MaturanaH,VarelaF(1979)Autopoiesisandcognition.Dordrecht:Reidel. Psychophylogenesis41MaynardSmithJ,HollidayR(1979)Preface.In:Theevolutionofadaptationbynaturalselection(May-nardSmithJ,HollidayR,ed),ppv-vii.London:TheRoyalSociety.MayrE(1974)Behaviorprogramsandevolutionarystrategies.AmericanScientist,62:650-659.MenzelJrEW(1986)Howcanyoutellifananimalisintelligent?In:Dolphincognitionandbehavior:Acomparativeapproach(SchustermanRJ,ThomasJA,WoodFG,ed),pp167-181.London:Erlbaum.MooreBR(1996)Theevolutionofimitativelearning.In:Sociallearninginanimals(HeyesCM,GalefJrBG,ed),pp245-265.SanDiego:AcademicPress.PlotkinH(1997)Evolutioninmind.London:Penguin.PlotkinHC,Odling-SmeeFJ(1981)Amultiple-levelmodelofevolutionanditsimplicationsforsocio-biology.BehavioralandBrainSciences,4:225-268.PovinelliDJ(1993)Reconstructingtheevolutionofmind.AmericanPsychologist,48:493-509.RenschB(1967)Theevolutionofbrainachievments.EvolutionaryBiology,1:26-68.RiedlR(1978)Orderinlivingorganisms.Chichester:Wiley.RiedlR(1995)Deficienciesofadaptationinhumanreason:Aconstmctivisticextensionofevolutionaryepistemology.EvolutionandCognition,1:27-37.RistauCA,ed(1991)Cognitiveethology.Hillsdale,NJ:Erlbaum.SchrodingerE(1944)Whatislife?Cambridge:CambridgeUniversityPress.SjolanderS(1993)Somecognitivebreakthroughsintheevolutionofcognitionandconsciousness,andtheirimpactonthebiologyoflanguage.EvolutionandCognition,3:1-10.SoberE(1998)Morgan"scanon.In:Theevolutionofmind(CumminsDD,AllenC,ed),pp224-242.NewYork:OxfordUniversityPress.SpencerH(1855)Principlesofpsychology.London:Longman.ThomasRK(1980)Theevolutionofintelligence:Anapproachtoitsassessment.Brain,BehaviorandEvolution,17:454-472.TimberlakeW(1993)Animalbehavior:Acontinuingsynthesis.AnnualReviewofPsychology,44:675-708.TinbergenN(1959)Comparativestudiesofthebehaviorofgulls:Aprogressreport.Behaviour,15:1-70.WhitmanCO(1898)Animalbehaviour.WoodsHole,MA. 3ModularityandtheEvolutionofCognitionSaraShettleworthWhenconditionsarepredictableacrossgenerationsbutunpredictablewithingener•ations,mechanismswillevolvethatalloweachindividualtoadjustitsbehaviortothedetailsofitsownenvironment(RichersonandBoyd,thisvolume;Huber,thisvol•ume).Forinstance,abeepredictablyfindsfoodsourcesthatareworthrevisiting,buttheirpreciseappearanceandlocationareunpredictableinadvanceofexperiencebyanindividualbeeorcolonyofbees.Suchpredictableunpredictabilitymakesitadap•tiveforindividualstobeabletosenseandretaininformationthatidentifieslocalprofitablefoodsourcesandtousethisinformationtorelocatethem.Similarmoreorlessinformalargumentscanbemadeinfavorofanimalsbeingabletoprocess,store,andactoninformationaboutphysicalandsocialcausation,time,rateofoccurrence,andotherfeaturesoftheworld(Dyer,1998;Shettleworth,1998).Atthesametime,however,speciesclearlydifferenormouslyinthevarietyofsensoryinformationavailabletothemandthewaysinwhichtheycanprocessandactonit.Forinstance,considertheone-celledorganismstentor,whichcanreacttovariousobjectsinitsenvironmentwithingestion,rejection,orescape,inaverysimpleyeteffectiveway(seeStaddon,1983).Tosomeextent,thedifferencesamong,say,stentorandabeeandachimpanzeearetheproductsofparticularevolutionaryhistoriesandecologicalniches.However,afullaccountoftheevolutionofcognitionshouldembraceallmechanismsthatinvertebratesandvertebrateshavefortakingininformationthroughthesenses,retainingit,andusingittoadjustbehaviortolocalconditions.Indiscussionsofcognitioninnonhumans,itisusualtotrytodistinguishbetween"cognitive"andothermind/brainprocesses.Forinstance,cognitionissaidtoinvolveexplicitrepresentationsofabsentstimuli(Terrace,1984),ordeclarativeasopposedtoproceduralknowledge(McFarland,1991).Reviewsofthecomparativepsychologyoflearningtypicallyplacelearningprocessesinahierarchy:habituation,thevariousformsofassociativelearning,and"highercognitiveprocesses"suchasconceptfor•mation,counting,andlanguage(e.g.,Thomas,1996).Amoreproductiveapproachtothinkingabouttheevolutionofcognitionistoconstruecognitionasinformationprocessinginthebroadestsense,fromgatheringinformationthroughthesensestomakingdecisionsandperformingfunctionallyappropriateactions,regardlessofthecomplexityofanyinternalrepresentationalprocessesthatbehaviormightimply(seeShettleworth,1998).Tolookmorenarrowlyrisksexcludingbehaviorsthatareproperlypartofacompletecomparativeandevolutionaryanalysis.Toseewhythisisso,considerbehaviorsthatappearatfirstglancetoinvolve"highercognitivepro•cesses"suchasplanningandforesightbutthatactuallymaynot.Forinstance,wildchimpanzeesinsomepartsofAfrica"fish"fortermitesbypokingsticksorgrass 44SaraShettleworthbladesintotermitemoundsinsuchawaythatthetermitesclingtothemandcanbeextracted.Termitefishingisalearnedbehavior,perhapsevidenceofchimpanzeeculture(McGrew,1992).Ithasbeensuggestedthatyounganimalslearnbyimitationtofishfortermites,butthisisdebatable.Pokingsticksandgrassesintotermitemoundsandgettingtermitesoutmaybeaninstrumentallyreinforcedskillthatislearnedbytrialanderrorandthatdoesnotinvolveeitherspecificallysociallearningorunderstandinghowtoolswork.Butifthechimpsare"merely"performingacomplexreinforcedoperant,doesthatmeanthattermitefishingisnolongerpartofthestudyoftheevolutionofcognition?Thisquestionisespeciallyappositeinthelightofincreasingevidencethatinstrumentallearningismediatedbysubtlerepresenta•tionalprocesses(seechaptersinthisvolumebyDickinsonandMackintosh).Asanotherexample,whenabeedepartsfromanewfoodsource,itturnsbackandfacestowardthepointofdeparturewhileloopingaroundinfrontofit.Thisbehaviormightbedescribedinwaysthatimputetothebeeconsciousforesightandpossessionofacognitivemap.However,itappearssufficienttoconcludethatduringthe"turnbackandlook"behaviorthebeeisrecordingavisualimagethatitusestorelocateagoalbymovinguntilthevisualinputmatchesitsmemorized"snapshot"(CollettandZeil,1998).Butifitturnsoutthatthebeeismatchingsnapshotsratherthanreferringtoacognitivemap,itsbehaviorisnolessanexampleofadaptiveinformationpro•cessing.Thus,tothinksensiblyabouttheevolutionofcognitionweneedtostartwithallthewaysinwhichanimalsprocessanduseinformationinnaturetoadjustindi•vidualbehaviortolocalconditions,localonaspatialandtemporalscale,fromtheanimal"slifetimedowntovariationsfrommomenttomoment.Inthisview,learningasusuallystudiedbycomparativepsychologistsisnottheessenceofcognition;norareexplicitinternalrepresentationsoftheworld,thoughpreprogrammedabilitiestoadjustadaptivelytoconditionsintheworldareinasenseimplicitrepresentationsofthoseconditions(Shepard,1994).TheapproachsketchedherediffersdrasticallyfromtheapproachoutlinedbyBitterman(thisvolume),amongotherwaysinbeinglessfocusedonassociativelearningandlessanthropocentric(Shettleworth,1998).Pro•gramslikethatreviewedbyBittermanillustrateimportantissuesthathavetobeaddressedinthisarea,butthecomparativestudyofassociativelearningisonlyonepieceofthemuchbiggerpictureofcognitiveevolution.Ifwetakeas"cognitive"allmechanismsthattakeininformationthroughthesensesandleadtobehavioraladjustmentstoconditionsthatarelocalintimeand/orspace,itfollowsthatcognitionmustbemodular.Thatis,ratherthanbeingasinglegeneralpurposecomputer,thenonhumanorhumanmindconsistsofacollectionofspecialpurposedevices,eachofwhichprocessesadistinctdomainofinputinadis•tinctandadaptivelyappropriateway.Thevarietyofwaysinwhichinformationmust ModularityandtheEvolutionofCognition45begatheredandused,evenbyarelativelysimpleanimallikeabee,istoogreatforonegeneralmechanismtodothejob.Evolutionbothfine-tunesspecies-generalmodulesandsomehowcombinestheoutputsofoldmodulesinnewwaysorproducesnewmodulestoallowanimalstosolvenewkindsofproblems.Thebestillustrationsofthisprinciplecomefromthesenses.Westartherenotonlybecausethecompara•tivestudyofperceptionispartofanybroadstudyoftheevolutionofcognition,butalsobecauseitprovidesapowerfulmodelofhowcognitionisputtogetherandvariesfromspeciestospecies.Afteroutliningsomepropertiesofsensorysystemsthatmakethemamodelofcognitivemodulesmoregenerally,Iwilldiscusswhatismeantbycognitivemodularityinmoredetailandgiveexamplesfromtheprocessingofinfor•mationabouttimeandspace.Thechapterconcludeswithsomespeculationsabouthowcognitionevolvesandwhatresearchcouldshedlightontheissuesraisedhere.Manyoftheideasinthischapterhavebeendiscussedbyothers(e.g.,SherryandSchacter,1987;HirschfeldandGelman,1994;Shepard,1994;Sperber,1994;CosmidesandTooby,1994;Gallistel,1995;Gigerenzer,1997).Theyaredevelopedingreaterdepthinarecentbook(Shettleworth,1998).PerceptionasaModelofCognitiveEvolutionAtthebasiclevelofgatheringinformationabouttheworld,itisclearthatdifferentorgansareneededfordifferentjobs.Eyes,ears,noses,tongues,electroreceptors,andsoonareallseparatemodulesthatgatherinformationaboutdifferentkindsofphysicalenergiesandprocessitindifferent,functionallyappropriate,ways.Thedif•ferentsenseorgansandtheirassociatedcentralprocessingareas,aswellassub-processeswithinthemsuchascolor,depth,andmotionperceptionwithinprimatevision,areanatomicallyaswellasfunctionallydistinct.However,theprimarycrite•rionformodularityinmostdiscussionsofcognitionisfunctional,withdifferentrulesofoperationinferredfrombehavior.Nevertheless,thefirstreasonwhyperceptionisamodelofcognitiveorganizationandevolutionasawholeisthatwithintherealmofcognitionitprovidesanillustrationofthegeneralbiologicalprincipleofseparateorgansforseparatejobs.Asecondprincipleisthatdifferentstructuresmaydothesamejobindifferentlineages.Consider,forinstance,thedifferencebetweenthecompoundeyesofbeesandthecamera-likeeyesofbirdsormammals.Hypotheticalcomputationalstruc•turesobeythisprinciplejustasdodifferentkindsoflight-sensingorgans.Wewillseeexamplesbelow.Asinusinggas,electricity,orwoodforcooking,allareequallyeffectivewaysofgettingdinner,butlocalandhistoricalconstraintsdeterminewhichaparticularcookuses. 46SaraShettleworthAthirdprincipleofmodularityisthatsharedstructuresmaybespecializedineachspecies.Thisisobviouswhenitcomestophysicalorgans.Forexample,birdsallsharebeaks,eyes,wings,andfeet,amongotherstructures.Yetabeakmaybeaprobe,ahammer,anutcracker,oraflesh-tearingtool,amongotherpossibilities.Feetmayhavewebs,lobes,talons,orclaws.Eyesandthephotoreceptorswithinthemmaybeplacedtodetectmovingpreyfromadistance,tolocalizesmallseedswhilesimultaneouslywatchingforpredators,andsoon(Lythgoe,1979).Theinformationeachspeciestakesinandthewaysinwhichitisusedarepartofapackagethatmakessenseintermsoftheproblemstobesolvedinthewild.Linkedadaptivemodificationsofsharedstructuresarealsoreflectedintherelativesizesofbrainareasprocessingdifferentkindsofsensoryinformation.Forinstance,complexrelation•shipsbetweenlifestyleandrelativesizesofvisualandolfactoryprocessingareasmaybeseeninprimates,insectivores,andbats(Bartonetal.,1995).Theprinciplessug•gestedbytheadaptivespecializationsofthesensesandtheirassociatedbrainareasarenolessapplicabletomore"central"informationprocessing,asisillustratedbyadaptationsofhippocampusinfood-storingbirdsandotheranimalsthatnaturallyfaceextraordinarydemandsonspatialmemory(Sherryetal.,1992).Thehippo•campusislargerrelativetotherestofthebraininsuchanimalsthanincloserelativesthatdonotstorefoodorhavelargeterritories.Afourthwayinwhichperceptionisamodelofcognitiveorganizationmoregen•erallyisthatinadditiontospecializations,therearegeneralprocesses,evenonesthatcutacrossmodules.Withineachlineage,aparticularsenseorgansuchasaneyeoranearisstructuredinthesamegeneralwayinallspeciesevenwhileitisadaptivelytweakedineachone.Thesensesshareanumberoffunctionalproperties,suchassensitivitytocontrast,Weber"slaw,atendencytoadaptorhabituate,andatendencytogiveabiggerresponsetoaphysicallybiggerstimulus(Barlow,1982).Thislastpropertyseemssoobviousthatitiseasytoforgetitisnotalogicalnecessitybutaverygeneraladaptationtothefactthatsomethingbiggerorcloserislikelytowarrantfasterandmoredecisiveaction.Finally,thesensesareamodelofcognitivemodularitybecausetheyexhibitpre-functionaladaptiveorganization.Ananimaldoesnotwaittosomehowdiscoverthatcertainwavelengthsorvolatilecompoundsareimportantinitslifebutcomesready-madewithsensesatleastroughlytunedtothekindsofenergiesitisgoingtoneedtoreacttoinitsspecies-specificniche.Thisdoesnotmeanexperienceisunimportant,buteventhecapacitytobemodifiedbythatexperienceinaparticularwayispartofthepackage.Notonlythefine-tuningofsensorysystemswithexperiencebutalsoappropriateprocessingofinformationrelevanttocircadianrhythms,deadreckoning,associativelearning,imprinting(Bateson,thisvolume),andsoon,isbuiltin. ModularityandtheEvolutionofCognition47Insummary,thecomparativestudyofperceptionisbothpartofcognitionandamodelforhowtoviewcognitionifwewanttounderstanditsevolution.Thestudyofsensationandperceptionismorecloselytiedtoconsiderationoffunctionthanisanyotherpartofcognitivescience(Marr,1982;Shepard,1994).Furthermore,compara•tiveandevolutionaryworkinthisareaisprobablybetterdevelopedthanisthestudyofanyotheraspectofcognitiveevolution(e.g.,Lythgoe,1979;Dusenbery,1992).CognitiveModularityinNonhumansWhatIsaModule?Theideathatcognitionismodularisnotnew,ascontemporarypsychologistsdis•cussingthemodularityofmindarewellaware(e.g.,Sperber,1994;Gigerenzer,1997).ThepaperbackeditionofFodor"s(1983)TheModularityofMindhasaphrenologicalheadonthecover,andarecentreview(LippandWolfer,1995)ofspecializationsinrodentbrainshas"microphrenology"initstitle,buttheideaofseparatepsychologicalfacultiespredatesthenineteenthcentury.ThecontemporarymetaphorformodularityisthemindasaSwissArmyknife(Wilson,1994;Mithen,1996),ananalogymeanttoshowthatthemosteffectivekindofgeneralinformationprocessorismadefromacollectionofspecializedparts,eachofwhichdoesapar•ticularjobverywell.Paradoxically,adaptivemodularityandtherelatedissueofseparablememorysystems(SherryandSchacter,1987)aredebatedmoreamongstudentsofhumanpsychology(cf.HirschfeldandGelman,1994)thanamongcom•parativepsychologists.Thedebateaboutgeneralprocessesasopposedtoadaptivespecializationsinanimalworkhasdealtmostlywithspecies-orcontent-specifictweakingsofassociativelearningmechanisms,asintaste-aversionlearning(seeBitterman,thisvolume),ratherthantheoverallmodularorganizationofanimalinformationprocessing,whatspecialissuesitraises,andhowtheymightbetackled.Yetsomeoftheclearestevidenceforcognitivemodularitycomesfromanimalbehavior.Moreover,onlywithnonhumanspeciesisitpossibletobringtogethercomparativedatafromecology,phylogeny,brainstructure,andbehaviortoaddressissuesaboutthefunctionandevolutionofmodularorganization.Whendistinctclassesofinput(domains)arecomputedonindistinctwaysasinferredfrombehavior,wehaveadistinctmentalmoduleormemorysystem.Com•putationaldistinctivenessistheprimarycriterionforcognitivemodularity,althoughotherssuchasanatomicalseparabilityarecommonlyassociatedwithit.However,anidentifiablecognitivemoduleneednotbelocalizedinasingleplaceinthebrain.Coversely,anatomicallydistinctsubstratesforcognitionneednotimplyinteresting 48SaraShettleworthcomputationaldistinctions.Forinstance,associativelearningcouldbesubservedbycircuitslocaltoparticularstimulusandresponsesystemseventhoughasinglesetofprinciplescoulddescribethatlearning.Associativelearninginhoneybeesandratsishardlylikelytoresideinsimilarstructuresdespiteitsfunctionalsimilarityintheseandotherspecies(Bitterman,thisvolume).TouseHeyes"s(thisvolume)terms,indenningcognitivemodules,"how"ismoreimportantthan"what,"eventhoughultimatelytherequirementsof"what"istobeprocesseddictate"how."Becausemodularityisseenasanevolvedadaptationforefficientinformationprocessing,cognitivemodulesaregenerallythoughtofasinnate.Indeed,evolution•arypsychologyisoftencontrastedwithapsychologythatseesthehumanmindasinfinitelymoldedbyculture(CosmidesandTooby,1994).However,thereisnologi•callynecessaryconnectionbetweeninnatenessandmodularity.Thedegreetowhichaparticularaspectofinformationprocessingisinfluencedbyexperience(i.e.,howitdevelops),whetherornotitiscomputationallydistinctive,whatinformationitdealswith,andwhatitscurrentorpastadaptivevaluemightbeareallseparablequestions(seeShettleworth,1998;Heyes,thisvolume).Experiencecouldmodifyprefunction-ally("innately")organizedcognitivemodulesincharacteristicways,presumablybyrulesintrinsictoeachmodule(seeBateson,thisvolume,foracandidate).Fodor"s(1983)criteriaformodularityincludeinformationalencapsulation:amoduleworksbyitself,disregardinginformationthatmaybeprovidedbyothermodules,evenwhentheresultisstupidbehavior.Visualillusionsareonekindofevidenceforthis.Someanalogsinanimalbehavioraredescribedbelow.Inhumans,thedegreeofmodularencapsulationmaychangeduringdevelopment(HermerandSpelke,1996;seebelow).Oneprovocativeviewofthehumanmind(Mithen,1996)suggeststhatcognitivemodularityhasbeenreducedduringevolution,allowingmorefluidcommunicationamongdistinct"intelligences."Consideringotherpossiblydenningfeatures,evidencethatacandidatemodulecanevolveindependentlyofothermodulesispotentiallyrevealing,asintheadaptivetweakingofolfaction,vision,andothersensesfordifferentniches.Inanimalbehavior,however,someofthebestillustrationsofcognitivemodularityinvolvesuperficiallysimilarinputsbeingpro•cessedindifferentways.Thenexttwosectionsdescribeexamplesfromtemporalandspatialcognition(seealsoGallistel,1995;Shettleworth,1998).ModularityinTemporalInformationProcessingNearlyeveryorganism,includingplants,bacteria,andhumanbeings,hasasystemforadjustingitsbehaviorandphysiologytolocaldayandnight,thecircadiantiming ModularityandtheEvolutionofCognition49system.Atleastinvertebrates,circadiantimingcanbecontrastedwithintervaltiming,whichallowsanimalstoanticipateeventsonascaleofsecondstoafewhours.Somesuggestordinaltiming,encodingeventsinasequence,asathirdtimingsystem(CarrandWilkie,1997).Circadiantimingisnotusuallyconsideredpartofcognition,butitisaninstructiveexampleofpreprogrammedadaptivebehavioralplasticity.Ithassomeverywell-describedproperties(Moore-Edeetal.,1982).Cir•cadiantimingallowsananimaltobeactiveatthespecies-appropriatephaseofthe24hourlight-darkcycle.Accordingly,theprimaryinputtothecircadiansystemislight.Organismsinconstantlightordarknesstypicallyexhibitanendogenous,free-runningrhythmofslightlymoreorlessthan24hours.Lightpushesorpullsthisintowhatwouldnormallybesynchronywithlocaldayandnight.Theeffectsoflightorotherentrainingstimuliaredescribedbyaphaseresponsecurveshowinghowmuchtherhythmisdelayedoradvancedasafunctionofwhenastimulusisappliedinthe24-hourcycle.Forexample,foranocturnalanimalsuchasagoldenhamster,apulseoflightjustastheanimalentersitsactivephasedelaystheonsetofactivitythenextday(asiftheanimalgotuptooearly).Incontrast,apulseoflightlateintheactivephaseadvancesactivitythenextday(asiftheanimalstayedactivetoolateinthemorning).Thecircadianrhythmhasalimitedrangeofentrainment:itwillsynchro•nizeonlytostimulihavingaperiodicitycloseto24hours.Becausethecircadiansystemcyclescontinuously,itprovidesonlyphaseinformation,thatis,informationabouthowlong(intermsofthephaseangleofthecycle)oneeventisbeforeorafteranother.Itsoutputisthereforevalidonanintervalscaleofmeasurement(CarrandWilkie,1997).Incontrast,intervaltimingallowsbehaviortoanticipatearbitraryeventsthatrecuroverarbitraryintervalssecondstohourslong,aswhenanimalstrackthetimesbe•tweenpreycaptureswhenforaging.Inthelaboratory,thepropertiesofintervaltimingarestudiedbyaskinganimalstotimetheintervalsbetweenreinforcersorthedurationsoflightsandtones(reviewinShettleworth,1998).Inthebest-establishedmodelofintervaltiming,animalsacquirerepresentationsofthedurationsofimpor•tanteventstowhichnewlyexperienceddurationsarecompared(Gibbonetal.,1997).Theserepresentationsareblurred,inawaydescribedbyWeber"sLaw.Thatis,shortintervalsaretimedwithsmallerabsoluteerrorthanlongerones,andtheerrorisproportionaltotheintervalbeingtimed.Unlikecircadiantiming,intervaltiminghasanaribitraryzeropoint(thebeginningoftheeventbeingtimed),anditsoutputcanberepresentedonaratioscaleofmeasurement.Becauseordinaltimingrepresentsonlywhetheroneeventisbeforeorafteranother,inthisanalysis(CarrandWilkie,1997),intervaltimingisthecomputationallyrichesttimingsystem. 50SaraShettleworthTosummarize,circadianandintervaltiminghavedifferentadaptivefunctions,andtheydifferaccordinglyinotherrespects.Effectiveinputstothecircadiansystemarelightandafewotherstimuliwithperiodicitiesaround24hours,comparedtostimulifromarangeofmodalitieswitharbitraryperiodicitiesfromsecondstohoursfortheintervaltimingsystem.Therelationshipbetweeninputandoutputisdescribedforcircadiantimingbyaphaseresponsecurve,whereasintervaltimingismediatedbyastoredrepresentationoftheintervaltosomeimportanteventassociatedwithapar•ticularsignal,muchasinotherformsofassociativelearning.Incontrast,acquisitioninthecircadiansystemisdescribedasentrainmentofacontinuouslyrunningendogenousrhythm.Anentrainingagentisnotencodedassuchbutratherisrepre•sentedinadjustmentsoftherhythmoveratmostafewdaysafteritsapplication.However,likealearningcurveintheRescorla-Wagnermodelofassociativelearning,phaseresponsecurvesmayreflectthediscrepancybetweenactualand"expected"events.Intheexampledescribedearlier,lightinthemiddleofthehamster"ssubjec•tiveday(evidencedbyitsfree-runningactivityrhythminconstantdarkness)haslessofaneffectthanlightanhourortwofromthebeginningorendofitssubjectivenight,whenlightmightbethoughtofasmore"unexpected."Circadianandintervaltimingareneuroanatomicallyandpharmacologically,aswellascomputationally,distinct.Inmammals,thecircadianclockislocatedinthesuprachiasmaticnucleusofthehypothalamus,whereasthecerebellum,frontalcortex,andbasalgangliaarethoughttobeinvolvedinintervaltiming(reviewinGibbonetal.,1997).Circadianrhythmsareveryancientandphylogeneticallyverywidespread,probablymoresothanintervalorordinaltiming—whichhave,however,beentestedinmanyfewerspecies.ModularityinSpatialInformationProcessingMysecondexampleofcontrastingmodules—thoseallowinganimalstogetbacktoarememberedplace—ismoreconventionallycognitive.Spatialcognitioninvolvesdifferentkindsofimplicitcomputationsonanumberofdistinctkindsofinforma•tion(Gallistel,1990;Shettleworth,1998).Perhapsthemostbasicandprimitiveway-findingmechanismislayingdownatrailandfollowingitbacktoastartingplacelikeanestorotherplaceofrefuge.Here,informationaboutwheretogoisdepositedintheenvironment.Thetrail-followerneedonlydetectthetrailanddeterminedirectionrelativetoit.Neitheroftheseaccomplishmentsencodesthemetricpropertiesofspace,anglesanddistances.Cognitivelymoreinterestingisdeadreckoning,asystemthatisphylogeneticallyverywidespread.Inawell-studiedexample,adesertant ModularityandtheEvolutionofCognition51(Cataglyphisfortis)leavesitsnestandwandershereandthere,inatwistingpath,untilitencountersanitemoffood.Itpicksupthefoodandheadsstraightbacktothevicinityofthenest,whereuponitbeginstocirclearounduntilitlocatesthenestentrance(CollettandZeil,1998).Similarly,ahamsterforaginginanovelarenaintotaldarknesstakesacircuitouspath,butwhenfrightenedbyasuddennoise,headsbacktoitsnestinastraightline(Etienneetal.,1998).Theabilityoftheantorthehamstertoheadstraightforthenestatanymomentimpliesthattheanimalsarecontinuouslyrecordingsomecorrelate(s)ofthedistanceanddirectiontraveledfromhome.Thisjobisdonewithdifferentorgansindifferentkindsofanimals.Insmallmammalslikethehamster,changesindirectionaresensedbythevestibularorgans,whereasdesertantsusethedirectionofthesuncorrectedfortimeofday,asuncompass.Geesederivedistancefromvisualflowpatterns,whereassomeinsectsusedistancewalked.Althoughdeadreckoningisoftenreferredtoaspathintegration,mostanimalsdonotuseamechanismthatliterallyintegrates.Avarietyofinsectsandmammalshavebeentestedbyallowingthemtotakeanoutwardjourneyalongapathwithtwosegmentsandthenheadhome.Whenthelengthsofthetwosegmentsoftheoutwardpathandtheanglebetweenthemarevaried,theerrorinhomewardheadingvariesinawaysuggestingthatdirectionfromhomeiscomputedaccordingtoarulethatweighseachdirectiontakenduringtheoutwardjourneybythedistanceoverwhichitismaintained.Bothmammalsandinsectsgenerallyturnmoresharplywhentakingthe"homeward"directionthantheyshould,inawaythatbringsthemacrosstheoutwardpath.Whenverydifferentkindsofmechanismsattheneurallevelproducethesamegeneralpatternoferror,itisreasonabletoaskwhetherthereissomefunc•tiontothatparticularkindoferror.Inthiscase,itmightbethatbycrossingtheoutwardpaththeanimalislikelytocomeintocontactwithlandmarksthatwillguideithome(Etienneetal.,1998).Withdeadreckoning,theanimal"spathisrecordedautomaticallyandimmedi•ately,inasingletrial.Indeed,theantorhamsterthatdidnotreturntoitsrefugeafteritsfirsttripinanewdirectionmightnotsurvivetomakeothertrips.However,deadreckoningaccumulateserror,soitneedsresettingperiodically,whentheanimalreturnstothenest.Thismakesitusefulforrepeatedtripsfromacentralplace,butlessusefulforlongjourneys.Deadreckoningcomputesdistancesanddirectionsinegocentriccoordinates,ascanbeseenfromexperimentsinwhichdesertantsarepickedupandplacedinanewlocationbeforebeingallowedtostart"homeward"(reviewinCollettandZeil,1998).Inunfamiliarbarrenterrain,antsrunapproxi•matelythecorrectdistanceanddirectiontoreachhomefromtheiroldlocation. 52SaraShettleworthLandmarkstietheanimaldirectlytotheenvironment,butmakinguseofthemrequiresimplicitcomputationsonmultipleperceivedandrememberedvectors.AclassicdemonstrationoflandmarkuseisTinbergen"s(1932/1972)studyofthediggerwasp.Tinbergenarrangedobjectssuchasacircleofpineconesaroundtheentrancetothenestwhilethewaspwasinside.Afterthewasphadmadeafewforagingtrips,Tinbergenmovedtheselandmarks.Thewaspnowsearchedforthenestinthemiddleofthepineconecircle.Differentanimalsencodeandusetheinformationfromland•marksindifferentways.Beesappeartomovesoastomatchthevisualimageoftheworldasitappearsfromagoalsuchasasourceofnectar.Animalsusingthis"snapshotmatching"mechanismsearchfartherfromorclosertothegoalwhenlandmarksaremadebiggerorsmaller,respectively.Animalswithdifferentkindsofdistanceperceptionbehavedifferently,anotherexampleofthesamefunctionbeingaccomplishedbydifferentmechanismsindifferentlineages(seeChengandSpetch,1998).Whenpigeonsaretrainedtofindgrainburiedinafixedrelationshiptooneormorelandmarks,andonelandmarkismoved,thebirdsbehaveasiftheyhaveencodedvectorsfromselfandgoaltothelandmarks.Inaconflict,aswhenjustonelandmarkismoved,pigeonsaveragevectorsindicatedbydifferentlandmarks.Theycomputedistanceanddirectioncomponentsofsuchvectorsseparately,acaseofmoduleswithinamodule(ChengandSpetch,1998).Whereasdeadreckoningapparentlygoesonautomatically,allthetime,featuresoftheenvironmentcanbeusedaslandmarksonlyafteronelearnstheirsignificance.Someattentionhasbeengiventohowanimalslearnaboutlandmarksandhow,ifatall,thislearningfollowsassociativeprinciples(e.g.,Redheadetal.,1997),butthereisscopeformoreworkonthissubject(ChengandSpetch,1998).Deadreckoningandlandmarkusequalifyasdistinctcognitivemodulesbecausethedifferentkindsofinputthattheytake(distancesanddirectionsoftheselffromastartingplacevs.vectorsinvolvingvisualorotherstimuliperceivedatadis•tance)demanddifferentimplicitcomputations(continuousupdatingofdistanceanddirectionfromastartingplacevs.adding,subtracting,and/oraveragingmultipleperceivedandrememberedvectorsor"snapshots"involvingarbitrarygoals).Land•markusealsotypicallyinvolvessomeformoflearning(possiblyassociative)aboutrelationshipsofperceptibledistalstimuliwiththegoalandperhapsalsowitheachother.Thatinformationisstoredinlong-termmemory,accessedwhencurrentland•marksmatchrememberedones(aprocessdependingonitsownsetofcomputations).Landmarkuseallowstheanimaltolocateitselfwithinaterrainthatitvisitsrepeat•edly,whereasdeadreckoningmaybeusedinitiallyinlearningthelocationsofobjectsinthatterrain(GallistelandCramer,1996).Butdeadreckoningandlandmarkusedonotexhaustthemechanismsbywhichanimalslocatethemselvesinspace(Shet- ModularityandtheEvolutionofCognition53tleworth,1998).Agoalmaybeidentifiedbyabeacon,anobjectthatcanbeper•ceivedfromadistance;theanimalneedonlylearnitssignificancethroughnormalmechanismsofclassicalconditioning.Itcanthenapproachthegoalusingasimplehill-climbingprocessinvolvingonlysignalsfromthebeacon.Anotherpossibilitywellstudiedinpsychologyisresponseorroutelearning,thatis,learningastereotypedseriesofmovesthatcarriesonebetweentwoormoreknownplaces.Ordinaltimingwouldhavearolehereinsequencingthemoves.Inaddition,disorientedrats,smallchildren,andsomeotheranimalsrespondonlytothegeometryofthespacesur•roundingthemanddisregardlandmarksorfeaturesofthesurfacesdefiningthespacethattheycanpatentlyuseunderothercircumstances(Cheng,1986;HermerandSpelke,1996).Thisuseofthegeometricmoduleparallelsevidenceofperceptualmodularityinthatapparently"stupid"behaviorresultsfromrelyingonjustonesourceofinformation.OtherCandidatesforDistinctCognitiveModulesTherearemanyothercandidatesfordistinguishableinformationprocessingmodules.Theseincludesomeaspectsofimprinting(Bateson,thisvolume)andsonglearning,whichcoexistinthebrainsofsomebirdspecies,alongwithmoregenerallearningabilities.Theyhavewellworkedoutbehavioralandneuralproperties(ClaytonandSoha,1999;Bateson,thisvolume).Imitationofseenmotorpatternsisanotherabilitythatmayappearinonlyafewplacesontheevolutionarytree(Moore,1996).Asso•ciativelearningisclearlyawidelysharedsystem,onewithanumberofdistinctsub-processes(suchasoccasionsetting,perceptuallearning,andlearningthepropertiesofreinforcers)notnecessarilysharedbyallspeciesorsystemscapableofformingsimpleassociations(Bitterman,thisvolume).Indeed,evenwithinrats,variousaspectsoflearningresultingfromexposuretosimplepredictiverelationshipsbetweenstimuliandreinforcersarelocalizedindifferentbrainareas(Holland,1997).Socialintel•ligence,orsomeaspectsofit,maybeadistinctmoduleormodulesfromphysicalintelligence,thoughexactlyhowtheseputativekindsofintelligencearerelatedisdebatable(Gigerenzer,1997;Kummeretal.,1997).Theoryofmindmaybeuniquetoafew(maybeonlyone)primatespecies,butthesubprocessesthatmakeitup,suchasrecognizingintentionalbeingsviaself-propelledmotion,detectingthedirectionofanother"sgaze,andsharedattention,maybemorewidespread(Baron-Cohen,1995;PovinelliandEddy,1996;Emeryetal.,1997).Language,reasoningaboutsocialobligations(Cosmides,1989),andperhapsconsciousness(MacPhail,thisvolume;Humphrey,thisvolume)maybeuniquetohumans. 54SaraShettleworthHowDoesCognitionEvolve?SpecializationofSharedModulesIfcognitionismodular,itcouldevolveinatleastthreeways.Mostcommonlyper•haps,sharedmodulesbecomespecialized.Examplesfrommorphologyandsensorysystemswerementionedearlierinthechapter.Asanotherexample,memoryhasgeneralpropertiesthatreflectthenatureoftheworld.Oldinformationislesslikelytobeusefulthanrecentinformation,somethingthathashappenedmoreoftenismorelikelytohappenagain,andsoon.Atthesametime,speciesmaydifferadaptivelyinwhattheyrememberbest.Givenredundantvisualcuestohiddenfood,food-storingspeciesofbirdstendtorememberpositionmuchbetterthancolorandpattern,whereasnonstoringspeciesrememberspatialandothercuesaboutequallywell(ShettleworthandHampton,1998).Inassociativelearning,too,weseespecies-specifictweakingforspecificjobs.ToborrowHeyes"s(thisvolume)termsagain,whatislearnedvariesadaptivelyacrossspecies,whereashowitislearneddoesnot.Forinstance,monkeys"fearofsnakesappearstobesociallytransmittedviaclassicalconditioning(CookandMineka,1990),buteventhough,asfarasisknown,itsabstractpropertiesarethesameasinothercasesofassociativelearning,theinputandoutputarespecies-specific.Thesightofafearfulmonkeyisnomorelikelytobeofsignificancetoaratorabirdthanthesmellofratbreathislikelytobeofimportancetoamonkey.Thesamecanbesaidofflavoraversionlearning,contrarytoitslonghistoryofbeingreferredtoasaspecialkindoflearning(seeBitterman,thisvolume).Nowadays,thegeneralityofassociativelearningbothwithinandacrossspeciesissaidtoreflectthegeneralpropertiesofphysicalcausation,acaseoftheadaptationistargumentforspecializationbeingturnedonitshead.DifferentWaysofCombiningInformationInFodor"s(1983)conception,modularityisprimarilyatthelevelofinputsystems,whereasmorecentralprocessingisgeneralacrossdomainsofinformation.Morerecentdiscussionsofmodularityindevelopmentalandevolutionarypsychology,however,seecognitionasmodularrightthroughfrominputtodecisionprocesses(seeGigerenzer,1995),andthisistheviewtakenhere.Inthisview—so-called"vertical"asopposedtoFodor"s"horizontal"modularity—animportantissueishowmodulesinteract.Inthesimplestcase,genuinelyencapsulatedmodulesareactivatedoneatatime.Eachmodulehasacharacteristictriggeringalgorithm(Gigerenzer,1997),soif ModularityandtheEvolutionofCognition55acreature"snervoussystemparsestheworldsothatnomorethanonealgorithmisevertriggeredatonce,howmodulesinteractisaproblemthatneverarises.Ingen•eral,however,evenverticalmodulescannotbecompletelyencapsulatedbecausethereneedstobeawaytoharmonizepotentiallyconflictingdecisionsfromdifferentmodules.Forinstance,whenadisorientedratortoddlerdisregardslandmarksthatitcanpatentlyuseunderothercircumstancesandrespondsonlytothegeometryofspace,itisprovidingevidenceforahierarchyofinformationuse.Ideasabouthowsuchhierarchies,orotherwaysofcombininginformation,workarenotverywelldeveloped(butseeGigerenzer,1995).Theproblemisnotunrelatedtotheclassicalethologicalproblemofhowdifferentdrivesystemsinteractincontrollingbehavior(cf.McFarlandandBosser,1993).Lookingacrossdifferentinformationprocessingsystemsandspecies,itisevidentthatthereisagreatvarietyofwaystocombineseparatesourcesofinformation,eitherwithinmodulesorbetweenthem.TheRescorla-Wagnermodelofassociativelearningdepictsdifferentsourcesofinformation,whichmightbetheoutputsofdif•ferentsensorymodules,competingforalimitedamountofassociativestrength.Evenwithinassociativelearning,however,thisisnottheonlypossibility.Forinstance,thetemporalpropertiesofconditionedstimuliandunconditionedstimulimaybepro•cessedinparallelwithotherinformation(WilliamsandLoLordo,1995).Circadiantimingisgoingalonginparallelwithotheractivitiesallthetime.Perhapsthetimeofdaywhenimportantthingshappenisstoredautomaticallywiththeirotherproperties(Gallistel,1990).Inoccasionsettingormodulation,onecuehasconditionalcontroloverhowothercuesareused,aswhenapigeonlearnsthatinoneenvironmentredmeansfoodandgreenmeansnothing,whereasinanotherenvironmenttheoppositeistrue.Configuringisanotherpossiblemodeofinteraction;compoundcuesmaybetreatedasseparateentitieswiththeirownpropertiesratherthanasthesumofparts(Pearce,1994).Forexamplesofhowdistinctcomputationalmodulesinteractduringtheuseofinformation,asopposedtoduringitsacquisition,wecanturntospatialbehavior,whichprovidescasesofparallelprocessing,averaging,hierarchicalorconditionalcontrol,andcompetitive,mutuallyinhibitory,interactions.Severalofthesepossibil•itiesareillustratedbywaysinwhichdeadreckoningandlandmarkuseinteractindifferentspecies.Thedeadreckoningdesertantalsorespondstolandmarksalongitsway.Inprinciple,deadreckoningandlandmarkusecouldgooninparallel,withtheinformationfromlandmarksinterpretedinthecontextofroughglobalpositioninformationfromdeadreckoning.However,theantsapparentlyrecallthelocalvectorassociatedwithaconspicuouslandmarkatdifferentglobalpositions.Theytaketheirdirectionfromthelandmarkforashortdistanceandthenresumethe 56SaraShettleworthdirectionappropriatefortheirpresumeddead-reckonedposition(Collettetal.,1998).Deadreckoningappearstobegoingoncontinuously,butitsoutputisinhibitedinthepresenceofwell-learnedlandmarks.Inhamsters,however,deadreckoningandlandmarksmaybeaveraged(Etienneetal.,1998).Whenahamsterhasmaderepeatedtripsfromitsnestintoacirculararenawithaconspicuouslandmarkononeside,andthelandmarkismoved90°aroundtheedge,thehamster"shomewardpathshifts,butnotbythefull90°.However,ifthelandmarkismovedhalfwayaroundthearena,solandmarkanddeadreckoningaretoodiscrepant,thenthehamsterfallsbackondeadreckoninganddisregardsthelandmark.Thatis,thereisshifttoahierarchicaluseofcues,withdeadreckoningtakingprecedence.Theanimalmaybeusingroughinformationprovidedbydeadreckoningtorecognizefamiliarlandmarks.Perhapstheevolutionarilypriormoduletakesprioritywhenoutputsconflict,butthisisonlyspeculation.EvolutionofNewModulesThecontrastsbetweenchimpanzee,bee,andstentoratthebeginningofthechaptermakeclearthatentiremodulesappearinsomelineages,makingpossiblewaysofdealingwiththeworldthatareimpossibleorverylimitedotherwise.Atthemostbasiclevel,speciesclearlydifferenormouslyinthevarietyofsensoryinformationavailabletothemandthewaysinwhichtheycanprocessandactonit.Whencog•nitiveskillshavesubcomponents,speciesmightshareoneormoreofthecomponentmoduleswithoutsharingthewholepackage.Forinstance,althoughhoneybeesbehavelikevertebratesinmosttestsofassociativelearning,theydifferinparadigmsthattaptheabilitytorepresentthequalityofrewards(Bitterman,thisvolume).Asanotherexample,iftheoryofmindhasanumberofsubcomponents,nonhumanprimatesorveryyoungchildrencouldpossessoneormorebutnotallofthem.Anintentionality(i.e.,self-producedmotion)detectorandaneyedirectiondetectormaybecommontoalotofspecies,buttheircombinationinasharedattentiondetectorlesscommon,andtheinterpretationofthatsharedattentionastheoryofmindcon•finedtohumans(Baron-Cohen,1995;HauserandCarey,1998).Intheirarticleontheevolutionofmultiplememorysystems,SherryandSchacter(1987)proposethatnewmodules(memorysystemsintheirterminology)arisewhenthereisfunctionalincompatibilitybetweenwhatexistingmodulescandoandtherequirementsofsomenewadaptiveproblem.Theyalsosuggestthatunderthesekindsofconditions,alreadyexistingsystemsmightturnouttobeexaptationsforsolvingthenewproblems.Thatis,eventhoughselectedviatheirroleinsolvingoneadaptiveproblem,existingmoduleswouldhavesomepropertiesthatcouldbeco-optedforotherproblems.ArelatedideaisRozin"s(1976)suggestionthatintelligence ModularityandtheEvolutionofCognition57evolvesthroughoriginallyspecializedmechanismsbecomingaccessibletonewkindsofinformation.Thesocialtheoryofintellect(seeKummeretal.,1997)mightbesaidtoassumeaccessibilitybecauseitimpliesthatcognitivemechanismsevolvedtodealwiththecomplexitiesofsociallifetransferedtononsocialcontent.Rozin(1976)suggeststhataccessibilityincreaseswithdevelopmentorevolution(seeMithen,1996).Acandidateexampleisthat,likerats,disorientedtoddlersrelyonthegeo•metricmoduleanddisregardlandmarks,whereasyoungadultsuseavailableland•marks(HermerandSpelke,1996).Similarly,theaccessibilityofassociativelearningtoawiderrangeofinputsandbehavioraloutputsmightchangethroughevolution.However,thisideashouldnotbeinterpretedassuggestingassociativelearningormemoryislocalizedinasingleplaceinthebraintowhichnewinputsgetconnectedoverthecourseofevolution.Moreover,totheextentthatmodularmechanismsaremoreefficientthangeneralpurposeones(andtheymaynotbe,Wilson,1994),anevolutionarypathtowardincreasedaccessibility—i.e.reducedmodularity—shouldnotbeexpected(Gigerenzer,1997;butseeMithen,1996).SummaryandConclusionsAbroadviewofcognitionasinformationprocessingofallsorts,fromsensationtodecisionmaking,inallspeciesseemstorequirethatcognitiveorganizationbemodular.Timingandspatialbehaviorproviderelativelywell-analyzedexamplesofadaptivedifferencesinprocessingofdifferentkindsofinformation.Animportantquestionforfutureresearchandtheoreticaldevelopmentinbothnonhumanandhumancognitionishowdistinguishablemodulesinteractinthewholebehavingindividual(Gigerenzer,1995).Themodular,adaptationistviewofcognitiveevolutionsketchedinthischaptercontrastswiththeideasfromtraditionalcomparativepsychologyreviewedbyBitter-man(thisvolume)inanumberofways.Itembracesnotonlyassociativelearning,butallwaysinwhichanimalsprocessandactoninformationabouttheworld.Theaimistoaccountforthewholepanoplyofevolvedmechanismsthatallowindivid•ualsofwhateverspeciestoadjusttheirbehaviortofeaturesoftheirlocalenviron•ment.Specializationandadaptationarenecessarypartsofthestory,butsoisgeneralityacrossspeciesand/orsituations.Theexpectationisnotthatasinglehier•archicalorderingofmechanismswillbefoundwithcapacitiespossessedonlybyhumansatthe"mostcomplex"or"mostadvanced"end.Likeresearchontheevolutionofotheraspectsofbehavior,asystematicresearchprogramontheevolutionofcognitionneedstoembracespeciesthatarebothclosely 58SaraShettleworthanddistantlyrelatedandexamplesofconvergenceaswellasdivergence.Broadcomparativestudiesofhowsensorysystemsandrelatedbrainareasarerelatedtoecologycanprovideamodelofhowotheraspectsofcognitiveevolutionmightbestudied(anexampleisBartonetal.,1995;seealsoLefebvre,thisvolume).Com•parisonsacrosswidelydifferentspecies,asemphasizedbyBittermanandinthepartofthischapteronspatialbehavior,mayalsodefinetherangeofphenomenatobeexplainedandmayprovideimportanttestsofadaptationisthypotheses.Atthesametime,detailedcomparisonsofcloserelativescanrevealhowdifferencesindevel•opmentalprogramsresultinspeciesdifferencesinbrainandcognition(Krubitzer,1995providesexamples)andperhapseventuallyshedlightonthemoleculareventsincognitiveevolution.Finally,althoughmyemphasishasbeentodrawasstarkacon•trastaspossiblewiththetraditionalpsychologicalapproachsketchedbyBitterman,carefulanalysesofcognitivemechanismsinthelaboratoryareneverthelesspartofthestory,alongwithnaturalhistoryandphylogeny."Howdoescognitionevolve?"impliesbothfunctionalandmechanisticquestionsaboutbehavior.Traditionallythesequestionshavebeentackledintheseparatedisciplinesofbiologyandpsychol•ogy,respectively.Asisevidentinthisbook,progresscanbemadewhenpeopleworkattheirinterface.ReferencesBarlowHB(1982)Generalprinciples:Thesensesconsideredasphysicalinstruments.In:Thesenses(BarlowHB,MollonJD,ed),pp1-33.Cambridge:CambridgeUniversityPress.Baron-CohenS(1995)Mindblindness.Cambridge,MA:MITPress.BartonRA,PurvisA,HarveyPH(1995)Evolutionaryradiationofvisualandolfactorybrainareasinprimates,bats,andinsectivores.PhilosphicalTransactionsoftheRoyalSocietyB,348:381-392.CarrJAR,WilkieDM(1997)Ordinal,phase,andintervaltiming.In:Timeandbehaviour:Psychologicalandneurobehavioralanalyses(BradshawCM,SzabadiE,ed),pp265-327.Amsterdam:Elsevier.ChengK(1986)Apurelygeometricmoduleintherat"sspatialrepresentation.Cognition,23:149-178.ChengK,SpetchML(1998)Mechanismsoflandmarkuseinmammalsandbirds.In:Spatialrepresenta•tioninanimals(HealyS,ed),pp1-17.Oxford:OxfordUniversityPress.ClaytonNS,SohaJA(1999)Memoryinavianfoodcachingandsonglearning:Ageneralmechanismordifferentprocesses?AdvancesintheStudyofBehavior,28:115-174.CollettM,CollettTS,BischS,WehnerR(1998)Localandglobalvectorsindesertantnavigation.Nature,394:269-272.CollettTS,ZeilJ(1998)Placesandlandmarks:Anarthropodperspective.In:Spatialrepresentationinanimals(HealyS,ed),pp18-53.Oxford:OxfordUniversityPress.CookM,MinekaS(1990)Selectiveassociationsintheobservationalconditioningoffearinrhesusmon•keys.JournalofExperimentalPsychology:AnimalBehaviorProcesses,16:372-389.CosmidesL(1989)Thelogicofsocialexchange:Hasnaturalselectionshapedhowhumansreason?StudieswiththeWasonselectiontask.Cognition,31:187-276. ModularityandtheEvolutionofCognition59CosmidesL,ToobyJ(1994)Originsofdomainspecificity:Theevolutionoffunctionalorganization.In:Mappingthemind(HirschfeldLA,GelmanSA,ed),pp85-116.Cambridge:CambridgeUniversityPress.DusenberyDB(1992)Sensoryecology.NewYork:W.H.Freeman.DyerFC(1998)Cognitiveecologyofnavigation.In:Cognitiveecology(DukasR,ed),pp201-260.Chicago:UniversityofChicagoPress.EmeryNJ,LorinczEN,PerrettDI,OramMW,BakerCI(1997)Gazefollowingandjointattentioninrhesusmonkeys(Macacamulatto).JournalofComparativePsychology,111:286-293.EtienneA,BerlieJ,GeorgakopoulosJ,MaurerR(1998)Roleofdeadreckoninginnavigation.In:Spatialrepresentationsinanimals(HealyS,ed),pp54-68.Oxford:OxfordUniversityPress.FodorJA(1983)Themodularityofmind.Cambridge,MA:MITPress.GallistelCR(1990)Theorganizationoflearning.Cambridge,MA:MITPress.GallistelCR(1995)Thereplacementofgeneral-purposetheorieswithadaptivespecializations.In:Thecognitiveneurosciences(GazzinigaM,ed),pp1255-1267.Cambridge,MA:MITPress.GallistelCR,CramerAE(1996)Computationsonmetricmapsinmammals:Gettingorientedandchoos•ingamulti-destinationroute.JournalofExperimentalBiology,199:211-217.GibbonJ,MalapaniC,DaleCL,GallistelCR(1997)Towardaneurobiologyoftemporalcognition:Advancesandchallenges.Currentopinioninneurobiology,7:170-184.GigerenzerG(1995)Thetamingofcontent:Somethoughtsaboutdomainsandmodules.ThinkingandReasoning,1:324-332.GigerenzerG(1997)Themodularityofsocialintelligence.In:MachiavellianintelligenceII:Extensionsandevaluation(WhitenA,ByrneRW,ed),pp264-288.Cambridge:CambridgeUniversityPress.HauserM,CareyS(1998)Buildingacognitivecreaturefromasetofprimitives:Evolutionaryanddevel•opmentalinsights.In:Theevolutionofmind(AllenC,CumminsD,ed),pp51-106.NewYork:OxfordUniversityPress.HermerL,SpelkeE(1996)Modularityanddevelopment:Thecaseofspatialreorientation.Cognition,61:195-232.HirschfeldLA,GelmanSA,ed(1994).Mappingthemind.Cambridge:CambridgeUniversityPress.HollandPC(1997)BrainmechanismsforchangesinprocessingofconditionedstimuliinPavloviancon•ditioning:Implicationsforbehaviortheory.AnimalLearningandBehavior,25:373-399.KrubitzerL(1995)Theorganizationofneocortexinmammals:Arespeciesdifferencesreallysodifferent?TrendsinNeurosciences,18:408-417.KummerH,DastonL,GigerenzerG,SilkJB(1997)Thesocialintelligencehypothesis.In:Humanbynature:Betweenbiologyandthesocialsciences(WeingartP,MitchellSD,RichersonPJ,MaasenS,ed),pp159-179.Mahwah,NJ:LawrenceErlbaumAssociates.LippHP,WolferDP(1995)Newpathstowardsolddreams:Microphrenology.In:Behavioralbrainresearchinnaturalisticandsemi-naturalisticsettings(AllevaE,FasoloA,LippHP,NadelL,RicceriL,ed),pp3-36.Dordrecht:KluwerAcademic.LythgoeJN(1979)Theecologyofvision.Oxford:ClarendonPress.MarrD(1982)Vision.NewYork:W.H.Freeman.McFarlandD(1991)Definingmotivationandcognitioninanimals.InternationalStudiesinthePhiloso•phyofScience,5:153-170.McFarlandD,BosserT(1993)Intelligentbehaviorinanimalsandrobots.Cambridge,MA:MITPress.McGrewWC(1992)Chimpanzeematerialculture:Implicationsforhumanevolution.Cambridge:Cam•bridgeUniversityPress.MithenS(1996)Theprehistoryofthemind.NewYork:ThamesandHudson. 60SaraShettleworthMooreBR(1996)Theevolutionofimitativelearning.In:Sociallearninginanimals:Therootsofculture(HeyesCM,GalefJr.BG,ed),pp245-265.SanDiego:AcademicPress.Moore-EdeMC,SulzmanFM,FullerCA(1982)Theclocksthattimeus.Cambridge,MA:HarvardUniversityPress.PearceJM(1994)Discriminationandcategorization.In:Animallearningandcognition(MackintoshNJ,ed),pp109-134.SanDiego:AcademicPress.PovinelliDJ,EddyTJ(1996)Whatyoungchimpanzeesknowaboutseeing.MonographsoftheSocietyforResearchinChildDevelopment,61:1-152.RedheadES,RobertsA,GoodM,PearceJM(1997)Interactionbetweenpilotingandbeaconhomingbyratsinaswimmingpool.JournalofExperimentalPsychology:AnimalBehaviorProcesses,23:340-350.RozinP(1976)Theevolutionofintelligenceandaccesstothecognitiveunconscious.ProgressinPsycho-biologyandPhysiologicalPsychology,6:245-280.ShepardRN(1994)Perceptual-cognitiveuniversalsasreflectionsoftheworld.PsychonomicBulletinandReview,1:2-28.SherryDF,JacobsLF,GaulinSJC(1992)Spatialmemoryandadaptivespecializationofthehippo•campus.TrendsinNeurosciences,15:298-303.SherryDF,SchacterDL(1987)Theevolutionofmultiplememorysystems.PsychologicalReview,94:439-454.ShettleworthSJ(1998)Cognition,evolution,andbehavior.NewYork:OxfordUniversityPress.ShettleworthSJ,HamptonRR(1998)Adaptivespecializationsofspatialcognitioninfood-storingbirds?In:Animalcognitioninnature(BaldaRP,PepperbergI,KamilAC,ed),pp65-98.SanDiego:AcademicPress.SperberD(1994)Themodularityofthoughtandtheepidemiologyofrepresentations.In:Mappingthemind(HirschfeldLA,GelmanSA,ed),pp39-67.Cambridge:CambridgeUniversityPress.StaddonJER(1983)AdaptiveBehaviorandLearning.Cambridge:CambridgeUniversityPress.TerraceHS(1984)Animalcognition.In:AnimalCognition(RoitblatHL,BeverTG,TerraceHS,ed),pp7-28.Hillsdale,NJ:Erlbaum.ThomasRK(1996)Investigatingcognitiveabilitiesinanimals:Unrealizedpotential.CognitiveBrainResearch,3:157-166.TinbergenN(1932/1972)OntheorientationofthediggerwaspPhilanthustriangulumFabr.I.In:Theanimalinitsworld(TinbergenN,ed),pp103-127.Cambridge,MA:HarvardUniversityPress.WilliamsDA,LoLordoVM(1995)TimecuesblocktheCS,buttheCSdoesnotblocktimecues.Quar•terlyJournalofExperimentalPsychology,48B:97-116.WilsonDS(1994)Adaptivegeneticvariationandhumanevolutionarypsychology.EthologyandSocio-biology,15:219-235. 4CognitiveEvolution:APsychologicalPerspectiveM.E.BittermanInthisinitialstakingoutofpositionswithrespecttotheevolutionofcognition(whichIinterpretas"knowing"intheclassicalsenseoftheterm,encompassingperception,learning,andunderstanding),mytaskistorepresentthepsychologicaltraditionrootedintheworkofThorndike(1911)andPavlov(1927)—theThorndi-kiantradition.TheviewsofThorndikeandPavlovdifferedconsiderably,ofcourse,andsoalsodidtheviewsoftheirvariousinfluentialsuccessors,suchasWatson,Lashley,Tolman,Guthrie,Skinner,Hull,Spence,andMowrer(Bitterman,1967).Skinner"swell-popularizedviewswereespeciallydeviant,althoughoutsidersoftenassumedthemtobetypical.Infact,itwouldbedifficulttofindsharpercriticismsofSkinnerthanthosemadebyinsiders.Iremember,forexample,H.L.Teuber"sdescriptionoftheSkinnerBoxasabloodlesstechniqueofdecortication,whichaffectsboththeexperimenterandtheanimal,andisirreversiblefortheexperimenter.AmongthecontemporarydescendantsofThorndikeandPavlov,thereisnolessdiversityofopinion,andyettherearesomeclearlydiscernablecommonassumptions.Forallthepublicizeddisagreementbetweenpsychologistsandethologists,thesameassumptionsaretobefoundinLorenz"sBehindtheMirror(1977).Thereisgeneralagreementthatinformationabouttheworldcomesfromsensorysystemswhoseoutputsarefilteredandorganizedincriticalwaysbygeneticallystructuredperceptualmechanisms;thatthemechanismsoflearning,whichpermitadaptationtoamuchwiderrangeofenvironmentsthancanpossiblybeprovidedforinthegenome,arethemselvesproductsofevolution;thatmuchoflearningisasso•ciativeincharacter;thatthemodificationofbehaviorbyrewardandpunishmentisofspecial—Lorenz(p.84)says"epoch-making"—importance;andthatunderstandingmustdependoninformationsuppliedbylearning.ThecentralconcernofpsychologistsworkingintheThorndikiantraditionhasbeenwithlearningandmemory.Theirearliestinterestwasinwhethersomethinglikehumanunderstandingcouldbefoundinanimals,buttheresultsofearlyexperimentswerelargelynegative.Onlyhighlyexperiencedsubjectssometimesseemedtosolveproblemsinsightfully,whichsuggestedthat,iftherewassomethinglikeunderstand•inginanimals,itwasgroundedinlearning,andthatwaswheretheinquiryshouldbegin.Questionsaboutlearningsoongaverisetoquestionsaboutperception(leadingtoagreatdealofinterestingworkongeneralization,discrimination,andattention),aswellastoquestionsaboutmotivationandabouttheinterplayoflearnedandunlearnedbehavior;foraproperappreciationofthescopeofthatwork,seethetextbookbyMackintosh(1974),alandmarkinitsfield.Interestinthediscoveryofconceptualabilitiesdidnot,however,entirelydisappearandinrecentyearshas 62M.E.Bittermanrevivedconsiderably,althoughstillwithoutclearoutcome;see,forexample,thecriticalanalysisbyHeyes(1998)ofevidencefor"theoryofmind"innonhumanprimates.AnoldcriticismofThorndikianresearchisthatitisdoneinunnaturalsituations,althoughlaboratoryproceduresforthestudyoflearninginanimalsweredesignedfromtheverybeginningwiththebehaviorofthesubjectsinmoreworldlysettingsinmind—Thorndike"spuzzleboxesforcats,Small"smazeforrats,andevenSkinner"skey-peckingapparatusforpigeons(nowcommonlyencounteredinecologicallabo•ratories).InBehindtheMirror,Lorenzdoesnothesitatetorelyonresultsofmazeexperimentswithratsandmice,and(inadiscussionofPavlov"sexperiments)recog•nizesthevalueofisolatingspecificresponses"artificially"(p.86)forpurposesofquantitativeanalysis.Intryingtounderstandananimal,wewanttoknowaboutitsbehaviorinaswidearangeofsituationsaspossible,andsettingsfarremovedfromthoseinwhichtheanimalnormallyisfoundmaybeespeciallyinstructive.H.F.HarlowonceremarkedthatfieldobservationsofrhesusmonkeysgavelittlehintoftheintellectualcapabilitiesdisplayedintheWisconsinGeneralTestApparatus.AnothercriticismofThorndikianresearchisthatithasbeenconcentratedonaverysmallnumberofspecies,andthewrongonesatthat.Particularlydisturbingforatimewastheamountofeffortdevotedtorats.Theanswertotheobjectionisthat,withlimitedresources,itismoreinstructivetostudyasmallnumberofspeciesintensivelythanalargenumberofspeciessuperficially;thelawsoflearninginanyspeciesarenotgivenimmediatelyinitsbehavior,butmustbeextractedpainstakinglyfromthedataofalongseriesofanalyticalexperiments.Inchoosingaspeciesforintensivestudy,wehavetoconsidersuchpracticalquestionsaswhetheritisreadilyavailable;whetheritliveswellinthelaboratory;andwhetheritssensory,motor,andmotivationalpropertiesaresuitableforthework.Havingdecidedsome40yearsagothatitwasimportanttohaveabodyofdetailedinformationaboutlearninginavertebrateverydifferentfromtheratinitsevolutionaryhistory,IwasledbyjustsuchpracticalconsiderationstochoosetheAfricanmouthbreedingfish,Tilapiamacro-cephala,whichL.R.AronsonbredingreatquantityatthenearbyAmericanMuseumofNaturalHistoryinNewYork,andwhichhadthespecialadvantageofalargeappetitefordryfood(Bittermanetal.,1958).Withthedevelopmentofautomaticdevicesforfeedingliquidfoodandsmallworms,thegoldfish(abundantalmosteverywhere)provedtobeevenmorepractical.Lateron,whenIdecidedthatitwasimportantalsotohaveabodyofdetailedinformationaboutlearninginaninverte•brate,Iwasledagainbypracticalconsiderationstothehoneybee.Andwhatoftheearlyfocusonlearninginrats?UsingdatafromSimpson(1945)andWalker(1964),R.B.Mastertononcesuggestedtomethatifwewerestarting CognitiveEvolution63overagainandhadtoselectasinglemammalianspeciesforintensivestudy,ratswouldbeprimecandidatesondemographicgrounds.Ofthelivinggeneraofmam•mals,morethanathirdarerodents—threetimesthenumberineitherofthenexttwolargestorders.Oftherodentgenera,almosttwo-thirdsareinoneortheotheroftwomainfamiliesgenerallyreferredtoas"rats"—theold-worldandnew-worldratsandmice.Rattus,oneoftheold-worldgenera,hasmanymorenamedformsthananyothermammaliangenus,andevenifonly50percentofthemweretruespecies,thenumberofRattusspecieswouldbelargerthanthenumberinanyothergenus.ItfollowsthatarandomsamplefromMammaliawouldmostprobablyyieldarodent,andasampleofRodentiawouldmostprobablyyieldarat—which,havingbeenselectedforconsideration,wouldpassthepracticaltests.Ifwewerelookingforanyvertebrate,theprobabilityissomewhatgreaterthatwewouldhitonabonyfish.General-ProcessTheoryOfcourse,therehasalsobeenconsiderableworkonlearninginmammalsotherthanrats,invertebratesotherthanmammals,andininvertebrates,andnothinghassoearnedtheskepticismofoutsidersasthehypothesis,conceivedearlyinthecourseofthework,thattherearesomequitegenerallawsoflearning.Thefirstformulationoftheso-calledgeneral-processviewweowetoThorndike(1911),whostudiedcats,dogs,chickens,monkeys,andevenaspeciesoffish(Fundulus),andwhofoundonlyquantitativedifferencesintheirlearning.Thevariousanimalslearneddifferentthings,andsomeseemedtolearnmorequicklythanothers,butallcouldbeunder•stoodassystemsofconnectionsgovernedbytheLawsofExerciseandEffect.Thorndike"sdistinctionbetweenthecontentandrateoflearningontheonehand,andthelawsoflearningontheother,isanimportantone.Hull(1945)putsthequestionaboutthegeneralityofthelawsoflearninginaveryclearway.Heaskswhethertheequationsthatdescribelearninginvariousspecies(whenwehavethem)willbeofthesameform,differingonlyinthevaluesoftheirconstants(as,inhisexample,thegravitationalconstantatHammerfestandMadras),orwhethertheequationsthemselveswillbedifferent.ConsidertheS-Scontiguityprinciple,accordingtowhichthepairingoftwostimuli—suchasthepairingofatone(theconditionedstimulus,orCS)andabitoffood(theunconditionedstimulus,orUS)inPavlovianconditioning—resultsintheformationofanassociationbetweenthem.BushandMosteller(1951)suggestthatthegrowthofassociativestrengthinthecourseoftrainingcanbedescribedbyasimplelinearequationcurrentlymorefamiliarinthenotationofRescorlaandWagner(1972): 64M.E.BittermanAV=a•Up(^-V)[1]withVrepresentingthestrengthofassociationatthebeginningofeachtrial;AV,thechangeinVproducedbythepairingoftheCSandtheUSonthattrial;a,thesalienceoftheCS;Up,thelearningrate;andX,themaximalstrengthofassociationthatcanbeachievedwithagivenUS—theasymptoteofthegrowthfunction.Onnonreinforcedtrials(thatis,ontrialswiththeCSalone),X=0,AVisnegative,andVdeclinesatarate(Dp)thatmaybedifferentfromUp.Theconstantsa,Up,Dp,andX,mayvarywidelyinvaluefromsituationtosituationandfromspeciestospe•cies,asmaytheconditionedresponse(CR)itself,andsoalsotherelationbetweenVandthemeasureofresponse(specifiedinasupplementaryperformanceequation);butthelearningequation(andtheconditioningprocessitdescribes)maybethesame.Thereisagooddealofcontemporaryinterestinthedevelopmentofquantitativetheoriesoflearningthatwillpermitexactratherthanmerelyordinalpredictionsofexperimentaloutcomes.Equation[1]hasprovedusefulinderivingtheresultsofexperimentsonsuchdiversephenomenaastransitiveinferenceinpigeons(CouvillonandBitterman,1992),avoidanceconditioningingoldfish(Zhuikovetal.,1994),and(byMartinShapiroindoctoralresearchattheUniversityofHawaii)risk-sensitiveforaginginhoneybees.Althoughwhatmanyhavecometothinkofastheethologicalviewisthattherearenogenerallawsoflearning,wefindgeneral-processthinkingbyLorenzhimselfinBehindtheMirror.Thereheconsidersatlengthsomephenomenaoflearninginani•malsasdiverseascuttlefish,flatworms,humaninfants,wasps,dogs,andbirdsofvariousspecies.Twoofthephenomena—facilitationbypracticeandsensitization—areinterpretedasfrequencyorpracticeeffectsandattributedtoaprocesslikethatof""runningin"anautomobile"(p.69).Thorndike"sLawofExercisecomestomindhere.Fourotherphenomena—habituation,habit,traumaticavoidance,andimprint•ing—weeexplainedintermsofassociation,whichisdennedas"theformingofalinkagebetweentwonervousprocesseshithertonotcausallyconnected"(p.81).Aseventhphenomenonisconditioningbyreinforcementor"learningthroughsuccessandfailure,"whichLorenzthinksofas"trueconditioning"ratherthan"mereasso•ciation."Itisbased,heassumes,onafeedbackprocessfoundinallanimalsexcept"unicellularandlowermulticellularcreatureswhichhavenocentralizednervoussystems"(pp.84-87).Here,ofcourse,Thorndike"sLawofEffectcomestomind.Insum,threegeneralprocessesarepostulated—afrequencyprocessandacontiguityprocessthatareperfectlygeneral,andareinforcementprocessthatoperatesinallbutthesimplestanimals.PsychologistsworkingintheThorndikiantraditionwouldquarrelwiththesubstanceofthetheory,whichissummarizedintable4.1,butwouldnotfinditsformorintentatallforeign. CognitiveEvolution65Table4.1General-processtheoryinBehindtheMirror(Lorenz,1977)ProcessLocusofoperationRepresentativephenomenaFrequencyAllanimalsFacilitationbypracticeSensitizationAssociationAllanimalsHabituationHabitTraumaticavoidanceImprintingReinforcementAllanimalswith"centralizedLearningbytrialandsuccess("trueconditioning")nervoussystems"General-processtheoryisoftenthoughttobecontradictedbyadaptivespecialization—theallegedevolutionarytailoringoflearningprocessestotheneedsofparticularspeciesinparticularsituations—althoughtheassumptionoftailoringimpliestheexistenceofgeneralprocessestobetailored,aswhenLorenzproposesthattheasso•ciationsinvolvedinimprintingandtraumaticavoidancemaybeespeciallypersistent.Onelineofevidenceforadaptivespecializationcomesfromexperimentsinwhichtheperformanceofthesamespeciesindifferentsituationsiscompared.Why,asksHuber(thisvolume),shoulditbeeasiertotrainratstoavoidshockbyrunningawayfromthedangersignalthanbymakingcopulatoryresponses?TheThorndikianansweristhataresponsemustoccurbeforeitcanberewarded,andthesignal,whichispairedwithshockwhenevertheanimalfailstoavoid,ismorelikelyinconsequenceofthepairingtoelicitrunningthancopulation.Theoncepopularconvictionthatfood-aversionexperimentspointtotheoperationofspecializedlearningprocessesisnowwidelyunderstoodtobegroundless(KlosterhalfenandKlosterhalfen,1985).Theprimordialfinding—thatratspoisonedaftereatinganovelfoodmaydevelopanaversiontothetasteofthefoodbutnottoitsappearance—followsdirectlyfromthecontiguityprinciple,giventhatthetracesofvisualstimulifaderapidlyintherela•tivelylongintervalbetweenpoisoningandillnesswhilefoodremainsinthegut(Bitterman,1975).Withoutcontrolsfortheeffectsofthemanyvariablesotherthanlearningthatinfluenceperformanceinsuchexperiments,conclusionsaboutlearningareunwarranted.Anotherlineofevidenceforadaptivespecializationcomesfromexperimentsinwhichtheperformanceofdifferentspeciesiscomparedinwhatispurportedtobethesamesituation,andheresimilardifficultiesareencountered.Betterperformanceoffood-storingbirdsascomparedwithnonstorersinspatiallearningtasksmightwellbedue,asShettleworth(1993)admits,toevolutionarytailoringoftheirsensoryormotivationalpropertiesratherthanoftheirlearning;spatialcues,shesuggests,maybemoresalientfortheblack-cappedchickadeethanforthedark-eyedjunco.Evenif 66M.E.Bittermanadifferenceinwhatshereferstoasthecapacityandpersistenceofspatialmemorycouldsomehowbedemonstrated,ofcourse,thefindingwouldnotgotothegeneral•ityofwhateverlearningprocessisinvolved—nottotheformofthelearningequa•tion,asHullputit,butonlytothevaluesoftheconstants.Itisoftenimpliedthatcertainseeminglyspecializedinstancesoflearning,suchasthemuch-advertisedsonglearningofbirds,involveprocessesthatareentirelyunique—thatthetailoringisqualitativeratherthanmerelyquantitative—butlittleinteresthasbeenevidencedinwhattheuniqueprocessesmightbe,orhowtheymightdifferfromthosedelineatedbyconventionallaboratoryexperiments.TheaccommodatingmodularviewexpoundedherebyShettleworthmaybeattractivebecauseitsuggeststhatweneednotworryverymuchaboutsuchmatters.TheLawofLeastEffortholdsformindaswellasformuscle.Scatteredallegationsofadaptivespecializationshouldnotbepermittedtoobscurethefactthatthereareagreatmanyphenomenaoflearningthattranscendparticularsensory,motor,andmotivationalcontexts,whicharefoundinmanydifferentspe•cies,andwhichpointtotheoperationofcommonprocesses—althoughwearestillnotclearastothenatureofsomeofthoseprocesses.ThestrategyofpsychologistsworkingintheThorndikiantraditionistocontinueingivenspeciestotrytodiscoverwhattheprocessesare,andincomparativeexperimentstoexaminetheirgeneralityoverarangeofwidelydivergentspecieschosentoprovidecluestotheirevolutionaryhistory.Becauseavarietyofgenerallearningprocessesseemtobeatworkinthespeciesthatwehavebeenstudyingintensively,andbecauseitisunlikelythatthewholesetofthemappearedtogether,orareinextricablylinked,weshouldnotbesurprisedtofindanimalsinwhichcertainofthemareabsent,orarepresentinadif•ferentform.Itispossible,however,asMacphail(1982)suggests,thattheremaybenodifferencesamongexistingvertebratesbecausethecriticaldevelopmentsoccurredatanearlystageofevolutioninanimalsnowextinct.LearninginVertebratesandHoneybeesThattherearemanyphenomenaoflearningcommontoourfavoritevertebratesubjects,asdiverseastheyare,isperhapsnotdifficulttounderstandontheassump•tionofcommonprocessesevolvedincommonancestors.Thatmanyofwhatappeartobethesamephenomenaarefoundalsoinhoneybeesisperhapsmoredifficulttounderstandinviewofthegreaterremotenessoftheevolutionaryrelationshipandthepresumedsimplicityofeventhemostadvancedcommonancestors.Because,asSimpson(1964)notes,convergencetothepointofidentityorevenofseriouslycon•fusingsimilarityisunlikelyinwhathereferstoaselaboratelypolygenicbehavioral CognitiveEvolution67Figure4.1Classicalconditioninginhoneybees(P,paired;U,unpaired;U-•P,unpairedtopaired.systems,itmightbeexpectedthatresemblancesinthelearningofhoneybeesandvertebrateswouldbenomorethansuperficial,butanexaminationofasmallsampleofthemwillshowthattheyarequitedetailed.Consider,forexample,theperformanceofagroupofhoneybeeforagers(GroupP)harnessedinsmalltubesandsubjectedtoaseriesofconditioningtrialswithodorastheCSandasmalldropofsucrosesolution,appliedtoantennaeandproboscis,astheUS.Asfigure4.1shows,theodorsooncomestoelicitextensionoftheproboscis(theCR),whichattheoutsetiselicitedonlybythesucrose(Bittermanetal.,1983).TheresultslookverymuchlikesalivaryconditioninginaPavloviandogstandingonlaboratorytableinPetrograd.Thestimuliandresponsesaredifferent,andsoalsoistherateofchangeinperformanceovertrials(whichisactuallymuchgreaterinthehoneybees),butthelearningprocessmaybethesame—theprocessdescribedbyequation[1].Thepossibilitythatthechangeinrespondingtotheodorisduetoexperiencewiththestimuliapartfromtheirpairing—say,tosensitizationoftheresponsebytheex•periencewithsucrose—isevaluatedwithacontrolprocedurecommoninworkwithvertebrates.Anothergroupofhoneybees(GroupU)isgiventhesameexperiencewiththeodorandsucrose,exceptthatthetwostimuliarepresentedseparatelyinrandomsequence(explicitlyunpairedtraining).Asfigure4.1shows,theseanimalsrespondverylittletotheodor,whichsuggeststhatthepairingofthestimulireallyisacriticalfactorintheperformanceofGroupP.Invertebrates,theexplicitlyunpairedprocedureisfoundnottobeassociativelyneutral,asevidencedbythefactthatastimulusexplicitlyunpairedwithaUSisslowtoconditionwhensubsequentlypairedwithit.Figure4.1showsthatthesameistrueforhoneybees;ascomparedwiththerateofconditioninginGroupP,therateofconditioninginGroupU—shown 68M.E.BittermaninthecurvelabeledU^P—isindeedslow.Astowhyexplicitlyunpairedtrainingshouldretardsubsequentconditioning,thereissomedisagreementinthevertebrateliterature.Nowthereisanotherquestionworthasking,whichhasbeenaskedalsoaboutappetitiveconditioninginvertebrates.IsthechangeinresponsetotheCSaproductofCS-UScontiguityperse,oroftheCR-UScontiguity(thecontiguityofresponseandsucrose)thatthepairingofthetwostimulioccasions?AlthoughthesucroseispresentedwhetherornottheCRoccurs,theCRisalwayscloselyfollowedbythesucrosewhenitdoesoccur,whichgivesusreasontosuspectthatwhatwemayreallyhavehereisinstrumentallearningindisguise(Lorenz"s"trueconditioning"ratherthan"mereassociation").Weanswerthequestionbytraininganewgroupofhoneybeeswithtwodifferentodors.Onsometrials,oneoftheodors(S+)ispairedwithsucrose,butonlywhentheCRfailstooccur(omissiontraining).Ontheremainingtrials,theotherodor(S-)ispresentedalone,servingtocontrolforthepossibilitythatanyrespondingtoS+isduesimplytosensitization.Theresultsarelikethoseforvertebrates.ThereislittleresponsetoS-,butagooddealofresponsetoS+,which(becausethereisnoCR-UScontiguity)mustbeattributedtothecon•tiguityofCSandUSonthefewtrialsonwhichtheCRdoesnotoccur.Itisinter•estingtonotethatPavlovfoundgoodsalivaryconditioningindogsevenwhentheCSwaspairedwiththeUSononlyasmallpercentageoftrials.AsPavlovalsofound,aCScanitselfactasaUSintheconditioningofanovelstimulus(second-orderconditioning),andthesameistrueofhoneybees.Afteroneodorhasbeenpairedwithsucroseinaseriesoftrials,asecondodorthatispairedwiththefirstodorsooncomestoelicittheCR.Thatdoesnothappenifthetwoodorsareexplicitlyunpairedinthesecondstageoftheexperiment.Itdoesnothappeneitherifthefirstodorhasbeenexplicitlyunpairedwithsucroseinthefirststageoftheexperiment—thepairedodorsmaybeassociatedinthesecondstage,butthesecondodorwillnotevoketheresponseifthefirstodordoesnot.Thevertebrateliteratureshowsthatcontiguousneutralstimuliare,infact,associated(thatclassicalcon•ditioningdoesnotrequireamotivationallysignificantUS),andthatthesameistrueofhoneybeeshasbeenshownbyexperimentsonwhatinthevertebrateliteratureiscalledwithin-compoundassociation(CouvillonandBitterman,1982).Theworkonwithin-compoundassociationinhoneybeeswasdonewithfree-flyingsubjectsthatwerepretrainedindividuallytoforageforsucrosesolutionatalabora•torywindow—feedingtorepletiononeachvisit,leavingforthehivetodepositthesucrose,andreturningoftheirownaccordafewminuteslatertocollectmore.(Thewindowusedinsuchexperimentsissosituatedastominimizefollowingbynest-mates,atechniquethatProfessorB.Hassensteintaughtmeinhislaboratoryat CognitiveEvolution69Figure4.2Within-compoundassociationinhoneybees.ResponsesinachoicetestwithAandBafterexperiencewithAXandBYfollowedbytrainingwithX+andY-.Freiburg,whereIfirstlearnedtoworkwithhoneybees.)Inthefirststageoftraining,thesubjectswereexposedonaseriesofvisitstotwograytargets,one(AX)labeledwithacolor(A)andanodor(X),andanother(BY)labeledwithadifferentcolor(B)andadifferentodor(Y).Inthesecondstageoftheexperiment,thesubjectsweretrainedtodiscriminatebetweentwograytargetslabeledwiththeodors,onlyoneofwhich(X)containedsucrosesolution.When,inthethirdstageoftheexperiment,thesubjectsweretestedwithtwograytargetslabeledonlywiththecolors,neithercon•tainingsucrosesolution,theyshowedastrongpreferenceforA(thecolorpairedinthefirststagewiththeodorreinforcedinthesecondstage).Infigure4.2,somesampleresultsareplottedintermsofthemeancumulativefrequencyofrespondingtoeachofthetargetsina10-minchoicetest.Itseemsreasonabletoconcludethatcolor-odorassociationsareformedinthefirststage,althoughnotthenevidentinbehavior—aninstanceofwhatiscalledlatentlearninginthevertebrateliterature.Experimentswithcompoundstimulishowavarietyofotherphenomenafirstdis•coveredinvertebrates.Oneofthemisblocking,whichhasbeenfoundbothinproboscis-extensionconditioning(SmithandCobey,1994)andintheperformanceoffree-flyingforagers(Couvillonetal.,1997).Afterreinforcedtrainingwithacom•poundoftwoodorsortwocolors(AB),thereislessresponsetoBaloneifAhaspreviouslybeenpairedwiththesamereinforcer;thatis,theconditioningofBisimpaired(blocked)bythepresenceofthepreviouslyconditionedA.Blockinginvertebratesiscommonlyexplainedontheassumptionthatthecomponentsofareinforcedcompoundcompetewitheachotherforassociativestrengthorforatten•tion,andthesameexplanationmaywellholdforhoneybees.Inanycase,itisclearfromthedatabothforvertebratesandforhoneybeesthatCS-UScontiguity,althoughnecessary,isnotsufficientforconditioning. 70M.E.BittermanFigure4.3Performanceofhoneybeesintwocolor-positiondiscriminationproblemswiththerelevantdimensionthesame(S)ordifferent(D)inthetwoproblems.Competitionforattentionhasbeendemonstratedinhoneybees,againinanexperimentpatternedafterworkwithvertebrates(Klosterhalfenetal.,1978).Free-flyingforagersweretrainedwithtargetsoftwodifferentcolorsintwodifferentpositions,somerewardedforchoosingoneofthecolorsindependentlyofposition(color-relevanttraining),andothersforchoosingoneofthepositionsindependentlyofcolor(position-relevanttraining).Inasecondproblem,withtargetsoftwonewcolorsintwonewpositions,learningwasmorerapidwhentherelevantdimensionwasthesameasinthefirstproblemthanwhenitwasdifferent.Infigure4.3,theresultsareplottedintermsofthemeannumberoferrorsmadeineachofthetwoproblems.Theresultssuggestthattheanimalslearnfirstofallinsuchproblemstosingleouttherelevantstimulusdimension(seealsoShapiroandBitterman,1998),andthatdimensionalselectionisinfluencedbypriortraining.Itisinterestingtonotethattheresultsoflikeexperimentsondiscriminativelearninginoctopuseshavebeeninterpretedinthesameway(SutherlandandMackintosh,1971).Anotherinterestingphenomenonofcompoundconditioningfoundinhoneybeesaswellasinvertebratesisconditionaldiscrimination(CouvillonandBitterman,1988).Honeybeescanlearntochoose,say,agreenratherthanabluetargetwhenbotharescentedwithgeraniol,butablueratherthanagreentargetwhenbotharescentedwithpeppermint.Becauseeachofthecomponentsisequallyoftenreinforcedandnonreinforced,differentialrespondingtothecompoundscannotbeunderstoodintermsofthesummedassociativestrengthsofthecomponents.Spontaneousdiscrimi•nationofcompoundsquacompoundsalsohasbeendemonstratedinhoneybees(CouvillonandBitterman,1982).Subjectsthathavefoundsucrosesolutiononeachoftwotargetslabeledwithdifferentcolor-odorcompounds(AXandBY)clearly CognitiveEvolution71preferthemtotwotargetslabeledwiththesamecolorsandodorspaireddifferently(AYandBX)inasubsequentchoicetest.Thesimplestinterpretationintheverte•brateliteratureisthatwearedealinginsuchexperimentswithperceptualratherthanconceptualeffects—thattheanimalslearnaboutconfiguralpropertiesinthesamewaytheylearnaboutcomponentproperties(whichalsomaysometimesbequitesubtle)—althoughthereissomedisagreementastohowtheperceptualeffectsarebestconceived.Thesedetailedsimilaritiesintheresultsforhoneybeesandvertebratesrepresentonlyaverysmallproportionofthosethathavebeenfoundinawiderangeofexperiments(Bitterman,1988,1996).Differencesalsohaveturnedupoccasionally.Theresultsofinitialblockingexperiments(Funayamaetal.,1995)werenegative,forexample,althoughforreasonsthatnowseemtobepurelyperceptual(Couvillonetal.,1997).Alongseriesofrecentexperimentsdesignedtolookforevidenceofinhibitoryconditioningproducedonlynegativeresults(Couvillonetal.,1999),buttheirmeaningisuncertainbecausetheevidenceofinhibitoryconditioninginverte•brates(commonlytakenforgrantedsincethetimeofPavlov)isunimpressive(PapiniandBitterman,1993).Noteworthy,too,isourfailureinotherrecentexperimentstofindevidencethatthecontrolofperformancebyshort-termmemorycanbemodifiedbylearning(Couvillonetal.,1998).Onthewhole,however,thedifferencesarefaroutweighedbythesimilarities.Whyshoulditbesodifficulttodiscoverdifferencesinthelearningofanimalswhosemostrecentcommonancestorlivedhalfabillionyearsagoandhadhardlyanybrainatall?Atleastpartoftheanswermaybethathoneybeescanbestudiedproperlyonlyinrelativelymassedtrialsandforrelativelybriefperiods,whichrulesoutasearchformanyinterestingandseeminglymorecomplexphenomenaofverte•bratelearning.Onthesegroundsalone,althoughforotherobviousreasonsaswell,itseemsnecessarytoextendtheworktootherinvertebrates,andthetrick,ofcourse,istofindasuitableone.Withahugeresearchbudget(becausetodotheworkwellwouldbeenormouslyexpensive),Imightbetemptedtoturnagaintooctopuses(Walkeretal.,1970).DivergenceinVertebrateLearningWhenweaskwhethertherearedifferencesinthelearningofhoneybeesandverte•brates,wearereferringtowhatweconceivetobegeneralphenomenaofvertebratelearning—phenomenathathavebeenfound,orthatweassumecanbefound,invertebratesofallclasses.Thequestiondoesnotimplythattherehasbeennodiver- 72M.E.Bittermangenceinvertebratelearning,althoughevidenceofdivergenceisscarceindeed,andthescarcityisnotduesimplytothefactthatpsychologistshavenotbotheredtolookforit.Macphail(1982)hasreviewedinconsiderabledetailtheresultsofalargenumberofcomparativeexperimentswithfishes,reptiles,birds,andmammals,includingprimates,whichinhisopinionprovidenocompellingevidenceofqualita•tivedifferencesinlearning.Iagreewithmanyofhisevaluations;thedifficultyinestablishingsuchdifferences,ofcourse,isthatperformanceinlearningsituationsisinfluencedbyvariablesotherthanlearning.Thereare,however,severalseriesofexperimentsthatIthinkdeservefurtherconsideration,amongthemexperimentsontheso-calledparadoxicalrewardeffectsdiscoveredinworkwithrats.Oneoftheeffectsisnegativeincentivecontrast,foundfirstinafamousexperimentbyElliott(1928).Twogroupsofratsweretrainedinamaze,onewithbranmashasreward,andthesecondwith(lessacceptable)sunflowerseeds.(Exceptwhereother•wisenoted,theintertrialintervalinalloftheinstrumentallearningexperimentscon•sideredherewas24hours—aprocedurethathastheimportantadvantage,amongothers,thatthesensoryantecedentsofresponseononetrialarenotcontaminatedbythesensoryconsequencesofresponseonprecedingtrials.)Theperformanceofthebranmashanimalswasmuchbetterthanthatofthesunflowerseedanimalsuntilthebranmashanimalswereshiftedtosunflowerseeds.Findingsunflowerseedsforthefirsttime,theyshowedagooddealofdisturbance,andtheirperformanceonsubsequentdaysfellprecipitouslybelowthatoftheanimalsrewardedwithsunflowerseedsfromtheoutset.Intheupperportionoffigure4.4,theperformanceofthetwogroupsisplottedintermsofmeanerrorspertrial.Theseresults,andthoseoflikeexperimentswithdifferencesonlyinquantityofreward,suggestthatinstrumentalbehaviorinratsismodulatedbyanticipationofitsrememberedconsequences-energizedbyrememberedrewardandimpairedbyrememberedfrustration(Amsel,1958)—allofwhichmayseemperfectlycommonplaceuntilitisappreciatedthatquitedifferentresultsareobtainedinanalogousexperimentswithotheranimals.LowesandBitterman(1967)trainedgoldfishtostrikeatargetloweredintothewateratthestartofeachtrial.Therewardwasaclusterof40(Tubifex)wormsforonegroup,whichthenwasshiftedtofourworms;asecondgroupwasrewardedthroughoutwithfourworms.Inthelowerportionoffigure4.4,theresultsareplottedintermsofthemeanloglatencyofresponse(inseconds).Performancewasbetterforthelargerrewardinthefirststage,butthegroupshiftedfromthelargertothesmallerrewardshowednodisturbancewhatsoever,continuingtorespondasrapidlyasbeforetheshift,andjustasrapidlyasathirdgroupthatcontinuedtofindthelargerreward.Theresultssuggestthatinstrumentallearningingoldfishisdifferentinsomeimpor•tantwayfrominstrumentallearninginrats. CognitiveEvolution73Figure4.4Performanceofratsandgoldfishshiftedfrompreferredtoless-preferredrewardscomparedwiththeper•formanceofunshiftedcontrolgroups.Itispossible,ofcourse,thattheexplanationlieselsewhere.Thetrainingsituationsaredifferent,theresponsemeasuresaredifferent,therewardsaredifferent,andsoalsomaybethedifferentialattractivenessoftherewards.Theonlywaytodealwithsuchpossibilitiesseemstobetodothesamekindofexperimentundersystematicallyvariedtrainingconditions(Bitterman,1975).Itmayneverbepossibletofindcon•ditionsofwhosefunctionalequivalenceforthetwospecieswecanbeconfident;buttotheextentthatthequalitativedifferenceintheirperformanceremainsthesamewithvariationinthetrainingconditions,wecangivelessweighttotheinterpretationintermsofcontextualvariablesalone.Consideranexperimentinwhichgoldfishweretrainedtoswiminarunwayratherthantostrikeatarget,witheither40wormsoronlyasinglewormasreward(Gonzalezetal.,1972).The40-wormgroupswammorerapidlythantheone-wormgroup,andwhenshiftedtothesmallerrewardcon•tinuedtoswimasrapidlyasbefore,althoughagroupshiftedfrom40wormstononesoonstoppedresponding.Withanewinstrumentalresponseandagreaterdiscrep•ancybetweenthetworewardmagnitudes,thecontrasteffectfailedagaintoappear. 74M.E.BittermanParallelresultsforthetwospeciesareobtainedinexperimentsinwhichcon-summatoryratherthaninstrumentalrespondingismeasured—ratslickingadrinkingtube,andgoldfishsuckingliquidfoodfromanipple.Ineachexperiment,therearetwogroupsofsubjects,onefedforabriefperiodeachdaywithapreferredfood,andthesecondwithalesspreferredfood.Boththeratsandthegoldfishtakemuchmoreofthepreferredfood.Thencomesadaywhentheanimalsthathavebeengiventhepreferredfoodfindthelesspreferredfoodinstead.Thefeedingbehavioroftheratsisdisrupted;theytakemuchlessofthelesspreferredfoodthandotheratsthathaveknownonlythatfood(e.g.,Flahertyetal.,1983)—thatis,theyshownegativecon•trast.Theperformanceoftheshiftedgoldfishisentirelyunaffected;theycontinuetotakeasmuchofthelesspreferredfoodastheyhavebeentakingofthepreferredfood(CouvillonandBitterman,1985).AvariationofElliott"sparadigmproducesanotherparadoxicaleffectinrats—aninverserelationbetweenamountofrewardandresistancetoextinction.Onegroupistrainedwithlargereward,asecondwithsmallreward,andthenbothgroupsareextinguished,whichistosaythattheyarenolongerrewardedatall.Theoutcomeofaconsiderablenumberofsuchexperimentswithrats(e.g.,GonzalezandBitterman,1969)isthatthelargerewardgroupperformsbetterintraining,butextinguishesmorerapidly;nonrewardseemstobemorefrustratingforratsthatanticipatealargerewardthanforratsthatanticipateonlyasmallreward.Inarunwayexperimentwiththreegroupsofgoldfish,onetrainedwith40worms,asecondwithfourworms,andathirdwithasingleworm,performancebothintrainingandinextinctionwasdirectlyrelatedtoamountofreward—thelargerthereward,themorevigoroustheperformanceinextinctionaswellasintraining(Gonzalezetal.,1972).AthirdparadoxicaleffectfoundrepeatedlyinratsisthepartialreinforcementeffectorPRE(e.g.,GonzalezandBitterman,1969).Theprocedureistotraintwogroupswithlargerewards(theeffectdoesnotoccurinwidelyspacedtrialswhentherewardissmall)—aConsistentgrouprewardedoneverytrial,andaPartialgrouponlyonhalfthetrials(theremainingtrialsunrewarded).Whentherewardforbothgroupsisthendiscontinued,thePartialgroupextinguisheslessrapidlythantheConsistentgroup;thatis,nonrewardinextinctionislessfrustratingforthePartialanimals,perhapsbecausetheyhavealreadyencountereditintraining,whereithasbeenfol•lowedeventuallybylargereward.Inanalogousexperiments,Africanmouthbreedersrewardedwithfoodpelletsforstrikingatarget(LongoandBitterman,1960)andgoldfishrewardedwithwormsinarunway(SchutzandBitterman,1969)havefailedtoshowthePRE.Thatwearenotdealingheremerelywithidiosyncraticpropertiesofthespeciesbeingcomparedissuggestedbytheresultsofanalogousexperimentswithavarietyof CognitiveEvolution75otheranimals.Negativeincentivecontrasthasbeenfoundinconsummatoryexperi•mentswithopossumsoftwospecies(Papinietal.,1988),andunmistakableevidenceofdisturbanceproducedbyunrealizedanticipationofapreferredrewardhasappearedalsoindelayedresponseexperimentswithrhesusmonkeys(Tinklepaugh,1928)andchimpanzees(CowlesandNissen,1937).ThePREhasbeenfoundinpigeons(Robertsetal.,1963),butneithernegativecontrastnortheinverserelation(Papini,1997).ThePREalsohasbeenfoundinpigeons,althoughnotingoldfish,underconditionsinwhichtheintertrialintervalswererelativelyshortbuttrialswithotherstimuliwereinterpolatedtocontrolforsensorycarryover(Bitterman,1994).Allthreeparadoxicaleffectshavefailedtoappearinexperimentswithanimalsoftwooldervertebratelines—turtlesoftwospecies(PapiniandIshida,1994;PertandBitterman,1970)andtoads(Schmajuketal.,1981;Muzioetal.,1992).Thepara•doxicaleffectsfailalsotoappearinveryyoungratstrained(necessarily)withrela•tivelyshortintertrialintervals,anditisinterestinginviewofthedissociationevidentinthepigeondatathatthePREistheearliestofthethreeeffectstoappearastheratsdevelop(Amsel,1992).Thisstrikingpatternofresults,whichisobscuredinMacphail"s(1982)treatment,suggeststhatinstrumentallearningmayhaveundergonesomeimportantchangesincommonreptilianancestorsofbirdsandmammals,anditisregrettablethatthereisstillsolittleevidenceastothenatureofthosechanges.FundingformyownworkontheproblemendedratherabruptlyseveraldecadesagowhenpeerreviewersofanewbreedmaligneditasaprimitiveAristotelianenterprise;comparisonsofratsandgoldfishcouldnotbeexpectedtotellusanythingaboutevolution,theyadvised,becauseratsarenotdescendedfromgoldfish.Inrecentyears,theintellectualclimatehasimprovedsomewhat—nobodyatleasthasyetfeltcompelledtowarntheagencysupportingmycurrentworkthatvertebratesarenotdescendedfromhoneybees—anditmaybethataproposalforcontinuationofthevertebrateexperimentswouldnowbemorefavorablyreceived,althoughmisunderstandinglingers.IambaffledbyShettleworth"sinsistencethattraditionalcomparativepsychologists,despitetheir"claim"tobeinterestedinthe"commonalityofcognitiveprocesses,"reallythinkofevolutionasa"ladderofimprovement"(1993,p.179)and"expectthatasinglehierarchicalorderingofmechanismswillbefoundwithcapacitiespos•sessedonlybyhumansatthe"mostcomplex"or"mostadvanced"end"(thisvolume,p.57).Whetherwearejudgedtobedisingenuousormerelyconfused,theimplicationisincorrectthattherecognitionbothofcommonalityanddivergenceisinherentlycontradictory.Astoimprovement,itwouldbeabsurdtodenyhistoricadvanceintheabilityofanimalstoknowtheworld,ortoquestionthecognitivepreeminenceofhumans.Onehasbuttocompare(atoneextreme)theanimalsoftheoldestlinesthat 76M.E.Bittermanhavebeenfoundinourexperimentstobeincapableevenofsimpleassociativelearning,and(attheother)ournearestprimaterelatives,howeverimpressivetheirperceptualandmnemoniccompetence,thatwouldbehardputtodesignsuchexperi•ments,ortoevaluatetheresults—whichisnot,ofcourse,tosaythatthemoleculargeometryofevolutionislinear.Rats,aswenowknow,arecertainlynotdescendedfromgoldfish,andearlyJapanesehorrorfilmshaveprepareduswellenoughforthepossibilitythatbeforethesunfadesortheearthisdestroyedbyanerrantcometsomesuperordinateintelligencewillappearinacrustacean.Isuspect,inanycase,thatfortheprosaicpresent,theirconfidenceundiminishedbyanairysermononmodularity,traditionalcomparativepsychologistswillcontinuetoworkinmuchthesamewayasbefore.SummaryComparativepsychologistshavebeeninterestedinperceptionandintheprecursorsofhumanunderstanding,butthemainfocusthusfaroftheirworkontheevolutionofcognitionhasbeenonlearningandmemory.Conditioningexperimentswithasmallnumberofwidelydivergentvertebratespecieshaveyieldedalengthylistofphenomenathatmayreasonablybeassumedonthebasisofthetaxonomicdiversityofthesubjectstobegeneralphenomenaofvertebratelearning;theresultsvaryquantitativelywithspeciesandwithtrainingtechniques,buttherearecommonqualitativepatternsthatareunderstandableintermsofcommonfunctionalprinci•plesandmaywellreflecttheoperationofhomologousmechanismsofinformationstorageandretrieval.Despitetheremotenessoftheevolutionaryrelationship,manyofthevertebrateprinciplesseemtoholdalsoforhoneybees,whoseperformanceinconditioningexperimentsshowsdetailedsimilaritiestothatofvertebrates,althoughherewemaysuspectthatthesimilaritiesare,atleastinlargemeasure,convergent.Inthelearningofvertebrates,thereisnotonlyextensivecommonality,butevidenceofbroadevolutionarydivergenceaswell;someofthatevidenceisprovidedbyasetofexperimentsonthecontrolofinstrumentalbehaviorbyitsrememberedandantici•patedconsequences,whoseresultsforbirdsandmammalsarequalitativelydifferentfromthoseforanimalsofoldervertebratelines.Familiarobjectionstothewayinwhichcomparativepsychologistshaveapproachedtheproblemofcognitiveevolu•tionandtheconclusionstowhichtheyhavebeenledarereviewedandevaluated.ReferencesAmselA(1958)Theroleoffrustrativenonrewardinnoncontinuousrewardsituations.PsychologicalBulletin,55:102-119. CognitiveEvolution77AmselA(1992)Frustrationtheory.Cambridge:CambridgeUniversityPress.BittermanME(1967)Learninginanimals.In:Contemporaryapproachestopsychology(HelsonH,BevanW,ed),pp140-179.NewYork:D.VanNostrand.BittermanME(1975)Thecomparativeanalysisoflearning.Arethelawsoflearningthesameinallanimals?Science,188:699-709.BittermanME(1988)Vertebrate-invertebratecomparisons.In:Intelligenceandevolutionarybiology(JerisonHJ,JerisonIL,ed),pp251-276.Berlin:SpringerVerlag.BittermanME(1994)Amsel"sanalysisofreward-scheduleeffects.PsychonomicBulletinandReview,1:297-302.BittermanME(1996)Comparativeanalysisoflearninginhoneybees.AnimalLearningandBehavior,24:123-141.BittermanME,MenzelR,FietzA,SchaferS(1983)Classicalconditioningofproboscis-extensioninhoneybees(Apismellifera).JournalofComparativePsychology,97:107-119.BittermanME,WodinskyJ,CandlandDK(1958)Somecomparativepsychology.AmericanJournalofPsychology,71:94-110.BushRR,MostellerF(1951)Amathematicalmodelforsimplelearning.PsychologicalReview,58:313-323.CouvillonPA,AblanCD,BittermanME(1999)Exploratorystudiesofinhibitoryconditioninginhoney•bees.JournalofExperimentalPsychology:AnimalBehaviorProcesses,25:103-112.CouvillonPA,ArakakiL,BittermanME(1997)Intramodalblockinginhoneybees.AnimalLearningandBehavior,25:277-282.CouvillonPA,ArincorayanNM,BittermanME(1998)Controlofperformancebyshort-termmemoryinhoneybees.AnimalLearningandBehavior,26:469-474.CouvillonPA,BittermanME(1982)Compoundconditioninginhoneybees.JournalofComparativeandPhysiologicalPsychology,96:192-199.CouvillonPA,BittermanME(1985)Effectofexperiencewithapreferredfoodonconsummatoryrespondingforalesspreferredfoodingoldfish.AnimalLearningandBehavior,13:433-438.CouvillonPA,BittermanME(1988)Compound-componentandconditionaldiscriminationofcolorsandodorsbyhoneybees:Furthertestsofacontinuitymodel.AnimalLearningandBehavior,16:67-74.CouvillonPA,BittermanME(1992)Aconventionalconditioninganalysisof"transitiveinference"inpigeons.JournalofExperimentalPsychology:AnimalBehaviorProcesses,18:308-310.CowlesJT,NissenHW(1937)Reward-expectanciesindelayed-responsesofchimpanzees.JournalofComparativePsychology,24:345-358.ElliottMH(1928)Theeffectofchangeofrewardonthemazeperformanceofrats.UniversityofCaliforniaPublicationsinPsychology,4:19-30.FlahertyCF,BeckerHC,CheckeS(1983)Repeatedsuccessivecontrastinconsummatorybehaviorwithrepeatedshiftsinsucroseconcentration.AnimalLearningandBehavior,11:407-414.FunayamaES,CouvillonPA,BittermanME(1995)Compoundconditioninginhoneybees:Blockingtestsoftheindependenceassumption.AnimalLearningandBehavior,23:429-437.GonzalezRC,BittermanME(1969)Spaced-trialspartialreinforcementeffectasafunctionofcontrast.JournalofComparativeandPhysiologicalPsychology,67:94-103.GonzalezRC,PottsAK,PitkoffK,BittermanME(1972)Runwayperformanceofgoldfishasafunctionofcompleteandincompletereductioninamountofreward.PsychonomicScience,27:305-307.HeyesCM(1998)Theoryofmindinnonhumanprimates.BehavioralandBrainSciences,21:101-134.HullCL(1945)Theplaceofinnateindividualandspeciesdifferencesinanatural-sciencetheoryofbehavior.PsychologicalReview,52:55-60. 78M.E.BittermanKlosterhalfenS,FischerW,BittermanME(1978)Modificationofattentioninhoneybees.Science,201:1241-1243.KlosterhalfenS,KlosterhalfenW(1985)Conditionedtasteaversionandtraditionallearning.PsychologicalResearch,47:71-94.LongoN,BittermanME(1960)Theeffectofpartialreinforcementwithspacedpracticeonresistancetoextinctioninthefish.JournalofComparativeandPhysiologicalPsychology,53:169-172.LorenzK(1977)Behindthemirror.NewYork:HarcourtBraceJovanovich.LowesG,BittermanME(1967)Rewardandlearninginthegoldfish.Science,157:455-457.MackintoshNJ(1974)Thepsychologyofanimallearning.London:AcademicPress.MacphailEM(1982)Brainandintelligenceinvertebrates.Oxford:ClarendonPress.MuzioRN,SeguraET,PapiniMR(1992)Effectofscheduleandmagnitudeofreinforcementoninstru•mentallearninginthetoad,Bufoarenarum.LearningandMotivation,23:406-429.PapiniMR(1997)Roleofreinforcementinspaced-trialoperantlearninginpigeons(Columbalivia).JournalofComparativePsychology,111:275-285.PapiniMR,BittermanME(1993)Thetwo-teststrategyinthestudyofinhibitoryconditioning.JournalofExperimentalPsychology:AnimalBehaviorProcesses,97:396-403.PapiniMR,IshidaM(1994)Roleofmagnitudeofreinforcementinspaced-trialinstrumentallearninginturtles(Geoclemysreevesii).QuarterlyJournalofExperimentalPsychology,47B:1-13.PapiniMR,MustacaAE,BittermanME(1988)SuccessivenegativecontrastintheconsummatoryrespondingofDidelphidmarsupials.AnimalLearningandBehavior,16:53-57.PavlovIP(1927)Conditionedreflexes.Oxford:OxfordUniversityPress.PertA,BittermanME(1970)Rewardandlearningintheturtle.LearningandMotivation,1:121-128.RescorlaRA,WagnerAR(1972)AtheoryofPavlovianconditioning:Variationsintheeffectivenessofreinforcementandnonreinforcement.In:ClassicalconditioningII:Currentresearchandtheory(BlackAH,ProkasyWF,ed),pp64-99.NewYork:Appleton-Century-Crofts.RobertsWA,BullockDH,BittermanME(1963)Resistancetoextinctioninthepigeonafterpartiallyreinforcedinstrumentaltrainingunderdiscrete-trialsconditions.AmericanJournalofPsychology,76:353-365.SchmajukN,SeguraE,RudiazA(1981)Rewarddownshiftinthetoad.BehavioralandNeuralBiology,28:392-397.SchutzSL,BittermanME(1969)Spaced-trialspartialreinforcementeffectandresistancetoextinctioninthegoldfish.JournalofComparativeandPhysiologicalPsychology,68:126-128.ShapiroMS,BittermanME(1998)Intramodalcompetitionforattentioninhoneybees.PsychonomicBulletinandReview,5:334-338.ShettleworthSJ(1993)Whereisthecomparisonincomparativecognition?Alternativeresearchprograms.PsychologicalScience,4:179-184.SimpsonGG(1945)Principlesofclassificationandaclassificationofmammals.BulletinoftheMuseumofNaturalHistory,85:1-350.SimpsonGG(1964)Organismsandmoleculesinevolution.Science,146:1535-1538.SmithBH,CobeyS(1994)TheolfactorymemoryofthehoneybeeApismellifera.JournalofExperimentalBiology,195:91-108.SutherlandNS,MackintoshNJ(1971)Mechanismsofanimaldiscriminationlearning.NewYork:Aca•demicPress.ThorndikeEL(1911)Animalintelligence:Experimentalstudies.NewYork:Macmillan. CognitiveEvolution79TinklepaughOL(1928)Anexperimentalstudyofrepresentativefactorsinmonkeys.JournalofCompar•ativePsychology,8:197-236.WalkerEP(1964)Mammalsoftheworld.Baltimore,MD:JohnsHopkinsUniversityPress.WalkerJJ,LongoN,BittermanME(1970)Theoctopusinthelaboratory:Handling,maintenance,train•ing.BehaviorResearchMethods&Instruments,2:15-18.ZhuikovAY,CouvillonPA,BittermanME(1994)Aquantitativetwo-processanalysisofavoidancecon•ditioningingoldfish.JournalofExperimentalPsychology:AnimalBehaviorProcesses,19:342-352. 5WhatMustBeKnowninOrdertoUnderstandImprinting?PatrickBatesonAstheresultofrelativelybriefexposuretoaparticulartypeofobjectearlyinlife,manybirdsandmammalswillformstrongandexclusiveattachmentstothatobject.Thisprocessisknownas"imprinting."Oneparticulartypeisknownas"filialimprinting,"becausetheobjecttowhichtheyounganimalbecomesattachedistreatedasaparent.Someofthecharacteristicsofimprintingareundoubtedlyduetothenaiveanimalsearchingforandrespondingselectivelytoparticularstimuli.Beforeimprintingtakesplace,theyoungbirdhasclearpreferencesforthetypeofstimulithatitwillsubsequentlylearnabout.Italsohasinplacearepertoireofmotoractivitiesthatfacilitatethelearningprocessandmaintainproximitytotheobjectofattachment.Learningtakesplaceatabiologicallyappropriatetimeinthelife-cycleandfewwoulddoubtthatthewholeprocesshasbeenadaptedduringevolutionforthekin-recognitionfunctionthatitservesundernaturalconditions.Theimageconjuredupbythetermimprintingisvividandsimple.Atacertainstage,thewaxoftheyounganimal"sbrainissoft;itreceivestheimprintofthefirstconspicuousthingtheanimalencounters.TheGermantermPrdgung(translatedas"imprinting")wasfirstusedbyHeinroth,(1911),althoughSpalding(1873)hadusedasimilarmetaphor,"stampingin."KonradLorenz(1935),whodidsomuchtomakethephenomenonfamous,likedtheimagebecauseitsuggestsaninstantaneous,irre•versibleprocess.Useofthetermalsoledtostrongclaimsthatimprintingisquitedifferentfromassociativelearning(Hess,1973).Asmoreevidencebecameavailable,theclaimsweredisputedandthetermwasheldtobemisleading(Bateson,1966;Sluckin,1972).Nevertheless,totheendofhislife,Lorenz(1981)continuedtotreattheprocessasspecial."Imprinting"hasbeenretainedintheliteraturebyadvocatesandcriticsalike.(Confusingly,"imprinting"hasalsobeenusedforaquitedifferentprocessoperatingatthegenomiclevel.Theinfluenceofonegeneonanothermaybedeterminednotbythedominanceofthegeneitselfbutbythesexoftheparentfromwhichitcomes.SeeConstanciaetal.,1998.)Leavingasidethematterofwhetherornottheterminologyisappropriate,whathappensasayounganimallearnsthecharacteristicsofitsparent?InthischapterIshallarguethat,tounderstandimprintingproperlyfromabehavioralstandpoint,itisnotnecessarytoknowhowgenesareswitchedonandofforanyoftheotherintricatemechanismsofcellularmachinery,interestingthoughsuchdetailsmightbe.Instead,whatisrequiredisagoodunderstandingofhowthevariousneuralsubprocessesinvolvedinlearningareactivatedindevelopmentandhowtheyfittogether. 86PatrickBatesonEarlyexperiencecanalsohavelong-lastingeffectsonsexualpreferences,buttheconditionsaredifferentfromthoseinwhichthefirstattachmentsareformed.Astonishingretentionofsexualpreferencesisfoundinthefaceofconsiderablesexualexperiencewithotherobjects(Immelmann,1972).However,thefinalhook-upbetweentherepresentationoftheimprintedobjectstoredinearlylifeandtheoutputsystemcontrollingsexualbehaviorprobablydoesnotoccuruntilmuchlaterthantheoriginalstorageoftherepresentation(HutchisonandBateson,1982;BischofandClayton,1991;KruijtandMeeuwissen,1991;Oettingetal.,1995).AsBischof(1997)argues,theparallelsbetween"sexualimprinting"andsonglearninginbirdsarestriking.Filialimprintingwithanovelandconspicuousobjectusuallyoccursmostreadilyataparticularstageofdevelopment(Bolhuis,1991)knownasthesensitiveperiod.Therangeofobjectsthatmotivateandelicitsocialbehaviorisrestrictedbytheani•mal"sexperience.Whentheyoungbirdbecomesfamiliarwithoneobject,thelikeli•hoodofitwithdrawingfromdissimilarconspicuousobjectsincreases.Thefirstpreferencesformedarelikelytobetheonesthatlast,withincertainconstraintssuchastheageoftheanimalatitsfirstexposureandthelengthofthatexposure(CherfasandScott,1981;ImmelmannandSuomi,1981;BolhuisandBateson,1990;Cook,1993).Whenabirdiswellimprinted,whenexposedtoanotherobject,itfirstwithdraws,showingeverysignofgreatalarm.Bydegreesthisalarmhabituates.Sometimesthebirdstartstodirectsocialbehaviortowardthenewobjectandmaybecomeattachedtoit.However,ifithasbeenwellimprintedwiththefirstobject,itdoesnotexpressanysocialbehaviortowardthesecondobject—itistamebutunattached.Thesensitiveperiodseemstobebroughttoanendbytheformationofasocialattachment(Bateson,1987).However,evendark-rearedchickseventuallyarelesseasytoimprintthantheywereshortlyafterhatching,whichmaysuggestthattheendingofthesensitiveperiodmaynotbeentirelyexperiencedependent(ParsonsandRogers,1997).FactorsInfluencingImprintingManyfactorshaverelativelyshort-termeffectsonresponsiveness.Forexample,PoltandHess(1966)foundthatdomesticchicksgiventwohoursofsocialexperiencewithsiblingsbeforehandfollowedamovingobjectmorestronglythandidisolatedbirds(LickliterandGottlieb,1985,1988).Stimulationinothermodalities,whenpresentedconcurrentlywithvisualstimuli,canhaveapowerfulmotivatingeffect.Gottlieb WhatMustBeKnowntoUnderstandImprinting?87(1971)foundthatindomesticchicksandmallardducklings,thesoundsmosteffectiveinelicitingpursuitofamovingvisualstimulusareconspecificmaternalcalls.Fur•thermore,youngbirdslearnthecharacteristicsofauditorystimuliplayedtothemshortlyafterhatching(Gottlieb,1988).Informingasocialattachmentundernaturalconditions,auditorysignalsareimportantinguidingtheprocess.TenCate(1989)alsofoundthatinJapanesequail,thepostureofaliveadultfemalehasapowerfulmotivatingeffectontheresponsetoherbythechicks.Atonetime,movementwasregardedasessentialin"releasing"thefollowingresponseofdomesticchicksandducklings,andsoininitiatingtheimprintingpro•cess.However,theeffectivenessofthemanyvisualstimuliusedintheimprintingsituationdependsonsuchpropertiesastheirsizeandshape,aswellasontheangletheysubtendandtheintensityandwavelengthoflighttheyreflect.Moreover,theratesatwhichthesevariableschangearealsoimportant—hencetheundoubtedeffectivenessofmovementandnicker.Thebirdclearlyrespondstoapatternofstimulation,andcharacterizationofthemosteffectivestimulusmustbecastintermsofcompounds.Gottliebandhiscol•leagues(JohnstonandGottlieb,1981;LickliterandGottlieb,1985)arguethattheconditionsunderwhichimprintingisstudiedinthelaboratoryaresoimpoverishedandartificialthattheresultscangiveaseriouslymisleadingviewofwhathappensinthewild.However,itdoesnotfollowthatexperimentalanalysisis,therefore,uselessorthatdifferentneuralsystemsarestudiedinlaboratoryandnaturalconditions.Acarthatisfilledwithlow-octanefuelandrunsbadlydoesnotbecomeanothercaronthataccount.Nevertheless,thewell-knownsensitiveperiodcurvesforchicksandducklings,withtheirpeakswithinthefirstdayafterhatching,areprobablymislead•ing.Mostprocessingandinformationstorageaboutthemotherprobablytakesplaceatleastadaylaterundernaturalconditions.Theworkonpredispositionshasincreasinglyfocusedonstimulusfeaturesfoundinthenaturalworld.Strongevidencesuggeststhatheadandneckfeaturesareparticu•larlyattractivetodomesticchicks(HornandMcCabe,1984;JohnsonandHorn,1988).Thediscoveryoftheheadandneckdetectorwasimportantbecauseitsuggestedadissociationoftheanalysissubsystemrequiredforimprintingfromtheoneinvolvedintherecognitionlearning.Underlaboratoryconditions,thenecessaryfeaturedetectorstakelongertodevelopthandotheonesdrivenbyflashinglightsandmovement(HornandMcCabe,1984;Bolhuisetal.,1985;Johnsonetal.,1985;Bolhuis,1989).Thedissociation,whichhadbeenanticipated(Bateson,1981),wasconfirmedbytheanalysisthatledtotheidentificationofaspecificregionofthebrainconcernedwithstoringarepresentationofimprintingobjects. 88PatrickBatesonIdentificationofaNeuralSiteforImprintingAnarrayofdifferentneurobiologicaltechniqueshaveimplicatedtheintermediateandmedialpartofthehyperstriatumventrale(IMHV)onbothsidesofthebrainasbeingsitesofaneuralrepresentationoftheimprintingobject(Horn,1985,1991,1998).Whenevidenceisopentoavarietyofinterpretations,greaterconfidenceinaparticularexplanationmaybeachievedbytacklingtheproblemfromanumberofdifferentangles.Eachpieceofevidenceobtainedbythedifferentapproachesmaybeambiguous,buttheambiguitiesaredifferentineachcase.Whenthewholebodyofevidenceisconsidered,therefore,muchgreaterconfidencemaybeplacedonapar•ticularmeaning.Ananalogyistryingtolocateonamapthepositionofavisiblemountaintop.Onecompassbearingisusuallynotenough.Twobearingsfromdif•ferentanglesprovideamuchbetterfix,andthreebearingsgivethemostreliablepositionforthetop.Animportantcomponentofthetriangulationprocedureistoexploittheasymp•toticcharacteroflearning:aphaseofrapidchangeisfollowedbyoneofmuchslowerchange.Therefore,animalsattherapidphasewillbelikelytoshowgreateractivityinbrainsitesspecificallyinvolvedinlearningthanthosethathavemovedontotheslowerphase,eventhoughmanyotheraspectsoftheanimals"experienceandactivitymatch.Animalsmaybepreparedinadvancebyunder-trainingthemorover-trainingthemonthetaskinquestion.Thistechniquewassuccessfullyexploitedwhenidenti•fyingtheroleofIMHVasasitefortheneuralrepresentationoftheimprintedobjectinimprinted(Batesonetal.,1973;Hornetal.,1979).ChicksthathavehadbothleftandrightIMHVremovedsurgicallyareunabletoimprint;ifbilaterallesionsareplacedimmediatelyafterimprinting,thebirdsshownorecognitionoftheimprintedobject(seeHorn,1985).Nevertheless,theselesionedchickswillshowapreferenceforastimulusthathasaheadandneckfeatureoveronethatdoesnot,therebydissociatingtheanalysiscomponentoftheimprintingprocessfromtherecognitioncomponent.Thelesioningexperimentsalsodissociatedrecog•nitionlearningfromlearninginvolvingexternalreward.ChickswilllearnavisualdiscriminationrewardedwithheatafterbilateralremovalofIMHV(Cipolla-Netetal.,1982;Honeyetal.,1995).Theywillalsolearntopressapedalwhenrewardedbytheviewofanimprintedstimulus,eventhoughtheydonotgoontolearnthechar•acteristicsofthatstimulus(JohnsonandHorn,1986).ManyofthedetailedcellularandmoleculareventsoccurringinIMHVarebegin•ningtobeworkedoutandtheconnectionsbetweenIMHVandotherstructureshavebeendescribed(Horn,1998).However,thelinksbetweenimprintingandotherlearningprocessesoccurringinparallelwithitarestillpoorlyunderstood. WhatMustBeKnowntoUnderstandImprinting?89DifferentLearningRulesTheuseoflesionssitedinIMHVisconsistentwiththeviewthatimprintingmightbeseparatedfromrewardedlearningonfunctionalgrounds.Manyofthetransactionsbetweenanimalsandtheirenvironmentsinvolveelementsofbothperceptuallearn•ing,occurringwithoutexternalreward,andofevent-relatinglearningthatdependsonexternalreward(orpunishment).Thesecomponentsofanoverallchangeinbehaviormaybeseenassubprocessesthatarenormallyusedinconjunction,butmaydependondifferentrules(Bateson,1990;Hollisetal.,1991).Whencharacterizingclassicalconditioning,Dickinson(1980)usedadefinitionthatrelatestotheutilityofthelearningprocess.Thelearningprocessservestouncoverthecausalstructureoftheenvironment.Thejobsoflearningtopredictandtocontroltheenvironmentarenotthesameasthatoflearningtocategorizeit.Atthephysio•logicallevel,similarifnotidenticalmechanismsmaybeusedtoachievethesediffer•entjobs.Atthebehaviorallevel,however,differentdesignruleswouldbeplausible.Detectingcausalstructuremayrequireclassification,butestablishingaclassificationdoesnotinvolveanassociationofcausewitheffect.Inuncoveringcausality,detectingorderisusuallycrucial.Ifthesupposedcausefollowsanevent,thenthenecessarycontingencyislikelytobemissing.Bycontrast,whenestablishingacategory,temporalcontiguitymaybeimportant,buttheorderinwhichthefeaturesoccurisnot.Undoubtedly,undersomeexperimentalarrange•ments,abackwardcontingencymaybeextractedinclassicalconditioning.Thisraisestwopossibilities:theregularityofanassociationmightallowthecomputationofacausallinkevenwhenthe"cause"appearstofollowthe"effect";alternatively,whenbackwardconditioningdoesnotoccur,theimpactofabiologicallysignificanteventmightdistractthesubject"sattentionfromeventsthatfollow.Thethrustofsometheoreticalapproacheshasbeentoexplainperceptuallearningandevent-relatinglearningprocessesinthesameterms(McLarenetal.,1989).Also,well-triedmethodologiesdevelopedfromthestudyofconditioninghavebeenappliedtoimprintingitself(AbercombieandJames,1961;Zolman,1982;Bolhuisetal.,1990;deVosandBolhuis,1990;vanKampenanddeVos,1995).Becauseneitherthetheorynortheexperimentalevidencehasdecisivelysuggestedaunitarymechanism,Ishallarguethecaseforlearningprocessesthataregovernedbydifferentrules.Intherealworld,acomplicatedobjectoftenpresentsasubstantiallydifferentsetoffeaturesfromoneviewthanitdoesfromanother.Inmanycircumstances,ananimalwouldbenefitfromtreatingthesedifferentsetsasthoughtheywereequivalent(Bateson,1973).Considertheproblemfacingabirdthathastogatherinformationaboutthefront,side,andbackviewsofitsmother.Alltheseviewsarephysically 90PatrickBatesondistinct,andtheymayalsotakeondifferentappearanceswhenviewedatdifferentdistances.Informationfromtwoseparatearraysoffeaturesmaybecombinedintoasinglerepresentationwhenthetwoarraysoccurinthesamecontextorwithinashorttimeofeachother(BatesonandChantrey,1972).Chantrey(1974)variedthetimebetweentheonsetofpresentationofoneimprintedobjectandtheonsetofpresentationofanother,andsubsequentlyrequireddomesticchickstodiscriminatebetweenthetwofamiliarobjectsinordertoreceiveafoodreward.Iftheobjectswerepresentedfiveormoreminutesapart,thebirdslearnedtodiscriminatebetweenthetwoobjectsmorequicklythanthoseinthecontrolgroupthathadnotbeenexposedtothesetwoobjects.However,whenthetwoobjectswerepresented30secondsorlessapart,theimprintedbirdstooklongerthanthecontrolgrouptolearnthediscrimination.Circumstancesarelikelytoarisewhenelementsofacompoundstimuluspresentedinrapidsuccessionareprocessedseparately.Stewartetal.(1977)wereonlyabletoobtainaclassification-togethereffectwhentheyreplicatedChantrey"sexperimentalconditionsexactly.Whentheyusedfeatureslesssalientthancolor,orpresentedthestimuliindifferentplaces,theydidnotgettheeffect—which,theyargued,wasfrag•ile.Nevertheless,thepointremainsthatwhentheelementsofacompoundaretreatedbyananimalaspartofawhole,theorderofpresentationdoesnotmatter.Ishallreturntothefragilityoftheeffectlater.Honeyetal.(1993),usingadifferenttechniquethanChantrey,double-imprintedchicksandthenrequiredthemtodiscriminatebetweenthetwoimprintedstimuli.Intheimprintingregime,thebirdswereeithergivenalternateexposureswithameaninterexposureintervalof14seconds—theMixedcondition—ortheywereexposedtoperiodicexposurestoonestimulusandthenafteragapoftwohourstoperiodicexposurestotheother—theSeparatecondition.Thepatternofimprintingwasother•wisethesameandthetotalexposuretothetwostimuliwasidenticalinthetwoconditions.ThebirdsimprintedundertheMixedconditiontooksignificantlylongertolearntheheat-rewardedvisualdiscriminationthanthechicksexposedundertheSeparatecondition.Theexplanationisthat,whenstimuliarepresentedinalternationclosetogetherintime,theyareclassifiedtogether;ifthebirdsaresubsequentlyrequiredtolearnthediscrimination,theyfirsthavetodisaggregatethetworepre•sentationsbeforetheyareabletomasterthetask.AModelofImprintingInordertounderstandmorefullytheclassification-togethereffect,itishelpfultohaveamodelofwhatmightbehappening.IshalloutlineonedevelopedbyGabriel WhatMustBeKnowntoUnderstandImprinting?91Hornandmyself(BatesonandHorn,1994).Thefirststepinthemodelsimulatesdetectionoffeaturesinastimuluspresentedtoayoungbird.Aspectsofthestimulusthatthebirdispredisposedtofindattractivearepickedoutatthisstage.Thesecondstepinvolvescomparisonbetweenwhathasalreadybeenexperiencedandthecurrentinput.Beforeimprintingtakesplace,nocomparisonisinvolved.Onceitoccurs,rec•ognitionofwhatisfamiliarandwhatisnoveliscrucial.Finally,thethirdstageinvolvescontrolofthevariousmotorpatternsinvolvedinexecutingfilialbehavior.ThebehavioralscaffoldingfortheimprintingprocessisprovidedbyadirectlinkbetweentheAnalysisandExecutivesystems.Figure5.1showsasimplifiedversionofthearchitectureoftheBatesonandHorn(1994)model.AllmodulesintheAnalysisSystemareinitiallylinkedtoallmodulesintheRecognitionSystemthat,inturn,arelinkedtoallmodulesintheExecutiveSystem,onlyoneofwhichisshownhere.Initialstrengthsoflinksareindicatedbythethicknessofthelines.AllmodulesintheAnalysisSystemarealsolinkedatmaxi•mumstrengthdirectlythroughaBy-PasstothemoduleintheExecutiveSystemthatcontrolsfilialbehavior(suchasapproachandfollowing).Thestartingconditionisshownfirst(figure5.1a).ThestrengthsoflinkagesbetweenmodulesafterthemodelhasbeenexposedtoastimulusthatactivatedAnalysismodule,Al,isshownnext(figure5.1b).ThespontaneousexcitabilityintheRecognitionmodule,R2,happenedtobehigherthanthatinRlatthetimetheinputfromAlarrivedandtheactivityinRlwasinhibited.Thestrengtheningruleisthatmodulesareconjointlyactive.Theweakeningruleisthattheupstreammoduleisinactivewhenthedownstreammoduleisactive.Thecompletedprocessisshowninfigure5.1c.Themodelcanreadilyperformaclassification-togetherprocessbyretainingtheexcitabilityoftheRecognitionmodulesforafiniteperiodaftertheyhadbeenacti•vated.Ifasecondsetoffeaturesarepresentedinalternationwiththefirst,thelevelofresidualexcitationintheRecognitionmodulesiscriticalindeterminingwhetherthetwostimuliaresubsequentlyrepresentedinthesamemodule.Thedegreeofoverlapinfeaturesbetweenthetwostimuliisalsocritical.Iftheoverlapishigh,theproba•bilityofthetwostimulisharingaRecognitionmoduleisalsohigh,evenwhenresid•ualexcitationfrompreviousstimulationiszero.Conversely,whentheoverlapoffeaturesislow,theprobabilityofsharingthesameRecognitionmoduleislow,evenwithmaximumlevelsofresidualexcitation.Themodelalsoprovidesareadyexplanationforsomenewempiricalevidence(BolhuisandHoney,1994,1998;HoneyandBolhuis,1997).Whenthematernalcallofthedomestichenaccompaniesthepresentationofavisualstimulus,thedomesticchickismoreresponsiveanddevelopsastrongerpreferenceforthevisualstimulus.However,iftheauditorystimulusisplayedintheabsenceofthevisualbeforethe 92PatrickBateson(a)AnalysisRecognitionExecutiveSystemSystemSystemFigure5.1SimplifiedarchitectureofBatesonandHorn"s(1994)modelforimprinting.AllmodulesintheAnalysisSystemareinitiallylinkedtoallmodulesintheRecognitionSystem,which,intheirturn,arelinkedtoallmodulesintheExecutiveSystem,onlyoneofwhichisshownhere.Initialstrengthsoflinksareindicatedbythethicknessofthelines.AllmodulesintheAnalysisSystemarealsolinkedatmaximumstrengthdirectlythroughaBy-PasstothemoduleintheExecutiveSystemthatcontrolsfilialbehaviour(suchasapproachandfollowing).Thestartingconditionisshowninpanel(a).Panel(b)showsthestrengthsoflinkagesbetweenmodulesafterthemodelhasbeenexposedtoastimulusthatactivatedAnalysismodule,Al.ThespontaneousexcitabilityintheRecognitionmodule,R2,happenedtobehigherthanthatinRlatthetimethattheinputfromAlarrived;activityinRlwasinhibited.Thestrengtheningruleisthatmodulesareconjointlyactive.Theweakeningruleisthattheupstreammoduleisinactivewhenthedownstreammoduleisactive.Thecompletedprocessisshowninpanel(c). WhatMustBeKnowntoUnderstandImprinting?93Figure5.2ResultsofanexperimentbyGriffiths(1998)inwhichdomesticchickswereimprintedwithamovingredtrianglefor120minutes.Onegroupwasgivennofurtherimprintingwhiletheotherwasimprintedwithamovingpurplecircleforafurther180min.Attheendofimprinting,chickswereeithergivenachoicebetweentheredtriangleandanovelbluecylinderorbetweentheredtriangleandthepurplecircle.Inthefirsttest,reductioninthepreferencefortheredtriangleastheresultofimprintingwiththepurplecircleisattributedtoaweakeningofthecontrolbytheredtriangle.Inthesecondtest,thereductionisattributedtoboththeweakeningofcontrolbytheredcylinderandthestrengtheningofthecontrolbythepurplecircle.compoundstimulus,thepreferenceforthevisualstimulusisweaker.Fromthestandpointofanimallearning,anevenmorestrikingresultisthatiftheauditorystimulusisplayedonitsownafterthecompound,thepreferenceforthevisualstim•ulusisalsoweakerthanwhenthepostcompoundexposurewasomitted.Somewhatsimilarresultshavebeenobtainedinothercontexts(e.g.,Dwyeretal.,1998)andhavebeenreferredtoas"retrospectiverevaluation."IntermsoftheBatesonandHornmodel,playingtheauditorystimulusonitsownweakensthelinkbetweentheAnalysismodulesprocessingthefeaturesofthevisualsystemandtheRecognitionsystem.Thisisbecausethedownstreammodulesareactivewhentheupstreammodulesareinactive.ThestrengtheningandweakeningaspectsofimprintinghavebeenexploredfurtherbyDanielGriffithsinhisPh.D.dissertation(Griffiths,1998).Chickswereexposedfor120minutestoamovingredtriangle.Halfofthemwerethenexposedforafur•ther180minutestoamovingpurplecircle,attheendofwhichtheirpreferencefortheredtrianglewascomparedwiththepurplecircleorwithanovelstimulus,amovingbluecylinder.Thesepreferenceswerecomparedwithchicksthatwerenotgivenasecondperiodofexposurewiththepurplecircle(figure5.2).Intermsofthe 94PatrickBatesonBatesonandHornmodel,thereductioninpreferencefortheredtriangleafterexpo•suretothepurplecircleisduetobothastrengtheningbetweentheAnalysismodulesprocessingthepurplecircleandtheRecognitionsystemandaweakeningbetweentheAnalysismodulesprocessingtheredtriangleandtheRecognitionsystem.Theextentofweakeningalonemaybeobtainedbycomparingthebirdsgivenachoicebetweentheredtriangleandthebluecylinderafternofurtherexposureorafterexposuretothepurplecirclefor180minutes.Nottoomuchshouldbemadeofthecalculationofthestrengtheningtoweakeningratiobecauseitisdifficulttoallowforgeneralizationandfortheinevitablenon-linearitiesintheunderlyingprocesses.However,throwingcautiontothewinds,aratioof4.3:1forstrengtheningtoweakeningisobtainedfromtheGriffithsdata,whichisclosetotheBatesonandHornguessof4:1.Itisnotformetopromoteotherneuralnetmodelsofperceptuallearningthatmightbeabletocopewellwiththesedata(e.g.,McLaren,etal.,1989;O"ReillyandJohnson,1994).Theirinterestliesinshowinghowamodelgeneratesanexperimentthatallowsestimatingparametervaluesinthemodel.OptimalTimeIntervalsIhavedwelledonsomeofthesuccessesoftheBatesonandHornmodel.Iwantnowtoconsideraninterestingfailure.IntheMixed/SeparatedesignthatwasusedtoreplicatetheChantreyresult,aMixedpresentationofapurplecircleandaredtri•angleduringimprintingledtosignificantlypoorerperformancethanaSeparatepre•sentationintheheat-rewardedvisualdiscriminationtestbetweenthepurplecircleandredtriangle(Honeyetal.,1993).However,theresultwasinvertedwhenthestrongpurplefeaturewassharedbythestimuliandthestimuliwereapurplecircleandapurpletriangle.NowtheMixedpresentationshowedasignificantlybetterperformancethantheSeparatepresentation(Honeyetal.,1994).Thecombinedresultsaresummarizedinfigure5.3.TheBatesonandHornmodelanticipatedthattwostimulisharingahighlyattractivefeaturewouldbemorelikelytoberepresentedinthesameRecognitionmodule,particularlyafterprolongedexpo•suretooneofthestimuli.ThisisbecauseastronglinkfromtheAnalysismodulerespondingtothehighstimulusvaluefeature,establishedduringexposuretothefirststimulus,increasesthelikelihoodthattheRecognitionmodulerespondingmoststronglytothefirststimuluswillalsorespondmoststronglytothesecondstimulus.However,theeffectoftheMixedconditionusingapurplecircleandapurpletrianglewasnotreadilyexplainedbytheBatesonandHornmodel. WhatMustBeKnowntoUnderstandImprinting?95Figure5.3Summaryoffourdoubleimprintingexperimentswithdomesticchicksinwhichexposuretothetwostimuliwaseither"mixed"or"separate"andthestimulieithersharedastronglyattractivefeatureincommon(color)ortheysharedrelativelyweakfeatures(suchasthepatternofmovement).Afterimprinting,thechickswererequiredtodiscriminatebetweenthestimuliwithwhichtheyhadbeenimprinted.Thechicksgiventheseparateimprintingconditioninwhichthesharedfeaturesofthestimuliwererelativelyweaklearnedthediscriminationmostquickly.Thechicksthathadbeengiventheseparateconditioninwhichthestimulisharedastrongfeaturelearnedthediscriminationmostslowly.TheHoneyetal.(1994)experimentwasrepeatedwithtwonaturalisticstimuli,asideviewandabackviewofajunglefowl.Thistimeacontrolwasincludedthathadnotbeenimprintedwiththetwostimuli.Onceagain,indiscriminationlearningtheMixedpresentationgaverisetosignificantlybetterperformancethantheSeparatepresentation,whichwasifanythingmarginallyworsethannothavinghadexperiencewitheitherstimuli(HoneyandBateson,1996).TheBatesonandHornmodelcanbemodifiedtocopewiththeseresultswhenanadditionalfeatureisadded.SupposingthatsomehabituationoccursintheAnalysismodules,thenthesharedfeaturesofthestimuliwillhabituatemorethanthenon•sharedfeatures.Asaresult,intheMixedcondition,thenonsharedfeatureswillstandoutmorerelativetoeachotherthanintheSeparatecondition.ThiseffecthastobesuperimposedonsomeresidualactivityintheRecognitionsystem.Ifsuchresidualactivitywassettodecaymorerapidlythanthedishabituationofanalysismodules,therevisedmodelsimulatedtheempiricaldata.Asthetimebetweenthepresentationsofthetwoimprintingstimuliwasincreased,thelikelihoodofthembeingclassifiedtogetherstartedhigh,thendeclined,andthenroseagain.Itisanempiricalmatterwhetherdecayoftheinferredresidualactivityandrecoveryofinferredhabituationhavedifferenttimecourses.Therefore,afurtherexperimentwascarriedout,varyingtheintervalbetweenpresentationtoameanof14secondsorameanof28seconds.Theresultsofthisexperimentshowedthatdoublingthetimeintervalledtoimprovedperformanceindiscriminationlearning(Honeyand 96PatrickBatesonBateson,1996).Thissupportedthenotionthatresidualactivitydeclinesmorerapidlythantheeffectsofhabituationdecays,leadingtoaloweredprobabilityofclassificationtogether.Theprobabilitythenclimbsagainasthepowerfulsharedfeaturedetectordishabitatuatesand,whenreactivated,pullstherepresentationsofthetwostimulitogether.Howgeneralaretheseresults?Iftheyrepresentauniversalfeatureofperceptuallearning,thepredictionisastrongone.Perceptuallearningwillimproveifthestimulithataretobediscriminatedbetweenarepresentedfairlyclosetogetherintime,butthegapbetweenthemmustnotbetooshort.Anoptimaltimeintervalbetweenpre•sentationsiscalledfor.LinksBetweenImprintingandRewardedLearningThepossibilityoftransferoftrainingafterimprintingmeansthattheneuralsystemunderlyingrecognitionlearningisconnectedwiththeoneunderlyingrewardedlearningintheintactanimal.Chicksshowastrongpreferencefortheimprintingobjectimmediatelyafterimprinting.Incontrast,whentransferoftraininginheat-rewardeddiscriminationlearningistestedimmediatelyafterimprinting,therateoflearningthediscriminationisunaffectedbyimprinting.However,ifdiscriminationlearningbetweenapurplecircleandaredtriangleisdelayedbysixhours,thenimprintingdoesaffecttherateoflearning,withthosegiventheMixedconditionlearningsignificantlymoreslowlythantheonesimprintedundertheSeparatecon•dition(Honeyetal.,1995).Althoughthememoryrequiredforrecognitionisformedquickly,thememorysustainingtransferoftrainingisnot.Lesionstudiessuggestthatanotherrepresentationoftheimprintingstimulus(knownasS")isconsolidatedinanotherregionofthebrainaboutsixhoursafterimprinting(McCabe,1991).TheformationortheuseofthisrepresentationcanbepreventedbyplacingalesionintherightIMHVsoonafterimprinting.Ifthelesionisdelayedformorethansixhours,theimprintedchicksretaintheirpreference.More•over,therepresentationmaybeusedintheheat-rewardeddiscriminationlearningtask(Honeyetal.,1995).Thelateralizationoftheprocessesinvolvedinformingthesecondstoreareofgreatinterestbecauseofthestrongevidenceaccumulatedovermanyyearsthatmanyprocessesinvolvedinthevisualcontrolofbehaviorarelater-alized(Andrew,1991;VallortigaraandAndrew,1991,1994).Thedynamicsofchangestakingplaceafterimprintinghavecertainsimilaritiestowhathappenswhenhumanslearnaboutfaces,becausetheleftprefrontalcortexisactivatedduringencodingnewmemoriesforfaces,whereastherightprefrontalcortexisactivatedduringlaterrecognitionofthosefaces(Haxbyetal.,1996).Inviewofthetransferof WhatMustBeKnowntoUnderstandImprinting?97trainingstudieswithIMHV-lesionedchicks(Honeyetal.,1995),thesecondstoreformedafterimprintingmightprovideapointofcontactbetweenimprintingandrewardedlearning.Whyarethetwoindependentmemorysystemsneeded?Possiblybecausetheasso•ciativeprocessdependentonexternalrewardrequiresdifferentrulesthanthesimplerecognitionprocessinvolvedinimprinting.Whyshouldthesystemsbelinked?Thebiologicaladvantagesofusingthesameinformationinavarietyofcontextscanbegreat.Inthecaseoftheyoungbird,themother"sactionsmayproveextremelyimportantinpredictingwhereandwhenitcanfindcrucialresourcesforitself.Whenthemothergivessignalssuchasthefoodcallthatthechicksrespondtowithoutlearning,allthechickshearthisandtheyareincompetitionwitheachother.Socapacityfortransferoftrainingislikelytopaytheindividualpossessingsuchamechanism.ConclusionImprintingisanexampleoftightlyconstrainedlearning.Paradoxically,itsgeneralinterestliesinitsparticularity.Thepredispositionstorespondtoparticularfeaturesandgiveparticularresponsestothestimulusarecentralinthecaseofimprinting.Mechanismsthatchangeasaresultofexperienceareobviouslydependentonmech•anismsthatdevelopedbeforeimprintinghastakenplace.Moreover,themechanismsthatexistedbeforeimprintingoccurredaresometimeschangedbytheexperienceandsometimesnot.Inotherexamplesoflearningwithdifferentfunctionsandinvolvedindifferentmotivationalsystems,theinterdependenceislessobvious,butpresent,nonetheless.Perhapsthemostimportantconclusionfromthebehavioralworkistheneedtothinkofagivenphenomenonintermsofaseriesofsubprocesses.BatesonandHorn(1994)referredtothesesubprocessesas"modules."Clearlysuchusagecancauseconfusionbecause"module"isawordthathascometohaveasmanymeaningsas"instinct"(seethechapterbyShettleworth).Aswithinstinct,beliefinthevalidityofonemeaningdoesnotimplybeliefinthevalidityofothermeanings.Forexample,manybelievethatthesubsystemsinvolvedinimprintinghaveevolvedastheresultofaDarwinianprocessofevolution(whichisonemeaningofmodule).Itdoesnotfollowthatthesubsystemsare"hardwired"anddonotchangeinthecourseofindi•vidualdevelopment(whichisanothermeaningofmodule).Nordoesitfollowthatthesubsystemsarededicatedtoonefunction(whichisyetanothermeaning).Despitetheambiguities,theconceptofamodularsubsystemgoessomewaytowardreconcilingthealternativeperspectivesrepresentedbyShettleworthinher 98PatrickBatesonchapterinthisbookandBittermaninhis.Bittermanclearlyrecognizestheneedtoexplaindifferencesinthewaysinwhichanimalslearnintermsofvariationintheperceptualandmotivationalmechanismsusedinthevariouscontextsinwhichlearningoccurs.Lookedatasawhole,thepropertiesoftheentiresystemaredif•ferent,allowingfortheevolutionofdifferencesinfunction.Shettleworth"spointisthatwholelearningsystemsmayhavedifferentuses.Nonetheless,thesubsystemsusedinonelearningcontextmayalsobeusedinanother.ToparaphraseBitterman"spointinmyownterms,thesubsystemsinvolvedinstorageofarepresentationoftheexternalworldmayoperateinthesameway.However,IdifferfromBittermaninsupposingthatrepresentationsofcausalityandrepresentationsofperceptualcatego•riesareachievedindifferentways.Theworkonimprintinghasfocusedontheanalysisofthefeaturesofthestimulithatstartofftheformationofthesocialattachment,theestablishmentofarepre•sentationofthatcombinationoffeatures,andthelinkingofsucharepresentationtothesystemcontrollingsocialbehavior.Thecommondenominatorwithagreatmanyotherlearningprocessesiscreatingarepresentationoftheobjecttowhichtheanimalhasbeenexposed.Representationsmustbeformedduringexploration,latentlearn•ing,and,indeed,virtuallyeverytransactionthatacomplicatedanimalhaswithitsenvironment.Theinferenceis,though,thatdifferentsubprocesseshavedifferentunderlyingrulesforplasticchange.Contiguityofthevariouselementsislikelytobeimportantinformingacategory,whereascontingencyiscrucialinlearningdepen•dentonexternalreward.Inferencesaboutthesubprocessesinvolvedinananimal"soveralltransactionwithitsenvironmentarecurrentlybeingexaminedattheneurallevel.Thebehavioraltheoriesundoubtedlymakeassumptionsaboutthenervoussystem,andtheseassump•tionsmayprovetobefalse.Astheneuralunderstandinggrows,theenquiryhastoreturntothebehaviorallevelsothatthepartsmaybereassembledand,ifneces•sary,newbehavioralexperimentsmaybedone.Thereturnflowofideasfromlowertohigherlevelsofanalysisseemsamuchmoreattractiveandplausiblepictureofcollaborationamongdisciplinesthanthatofrelentlessreductionisminwhichthebehavioralpeoplehandaproblemtotheneuralpeoplewho,havingdonetheirstuff,handitontothemolecularpeople.SummaryAlongdebatehasrevolvedaroundwhetherimprintingisspecial.Thetimingoftheprocess,thefeaturesthatmostreadilytriggerlearning,andthemotorsystemsthatarelinkedtorepresentationsstoredasaresultoflearningareallspecifictothe WhatMustBeKnowntoUnderstandImprinting?99functionalcontextofformingasocialattachmenttooneorbothparents.Theunderlyingneuralmechanismsmightbethesameasthoseinvolvedinotherlearningprocesses.Nevertheless,itisworthaskingwhethertherulesinvolvedinlearningaboutthecausalstructureoftheenvironmentaredifferentfromthoseusedinper•ceptuallearning(ofwhichimprintingisaspecialcase).Timeplaysadifferentroleinclassicalorinstrumentalconditioningthanitdoesinperceptuallearning.Theorderinwhichdifferenteventsareexperiencedmaymatteralotwhenoneeventcausestheother.However,theorderdoesnotmatteratallwhentheexperiencesaredifferentviewsofthesameobject.Somebehavioralandphysiologicalevidencefromstudiesofimprintinginchickssuggeststhatthesetwobroadfunctionsareservedbydifferentsubprocessesbutthatthesubprocessesare,nevertheless,intouchwitheachother.AcknowledgmentsIamgratefultoRobHoney,NickMackintosh,andtheeditorsfortheircommentsonearlierversionsofthemanuscript.IalsothankDanGriffithsforallowingmetorefertohisunpublisheddata.ReferencesAbercombieB,JamesH(1961)Thestabilityofthedomesticchick"sresponsetovisualflicker.AnimalBehavior,9:205-212.AndrewRJ,ed(1991).Neuralandbehaviouralplasticity.Oxford:OxfordUniversityPress.BatesonPPG(1966)Thecharacteristicsandcontextofimprinting.BiologicalReview,41:177-220.BatesonPPG(1973)Internalinfluencesonearlylearninginbirds.In:Constraintsonlearning:Limitationsandpredispositions(HindeRA,StevensonHindeJ,ed),pp101-116.London:AcademicPress.BatesonPPG(1981)Controlofsensitivitytotheenvironmentduringdevelopment.In:Behavioraldevel•opment(ImmelmannK,BarlowGW,PetrinovichL,MainM,ed),pp432-453.Cambridge:CambridgeUniversityPress.BatesonPPG(1987)Imprintingasaprocessofcompetitiveexclusion.In:Imprintingandcorticalplasticity(RauscheckerJP,MarlerP,ed),pp151-168.NewYork:Wiley.BatesonPPG(1990)Isimprintingsuchaspecialcase?PhilosphicalTransactionsoftheRoyalSocietyB,329:125-131.BatesonPPG,ChantreyDF(1972)Discriminationlearning:Retardationinmonkeysandchickspreviouslyexposedtobothstimuli.Nature,237:173-174.BatesonPPG,HornG(1994)Imprintingandrecognitionmemory—aneural-netmodel.AnimalBehav•iour,48:695-715.BatesonPPG,RoseSPR,HornG(1973)Imprinting:Lastingeffectsonuracilincorporationintochickbrain.Science,181:576-578.BischofHJ(1997)Songlearning,filialimprinting,andsexualimprinting:Threevariationsofacommontheme?BiomedicalResearchTokyo,18:133-146. 100PatrickBatesonBischofHJ,ClaytonN(1991)Stabilizationofsexualpreferencesbysexualexperience.Behavior,118:144-155.BolhuisJJ(1989).Thedevelopmentandstabilityoffilialpreferencesinthechick.Ph.D.dissertation,UniversityofGroningen.BolhuisJJ(1991)Mechanismsofavianimprinting:Areview.BiologicalReview,66:303-345.BolhuisJJ,BatesonP(1990)Theimportanceofbeingfirst:Aprimacyeffectinfilialimprinting.AnimalBehaviour,40:472-483.BolhuisJJ,deVosGJ,KruijtJP(1990)Filialimprintingandassociativelearning.QuarterlyJournalofExperimentalPsychologyBComparativePhysiologyandPsychology,42:313-329.BolhuisJJ,HoneyRC(1994)Within-eventlearningduringfilialimprinting.JournalofExperimentalPsy•chology:AnimalBehaviorProcesses,20:240-248.BolhuisJJ,HoneyRC(1998)Imprinting,learninganddevelopment:Frombehaviourtobrainandback.TrendsinNeurosciences,21:306-311.BolhuisJJ,JohnsonMH,HornG(1985)Effectsofearlyexperienceonthedevelopmentoffilialprefer•encesinthedomesticchick.DevelopmentalPsychobiology,18:299-308.ChantreyDF(1974)Stimuluspreexposureanddiscriminationlearningbydomesticchicks:Effectofvary•inginterstimulustime.JournalofComparativeandPhysiologicalPsychology,87:517-525.CherfasJJ,ScottAM(1981)Impermanentreversaloffilialimprinting.AnimalBehaviour,29:301.Cipolla-NetoJ,HornG,McCabeBJ(1982)Hemisphericasymmetryandimprinting:Theeffectofsequentiallesionstothehyperstriatumventrale.ExperimentalBrainResearch,48:22-27.ConstanciaM,PickardB,KelseyG,ReikW(1998)Imprintingmechanisms.GenomeResearch,8:881-900.CookSE(1993)Retentionofprimarypreferencesaftersecondaryfilialimprinting.AnimalBehaviour,46:405-407.deVosGJ,BolhuisJJ(1990)Aninvestigationintoblockingoffilialimprintinginthechickduringexpo•suretoacompoundstimulus.QuarterlyJournalofExperimentalPsychologyBComparativePhysiologyandPsychology,42:289-312.DickinsonA(1980)Contemporaryanimallearningtheory.Cambridge:CambridgeUniversityPress.DwyerDM,MackintoshNJ,BoakesRA(1998)Simultaneousactivationoftherepresentationsofabsentcuesresultsintheformationofanexcitatoryassociationbetweenthem.JournalofExperimentalPsychol•ogy:AnimalBehaviorProcesses,24:163-171.GottliebG(1971)Developmentofspeciesidentificationinbirds.Chicago:UniversityofChicagoPress.GottliebG(1988)Developmentofspeciesidentificationinducklings:XVIndividualauditoryrecognition.DevelopmentalPsychobiology,21:509-522.GriffithsDP(1998).Thedynamicsofstimulusrepresentationduringfilialimprinting:Behaviouralanalysisandmodelling.Ph.D.Dissertation,UniversityofCambridge.HaxbyJV,UngerleiderLG,HorwitzB,MaisogJM,RapoportSI,GradyCL(1996)Faceencodingandrecognitioninthehumanbrain.ProceedingsoftheNationalAcademyofScienceUSA,93:922-927.HeinrothO(1911)BeitragezurBiologie,namentlichEthologieundPsychologie,derAnatiden.In:Verh.5Int.Orn.Kongr.,pp589-702.HessEH(1973)Imprinting.NewYork:VanNostrandReinhold.HollisKL,tenCateC,BatesonP(1991)Stimulusrepresentation:Asubprocessofimprintingandcon•ditioning.JournalofComparativePsychology,105:307-317.HoneyRC,BatesonP(1996)Stimuluscomparisonandperceptual-learning—furtherevidenceandevalu•ationfromanimprintingprocedure.QuarterlyJournalofExperimentalPsychologyBComparativePhys•iologyandPsychology,49:259-269. WhatMustBeKnowntoUnderstandImprinting?101HoneyRC,BatesonP,HornG(1994)Theroleofstimuluscomparisoninperceptuallearning:Aninves•tigationwiththedomesticchick.QuarterlyJournalofExperimentalPsychologyBComparativePhysiol•ogyandPsychology,47:83-103.HoneyRC,BolhuisJJ(1997)Imprinting,conditioning,andwithin-eventlearning.QuarterlyJournalofExperimentalPsychologyBComparativePhysiologyandPsychology,50:97-110.HoneyRC,HornG,BatesonP(1993)Perceptual-learningduringfilialimprinting—evidencefromtransferoftrainingstudies.QuarterlyJournalofExperimentalPsychologyBComparativePhysiologyandPsy•chology,46:253-269.HoneyRC,HornG,BatesonP,WalpoleM(1995)Functionallydistinctmemoriesforimprintingstimuli:Behavioralandneuraldissociations.BehavioralNeurosciences,109:689-698.HornG(1985)Memory,imprinting,andthebrain.Oxford:ClarendonPress.HornG(1991)Cerebralfunctionandbehaviourinvestigatedthroughastudyoffilialimprinting.In:Thedevelopmentandintegrationofbehaviour(BatesonP,ed),pp121-148.Cambridge:CambridgeUniversityPress.HornG(1998)Visualimprintingandtheneuralmechanismsofrecognitionmemory.TrendsinNeuro•sciences,21:300-305.HornG,McCabeBJ(1984)Predispositionsandpreferences.Effectsonimprintingoflesionstothechickbrain.AnimalBehaviour,32:288-292.HornG,McCabeBJ,BatesonPPG(1979)Anautoradiographicstudyofthechickbrainafterimprinting.BrainResearch,168:361-373.HutchisonRE,BatesonP(1982)SexualimprintinginmaleJapanesequail:Theeffectsofcastrationathatching.DevelopmentalPsychobiology,15:471-477.ImmelmannK(1972)Sexualandotherlong-termaspectsofimprintinginbirdsandotherspecies.AdvancesintheStudyofBehavior,4:147-174.ImmelmannK,SuomiSJ(1981)Sensitivephasesindevelopment.In:Behavioraldevelopment(Immel•mannK,BarlowGW,PetrinovichL,MainM,ed),pp395-431.Cambridge:CambridgeUniversityPress.JohnsonMH,BolhuisJ,HornG(1985)Interactionbetweenacquiredpreferencesanddevelopingpredis•positionsinanimprintingsituation.AnimalBehaviour,33:1000-1006.JohnsonMH,HornG(1986)Dissociationbetweenrecognitionmemoryandassociativelearningbyarestrictedlesiontothechickforebrain.Neuropsychologia,24:329-340.JohnsonMH,HornG(1988)Developmentoffilialpreferencesindark-rearedchicks.AnimalBehaviour,36:675-683.JohnstonTD,GottliebG(1981)Developmentofvisualspeciesidentificationinducklings:Whatistheroleofimprinting?AnimalBehaviour,29:1082-1099.KruijtJP,MeeuwissenGB(1991)Sexualpreferencesofmalezebrafinches:Effectsofearlyandadultexperince.AnimalBehaviour,42:91-102.LickliterR,GottliebG(1985)Socialinteractionwithsiblingsisnecessaryforthevisualimprintingofspecies-specificmaternalpreferenceinducklings.JournalofComparativePsychology,99:371-379.LickliterR,GottliebG(1988)Socialspecificity:Interactionwithownspeciesisnecessarytofosterspecies-specificmaternalpreferenceinducklings.DevelopmentalPsychobiology,21:311-321.LorenzK(1935)DerKumpaninderUmweltdesVogels.JournalofOrnithology,83:137-213,289-413.LorenzKZ(1981)Thefoundationsofethology.NewYork:Springer-Verlag.McCabeBJ(1991)Hemisphericasymmetryoflearning-inducedchanges.In:Neuralandbehaviouralplasticity(AndrewRJ,ed),pp262-276.Oxford:OxfordUniversityPress.McLarenIPL,KayeH,MackintoshNJ(1989)Anassociativetheoryoftherepresentationofstimuli:Applicationtoperceptuallearningandlatentinhibition.In:Paralleldistributedprocessing—implicationsforpsychologyandneurobiology(MorrisRGM,ed),pp102-130.Oxford:OxfordUniversityPress. 102PatrickBatesonO"ReillyRC,JohnsonMH(1994)Objectrecognitionandsensitiveperiods—acomputationalanalysisofvisualimprinting.NeuralComputation,6:357-389.OettingS,ProveE,BischofHJ(1995)Sexualimprintingasa2-stageprocess—mechanismsofinformation-storageandstabilization.AnimalBehaviour,50:393-403.ParsonsCH,RogersLJ(1997)PharmacologicalextensionofthesensitiveperiodforimprintinginGallusdomesticus.PhysiologyandBehavior,62:1303-1310.PoltJM,HessEH(1966)Effectsofsocialexperienceonthefollowingresponseinchicks.JournalofComparativeandPhysiologicalPsychology,61:268-270.SluckinW(1972)Imprintingandearlylearning,2ndedition.London:Methuen.SpaldingDA(1873)Instinctwithoriginalobservationsonyounganimals.Macmillan"sMagazine,27:282-293.StewartDJ,CaprettaPJ,CooperAJ,LittlefieldVM(1977)Learningindomesticchicksafterexposuretobothdiscriminanda.JournalofComparativeandPhysiologicalPsychology,91:1095-1109.tenCateC(1989)Stimulusmovement,henbehaviorandfilialimprintinginJapanesequail(Coturnixcoturnixjaponica).Ethology,82:287-306.VallortigaraG,AndrewRJ(1991)Lateralizationofresponsebychickstochangeinamodelpartner.AnimalBehaviour,41:187-194.VallortigaraG,AndrewRJ(1994)Differentialinvolvementofrightandlefthemisphereinindividualrec•ognitioninthedomesticchick.BehaviouralProcesses,33:41-57.vanKampenHS,deVosGJ(1995)Astudyofblockingandovershadowinginfilialimprinting.QuarterlyJournalofExperimentalPsychologyBComparativePhysiologyandPsychology,48:346-356.ZolmanJF(1982)Ontogenyoflearning.In:Perspectivesinethology,vol.5:Ontogeny(BatesonPPG,KlopferPH,ed),pp275-323.NewYork:PlenumPress. 6StimulusEquivalenciesThroughDiscriminationReversalsJuanD.Delius,MasakoJitsumori,andMartinaSiemannThesensorysystemsofadvancedanimalsfrequentlyinputmorestimulusinforma•tionintothenervoussystemthanthemotorsystemcanpossiblyoutputasbehaviorpatterns.Thisbottleneckdemandsadrasticinformationreduction.Twotypesofreductioncanbedistinguished:selectiveattentioninvolvingbehavioralcontext-dependentinformationcensoring(aswhenduringsexualbehaviorfoodstimuliareusuallynotreactedto);andcategorization,apoolingintofewerresponseoutputs(aswhendifferentfoodsallelicitthesameingestiveresponse).Hereweareconcernedwithprocessesunderlyingthislatterkindofinformationreduction,thatis,howthenervoussystemmanagestoclassifystimulisothattheyresultinarestrictednumberofbehaviors.Inhumanpsychology,suchinformationpoolinghasbeenmuchstudiedundertheheadingofconceptformation,wheretheresponsesofinteresthavebeenwords,whichthroughsuitableexperience,cometocorrespondsemanticallytocollectionsofstimuli(SlomanandRips,1998).Thereareseveraltheoriesabouttheformationofconceptsbutnosingleonehasemergedasbeinguniquelycorrect(Fodor,1998).Thedeviousnessofthehumanmindatconceivingnamedcategoriesexceedsanysuccincthypothesisthatcanbeputtopaper.Considerthetermsmaleandfemaleasappliedtoorganismsfromalgaetohumansandtoobjectssuchasthesunandthemoon."Lesoleil"and"lalune,""dieSonne"and"derMond"areconverselymaleandfemaletofranco-andgermanophonepeople.Mosttheoriesassumethatthestimulithatcometobecategorizedtogetherdosoonthebasisofperceptualsimilarities,eitherbysimplybeingneighborsalongaphysicaldimensionsuchassizeorwavelength,orbysharingsomephysicalfeaturessuchasfeathersorlegs.Thecontributionoflessim•mediatesimilaritiesofstimulioritems,suchastheabilitytoflyortokill,orwhateverconfersgenderstothesunandthemoon,havealsobeenconsideredtheoreticallybuthavehardlybeenexaminedempirically.Astimuluscategorizationbythesekindoffunctionalattributesinterestsushere,thoughnotsomuchinhumansasinpigeons.Theabilityofpigeonstolearntodistinguishsetsofstimulithatarephysicallysimilarwithinsetsandphysicallydifferentbetweensets(forexample,slidescon•taininghumansornotcontaininghumans)inadirectlyperceivablewayasbelongingtodifferentcategories(asbeingworthornotworthpeckingforfoodreturns,forexample),andthecompetencetothenspontaneouslygeneralizethisdiscriminationtonovelbutstillcorrespondinglysimilaranddifferentstimuliwithoutadditionaltrain•ing,wasfirstestablishedbyHerrnsteinandLovelandin1964.Analogousresultshavesincebeenreplicatedwithawidevarietyofstimuli(e.g.,Lubow,1974;Delius,1992).Pigeonshaveprovennotonlycapableofcategorizingpicturesofnaturalobjectsbut 104JuanD.Deliusetal.alsoabletocategorizeartificialscenessuchasimpressionistandcubistpaintings(Watanabeetal.,1995).Moreover,Bhattetal.(1988)showedthattheycouldnotonlyclassifypicturesdichotomouslybutthattheyalsocouldconcurrentlycategorizepicturesoffourclassesofobjects.Kirpatrick-StegerandWasserman(1996)hadnoparticulardifficultyinhavingpigeonscategorizestimuliaccordingtowhetheroneoftwoshapeswasabove,below,totheleft,ortotherightoftheother.Thelimitsmightlieintheabstractnessofthepropertiesthatmakethestimuliinanintendedcategorysimilar.Herrnsteinetal.(1989)founditdifficulttotrainpigeonstocategorizedrawingsofdotswithinclosedcurvedlinesfromdrawingsofdotsoutsidesuchlines.Theselimitsaside,foralongtimetheremarkablecategorizationperformancesofpigeonswereequatedwithanabilitytoconceptualize,followingadefinitionbyKellerandSchoenfeld(1950)affirmingthatconceptsinvolvedageneralizationwithinclassesandadiscriminationbetweenclassesofstimuli.ConceptsThroughReversalsLea(1984)arguesthatthetermconceptshouldbeusedonlyifthestimulipertain•ingtoacategoryareshowntobeassociatedwitheachotherandnotonlywitharesponseorareinforcement.Heproposesthatsuchinter-stimulusassociationcanbedemonstratedbyretrainingtheanimalswithasubsetofthecategorystimuli(orthefeaturescomposingthem)withreversedreinforcementallocationsandthentestingforwhetherthereversedresponsewouldspontaneouslytransfertotheremainingstimuli(features).Ifallstimuli(features)belongingtoacategorywerebondedbyassociations,theresponseswitchshouldtransferfromtheleadingsubsettothetrail•ingsubset.Theprocedureyieldedmixedresults.Leaetal.(1990)hadpigeonsdiscriminatesetsofletters,anduponasinglereversaltheyfoundsomeevidenceofreversaltransfer.FersenandLea(1990)trainedpigeonstodiscriminatetwosetsoftownscapesinvolvingseveralfeatures.Whenreversedwithrespecttoonepairoffeaturestheyshowednoevidenceoftransfertotheotherfeatures.BhattandWasserman(1989)hadpigeonscategorizepicturesoffourdifferenttypesofobjects,butfoundthatreversaltrainingdidnottransfer.Jitsumori(1993)gotpigeonstosuccessfullycate•gorizeartificialmultifeaturedstimuliaccordingtofeatureadditionprinciplesbutfoundnoreversaltransferacrossfeatures.AstleyandWasserman(1998)hadpigeonslearntoco-categorizepicturesofpeople,flowers,cars,andchairsintotwoclassesandsimilarlyfoundonlyratherweakevidenceofinter-stimulusassociationsusingaresponsereassignmentprocedure.Allthesestudiesexposedtheanimalstoonlya StimulusEquivalenciesThroughDiscriminationReversals105singlereversalorreassignmentbeforetestingforchoicetransfer.Theywerethusexpectingthatinter-stimulusassociationswouldhavearisenthroughtheearliercat•egorizationtrainingratherthanbythereassignmentorreversalprocedureitself.Vaughan(1988)usedthereversalmethodforbothstrengtheningtheinter-stimulusassociationsandtestingthem.Pigeonslearnedtodiscriminateslidesalldepictingtreesbutwhichwerearbitrarilydividedintoasetof20positiveandasetof20neg•ativeslides.Theslideswereprojectedinarandomorderontoapeckingkey.Onlypecksonthepositivepicturesyieldedfoodrewards.Whenthepigeonshadlearnedtodiscriminate,theallocationofrewardandnorewardwasexchangedbetweenthesets.Suchreinforcementreversalswererepeateduntilthebirdsbecameproficientatswitchingtheirchoicesaccordingtothereigningallocations.Vaughanthenshowedthatuponareversal,experiencewithafewinitialslideswassufficienttocausethepigeonstorespondcorrectlytoalltheremainingones.Thatis,whenthepigeonsdetectedthatsomeoftheslidesofthesetshadexchangedtheirfunctionalsignifi•cance,theyspontaneouslytransferredtheadequateresponsetotheremainder.ThisledHerrnstein(1990)toexpandhisviewsaboutconceptualizationinanimals.Tothepreviouslyacceptedlevelsof(a)stimuluscategorizationbasedonstraight•forwarddiscrimination,(b)onbrute,multipleby-rotelearning,(c)onopen-endedextensionthroughgeneralizationgradients,and(d)oncommonabstractrelationsbetweenstimuluscomponents,headdedanewlevel(e)wherethecategorizationofstimuliwasnotbasedonperceptualsimilaritiesbetweenthembutbasedonreinforce•mentcontingenciesthatdifferentstimulicouldshare.ThislatterlevelcorrespondedtothestimulusclassificationbymultiplereinforcementreversalsdemonstratedbyVaughan(1988).Note,however,thatthereisnothingtopreventtheprocessesre•sponsibleforthedifferentlevelsofcategorizationfromoperatingconjointly.EquivalenciesbyMatchingVaughan(1988)maintainedthathispigeonshadformedtwoequivalencesetsinthesensethatexperiencewithanyexemplarofeachclasswasequallycapableofelicitingachoiceswitch.ThisviewconflictedwithanotherresearchtraditiondevelopedbySidman(1992)andothers(DougherandMarkham,1996).Employingtheso-calledsymbolicmatching-to-sampleconditioningprocedure,Sidmanbeganbyattemptingtoteachverballybackwardhumansthecorrespondenceamongpictogramsofitemssuchasacar,abed,anear,andsoonwiththewrittenwords"car,""bed,""ear,"andsoforth.Laterheusedpurelyarbitrary,initiallymeaninglessstimuli.AlthoughSidmanhimselfdidlittleanimalworkonthis(Sidmanetal.,1982),hisdesignslendthemselvestosuchanenterprise. 106JuanD.Deliusetal.Table6.1Minimalequivalenciesdesignemployingasymbolicmatching-to-sampleprocedure1.Matchingtraining2.Reflexivitytesting-LAB+A=BoKAAoA=A?+BAL-A=BoAAKoA=A?-MBC+B=CoLBBoB=B?+CBM-B=CoBBLoB=B?-BKL+K=LoKKAoK=K?+LKB-K=Letc....-CLM+L=M+MLC-L=M3.Symmetrytesting4.TransitivitytestingoKBAoB=A?oMACoA=C?oABKoB=A?oCAMoA=C?oLCBoC=B?oCKMoK=M?oBCLoC=B?oMKCoK=M?etc....etc....Thesamplestimuliarepresentedonamiddlepeckingkeyandthecomparisonstimuliarepresentedontwosidekeys.TherearesixdifferentstimuliA,B,CandK,L,M.Thesymbols+,-,andostandforreward,penalty,andnonreinforcement,respectively(modifiedfromSidman,1992bytheinclusionofunreinforcedtesttrials).Adaptedforpigeons,theprocedurecouldrunasfollows:OnagiventrialstimulusA,saypattern#mightbeshownonthemiddlekey;whenpeckeditwouldswitchonstimulusB,saypattern§ononeside-keyandstimulusK,saypattern@ontheotherside-key.ThecomparisonstimulusBisdennedasmatchingthesamplestimulusA,andifthepigeonpecksit,itisrewarded.IfitpecksthecomparisonstimulusL,dennedasnotmatchingthesamplestimulusA,itispenalized.ThisteachesthebirdstomatchthesamplestimulusAbychoosingthestimulusBandthustopossiblylearntheequivalencerelationA=B.Table6.1sketchesadesignattemptingtoteachpigeonstheequivalencerelationsA=B,B=CandK=L,L=M,eachletterstandingforadifferentstimulus.Part2delineatestestsforreflexivity,thatis,forwhetherthesamestimuliservingbothasasampleandacomparisonarerecognizedasequivalent.Part3sketchestestsforsymmetry,thatis,whetherwhenthestimulithatpreviouslyservedassamplesarenowusedascomparisonstimuli,andconverselyarestillrecognizedasequivalent.Part4sketchestestsfortransitivity,whetherpigeonscanderivefromthepreviouslylearnedequivalenciestheemergentequiv•alenciesA=CandK=M.OnlywhenpigeonspassallthesetestswouldtheycommandwhatSidmancallstheequivalenceclasses{A,B,C}and{K,L,M}.Ifthepigeonslearnedpart1onthebasisofaconfiguralratherthanarelationalstrategy,theywouldofcoursenotpassthem. StimulusEquivalenciesThroughDiscriminationReversals107Languagecompetenthumanspassthetestswithoutmuchdifficulty,butlanguagedeficientsubjectstendtofailononeoranothertestandonlymasterthecorre•spondingstagesafterremedialreinforcedtrainingwiththem.Normalchildrenbegintofullycommandequivalenceclasseswhentheyarelinguisticallycompetent,attheageofaboutsixyears(Sidman,1992).Youngerchildrentendtohavedifficultieswiththesymmetryandtransitivitytests(Valero-AguayoandLuciano-Soriano,1996).WhilenonhumanprimateshavemanagedtopasssomeoftheSidmancriteria(D"Amatoetal.,1985;YamamotoandAsano,1995),pigeonshavemostlyprovenincapableofpassinganyofthem(Lipkensetal.,1988;Jitsumori,1990;Fersenetal.,1992).OnlyKunoetal.(1994),usingadesignbypassingthereflexivityandsymmetryhurdle,gotoneoffourpigeonstopassthetransitivitytest.ZentallandUrciuoli(1994)arguethatpigeonsvariouslypassedthereflexivity,symmetry,andtransitivitytests,buttheirevidenceispiecedtogetherfromamethodicallyheterogeneouscollec•tionofexperiments.ThefactthatVaughan(1988)usedadifferentparadigmanddidnottestforreflexivity,symmetry,andtransitivitycausedSidmantobeinitiallycriticalofhisequivalenceclassdemonstration.However,whenhefoundoutthathumansubjectsbenefitedfromamultiplereversalpretrainingwithregardstopassingthesetests,herelaxedhisviewssomewhat(Sidmanetal.,1989;Sidman,1992).ReversalsContinuedAllstudiesusingthereversaltechniquewithpigeons,whethersuccessfulornot,employedmanyand/orcomplexstimuli.Becausethishamperstheanalysisofequiv•alenceclassformation,Deliusetal.(1995)rananexperimentemployingonlyfoursimplestimuli.Usingasimultaneoustwo-keyconditioningprocedure,pigeonsweretaughttoconcurrentlydiscriminatethecolorsred/greenandblue/yellowaccordingtotheschemeA+K-,B+L-,wherethe+and-signsmeanthatpeckswererewardedwithfoodorwerepenalizedwithtime-out.Whenthebirdshadlearnedthistask,thereinforcementallocationswerereversedtoA-K+,B-L+,andwhentheyhadlearnedthistheywereagainreversedtoA+K-,B+L-(table6.2[1]).Thereversalprocedurewasrepeatedabout35times,untilthebirdsbecameproficientinswitchingtheirstimuluschoices.ThiswastoensurethatthebirdslearnedthattheA,BandK,Lstimuliwereconsistentlyyokedtogether.SubsequenttestsfortheequivalenciesA=BandK=Linvolvedspecialreversalsessionswhereonlyoneofthetwodis•criminationpairswaspresentedduringthefirsttenorsotrials.Theseleadingtrialsgavethebirdstimeenoughtoadoptanatleast85percentcorrectresponding.Duringtheremainderofthesessiontheother,trailingdiscriminationpairwasadditionallypresented.Thefirsttwotrialswiththistrailingpairwentunreinforced.Thepigeons 108JuanD.Deliusetal.Table6.2Basicdesignofmultiplereversalequivalencetrainingandtesting,typicalsequencesofsimultaneousdis•criminationtrials1.EquivalenciestrainingA/K=B/LdiscriminationreversalreversalreversalA+B+B+B+A+A+B+...A+A-B-B-A-B-...A-B+B+A+...B-...K-L-L-L-K-K-L-...K-K+L+L+K+L+...K+L-L-K...L+...2.ReversaltestforA/K=B/L?andB/L=A/K?leadingtrailingleadingtrailingA+A+A+A+...A+BoA+BoA+B+...A+B-B-B-...B-AoAoA-B-...K-K-K-K-...K-LoK-LoK-L-...K-L+L+L+...L+KoKoK+L+...3.FullandhalfreversaltestsforA/K=B/L?fullreversalhalfreversalnormalizationB-B-A-B-A-B+B+A+A+...A+B-B-B-A+A+B-...B-B+A+B+...L+L+K+L+K+L-L-K-K-...K-L+L+L+K-K-Z.+...L+L-K-L-...Thestimuliformingthediscriminatorypairsareshownoneabovetheother.Reinforcementallocationreversalsaremarkedbyswitchesfromstandardtoitalicfontandback.Thesymbols+,-,andostandforreward,penalty,andnonreinforcement.Forsimplicitytherandomizedleft-rightpositionofthestimuliisnotrepresented(afterDeliusetal.1995).showed37percentcorrectinitialtrialswiththeleadingpairand48percentcorrectinitialtrialswiththetrailingpairtrials.Inaccordancewiththeequivalencehypothe•sis,thereversalexperiencewiththeleadingpairfacilitatedthechoiceswitchwithrespecttothetrailingpair.Afterthepigeonswereadditionallytrainedwithreversalsincorporatingsometrialswithhalf-keyred/half-keyblue(A|B)andhalf-keygreen/half-keyyellow(K|L)stimuluspairsmeanttostrengthentheassociationsthroughspatialcontiguity,renewedtestsyieldedamoreclearlysignificant39percentcorrectleadingand49percentcorrecttrailingpairdifference.However,thetestresultscouldalsohavearisenifthebirdshadadoptedanon-discriminative,50percentcorrectrespondingtowardtheinitialtrailingpairpre•sentationsafternoticingtheleadingpairreversal.AnadditionaltestsessionseriesusingadesignbyNakagawa(1992)andnotopentothisoptionwasthusrun(table6.2[3]).Halfofthesessionsimplementedfullreversalsaffectingbothpairsinalikemanner.Theyalternatedwithsessionsinvolvinghalfreversalswhereonediscrimi•nationpairwassubjecttoareinforcementreversalbuttheotherwasnot.Becausethesehalfreversalswereatoddswiththeequivalenciespresumablyinducedbythefulltrainingreversals,itwasexpectedthebirdswouldnotadjustaswelltothemastothefullreversals.Thishappened:thepigeonsshowedanaverage63percentand70percentcorrectchoicesduringthefirsttentrialsofthesetwotypesofsessions.This StimulusEquivalenciesThroughDiscriminationReversals109Figure6.1(a)conditioningplatformusedfortheSiemannandDelius(1998a),Siemann(1998),andJitsumorietal.(1999)studies(fromXiaetal.1996);(b)stimuliemployedandaverageperformanceduringtheinitialtrialswiththeleadingandtrailingpairsduringthetestreversalintheSiemannandDelius(1998a;clear)andtheSiemann(1998;stippled)studies.significantdifferencecontradictedthesimplerexpectancythatarelearningoftworeversedpairsdiscriminationwouldbemoredifficultthantherelearningofasinglereversedpairdiscrimination.TheresultthusrathersupportsthenotionthatthecolorstimulihadbecomepartlyequivalentinanA=B,K=Lmanner.ComparableresultswereobtainedinasimilarhalfandfullreversalexperimentcarriedoutbyZentalletal.(1991).Becauseredisnexttoyellowandgreenisnexttoblueonthewavelengthspectrum,itseemedpossiblethatstimulusgeneralizationcouldhaveworkedagainsttheintendedred=blueandyellow=greenequivalencies.SiemannandDelius(1998a)thereforeconductedamultiplereversalexperimentusingshapes.Pigeonswerecon•ditionedusingplatformsattachedtothecages.Thestimuliwerelightdiodearraypatternspresentedbelowthetwotransparentkeys.Feedersdeliveredafewgrainsofmilletontoeitherkey(figure6.1a).ShapeschosentobephysicallydissimilarformedthepairsA+-K-+andB+-L-+,thesymbolcombinations+-and-+indicat•ingtherepeatedandsynchronousreinforcementreversalstheyweresubjectedto.Thesecameintoeffectwheneverthepigeonshadreachedacriterionperformanceof70percentcorrectchoiceswithina40-trialblock.Testblockswereintroducedwhenthenumberoftrialsthatthebirdsneededtoreachthiscriterionroughlystabilizedaftersome100reversals.Therewere20blocksofabout120trialsstructuredaccordingtotheleading/trailingdesignexplainedearlier.Onlywhentheleadingpairhadbeenpresentedsome20timesinarowwasthetrailingpairpresentedaswell.Allthreepigeonsyieldedabetterperformanceduringtheinitialtrialswiththetrailingpairthanduringtheinitialtrialwiththeleadingpair.Theconditional 110JuanD.Deliusetal.extinctionprocessconsideredearlierisunlikelytohaveappliedbecausetheyyieldedabetterthan50percentchanceperformancewiththetrailingpair.Themeanscoreswere63percentand17percent,respectively(figure6.1b).Thelargedifferencebe•tweenthescoresindicatesasizeablereversaltransferacrossthediscriminationpairsandthusamarkedA/K=B/Lequivalenciescommand.Siemann(1998)ranarepeatexperimentwithasimilarprocedurebutemployingdifferentstimuli(figure6.1b).Thethreenewpigeonsagainyieldedaperformanceadvantagewiththetrailingpairovertheleadingpair,themeanscoresbeing45percentand32percentcorrect(figure6.1b).Thelesserdifferencebetweenthescorescouldbeduetothefactthatoneofthestimuluspairsconsistedoftwosomewhatsimilarshapes.Althoughsupportingtheequivalenceformationthroughmultiplereversals,ourstudiescouldnotreproducethelargereversaltransfereffectobtainedbyVaughan(1988).Apartfromsomeproceduraldifferencesofdoubtfulrelevance,thereisthefactthatweemployedafewsimplestimuliselectedtobedissimilarwithineachofthestimulussets,whereasVaughanemployedmanycomplexbutgenerallysimilarstim•uliwithineachofthestimulussets.Thereisevidencethatsimilarityandthenumberofstimuliusedfortrainingplayaroleintheestablishmentofrelationalprinciplesinpigeons(Wrightetal.,1988;Delius,1994).Thisisundoubtedlybecausethesimilarityfactorallowsthegroupingprocesstogrowuponalreadypresentstimulusgeneral•izationlinksandbecausethenumberfactorhelpstooverloadthenonrelational,rotelearningoptiontowhichpigeonsareotherwiseprone(VaughanandGreene,1984;FersenandDelius,1989).EquivalenciesbyReversalsJitsumorietal.(2000)employedmorestimuliofacontrolledsimilarity,aswellasanimprovedmultiplereversalprocedure.Fivepigeonsmadeupthesimilaritygroupaboutwhichwemainlyreporthere.Twosetsoffourlightdiodestimuliservedtobeginwith(figure6.2a,b).Humanobserversjudgedthemtobesimilarwithinsetsandtobedissimilarbetweensets.TheallocationofthedifferentpatternswithintheA,B,C,DandK,L,M,Nsetsofstimuliwasrandomizedacrossthebirds.ThepigeonswerefirsttaughtthetwinequivalenciesA=BandK=L(abbre•viated:A/K=B/L)withthestimuluspairsA+K-,A+L-,B+L-,B+K-pre•sentedinrandomorder(table6.3[1]).Thetrainingcontinueduntilthebirdsconsistentlyachievedabetterthan80percentcorrectperformance.Thereinforce•mentallocationtothediscriminationpairswasthenreversedtoA-K+,A-L+,B-L+,B-K+.Trainingcontinueduntiltheabovecriterionwasreattained.Rein- StimulusEquivalenciesThroughDiscriminationReversals111Figure6.2LightdiodestimuliusedbyJitsumorietal.(2000).Thesimilaritygroupofpigeonsfirstdealtwiththepatternsshownintherowsaandbandlateradditionallywiththoseshowninrowc.Thedissimilaritygrouptobementionedlaterdealtwiththepatternsshowninrowsdande.forcementswerethenreversedagain,andsoonuntilthebirdsexhibitedmorethan80percentcorrectwithinthefirstblocksafterareversalduringthreesuccessivereversals.ThenthepigeonswheretaughttheequivalenciesC/M=D/NusingthesameprocedurewiththediscriminationpairsC+-M-+,C+-N-+,D+-M-+,D+-N-+,wherethesymbols+-and-+indicatethemultiplereversalstheyweresubjectedto(table6.3[1]).Thebirdsachievedthecriteriononbothtaskswithin30-70reversals.TotestwhethertheequivalenciesA/K=B/L,D/N=C/MhadbeenformedweexaminedwhetherareinforcementreversalaffectingthestimuliA,KandD,NwouldtransfertothestimuliB,LandC,M,andtheconverse.Testsessionsbeganwithreinforcedpresentationsofonesetofstimuluspairs,theleadingpairs,untiltheusualcriterionwasreached.Atestblockfollowed.Itconsistedoffurthertrainingtrialswiththeleadingpairsandrandomlyinterspersed,unreinforcedprobetrialswiththeothersetofpairs,thetrailingpairs.Asecondtestsessionproceededidentically,exceptthatthereinforcementsoftheleadingpairswerereversed(table6.3b).OnetestsessiondyadinvolvedtheleadingpairsA+-K-+,D+-N-+,A+-N+-,D+-K+-,andthetrailingpairsBoLo,CoMo,BoMo,CoLo.TheotherdyadinvolvedtheleadingpairsB+-L-+,C+-M-+,B+-M-+,C+-L-+andthetrailingpairsAoKo,DoNo,AoNo,DoKo.Thesessionpairswerearrangedsothatwhentheaveragetestscoreacrossthemwasover50percentcorrectthisindicateda 112JuanD.Deliusetal.Table6.3Sketchofthebasicexperimentaldesign1.EquivalenciestrainingA/K=B/LandC/M=D/NtrainingreversalA+A+B+A+B+...A+A+B+B-A-A-B-...A-B-B-K-L-K-K-L-...L-K-L-L+K+K+K+...K+L+L+trainingreversalC+D+D+D+C+...C+D+D-D-C-C-...N-M-N-N-M-...M-¥S-M+M+N+M-...2.EquivalenciestestingA/K=B/L?andC/M=D/N?trainingtesttesttestA+A+D+A+...A+D+BoA+BoD+D+CoA+A+D+K-K-N-N-...N-K-LoK-MoN-K-LoK-K-N-testreversaltestiesttestCo...B-B-C-B-...C-AoAoB-Do...Mo...L+M+L+L+...M+KoNoM+No...Seetable6.2andthetextforfurtherexplanations(afterJitsumorietal.,2000). StimulusEquivalenciesThroughDiscriminationReversals113Figure6.3Equivalenciesthatweretrained(thicklines)andequivalenciesthatcouldemergebytransitivity(thinlines).Theequivalenciestestedareflankedbythenumberofbirdscommandingthem.Equivalenciesinferredtohaveornothavebeencommandedbymostbirdsaremarkedwith!and?symbols;(a)beforeand(b)aftertheincorporationofdissimilarstimuli.(ModifiedfromJitsumorietal.2000)choicetransferfromtheleadingpairstothetrailingpairs,anapproximately70per•centcorrectscoresignalingasignificanttransfer.Theactualscoresforallpairsandeachbirdexceededthatlevelbybeingbetween78percentand100percentcorrect,implyingacommandovertheequivalenciesA/K=B/L,D/N=C/M.Figure6.3asummarizesthesefindings.ThepigeonswerenexttaughttheequivalenciesA/K=D/N,B/L=C/M,usinganalogousprocedures.Theircommand,however,wasnotdirectlytested.Instead,wetestedwhetherthebirdsdominatedtheequivalenciesA/K=C/M,B/L=D/Nthathadnotbeenexplicitlytrainedbutwhichthepigeonscouldinprinciplederivebytransitivityiftheyhadformedtheequivalenceclasses{A,B,C,D}and{K,L,M,N}(figure6.3a).Thetestsessionswerestructuredasbeforebutinvolvedtheleadingtraining/trailingtestpairsA+-K-+/C0M0,C+-M-+/A0K0,B+-L-+/D0N0,andD+-N-+/B0L0.Thebirdsachievedsignificantabovechancelevelperfor•mancesofbetween72percentand100percentcorrecttrialswiththevarioustestpairs,exceptthatonebirdthatwasweakontheD0N0pair.Figure6.3asummarizestheresults.ThemasteryoftheseA/K=C/MandB/L=D/NequivalenciesbyfourbirdsindirectlydemonstratedthattheyalsocommandedtheequivalenciesA/K=D/NandB/L=C/M.Theyhadthusintegratedtheequivalenciesintoanetworksothatthestimuliwereassociativelyclusteredintwoseparate{A,B,C,D}and{K,L,M,N}classes.Withintheseclasseseachandallmemberstimuliwerecapableofsignalingareinforcementswitchcausingaresponseswitchtotheremainingmembers. 114JuanD.Deliusetal.DissimilaritiesInterfereTheequivalenceformationreflectedbytheseresultswithsimilar-withinanddissimilar-betweenstimuliisfarbetterthanthatobtainedduringourearliereffortsusingasimilarprocedurebutwithdissimilarwithin-classstimuli.Indeed,theresultscomparewiththoseofVaughan(1988)thatalsoinvolvesimilarwithin-classstimuli.Tofurtherassesstheroleofstimulussimilarity/dissimilarity,ourpigeonswerenowconfrontedwithadditionalequivalencetasksinvolvingdissimilarstimuli.Fourpat•ternsthathumanobserversjudgeddissimilartooneanotheranddissimilartothepatternsconstitutingthetwoearliersetswereconstructed(figure6.2c).Theassign•mentofthesepatternsasV,WX,YstimuliwasrandomizedacrossthebirdsthatweretrainedtolearntheequivalenciesA/K=V/XandC/M=W/Yaccordingtothesameprocedureemployedbefore.Althoughthebirdswerenowexperiencedwiththediscriminationreversalroutine,theytooklongertolearnwiththesestimulithanwiththeearlierones.ThebirdswerethentestedintheusualmannerusingtheleadingpairsB+-L-+andD+-N-+andthetrailingpairsVoXoandWoYo.ThistestedthemforthemasteryoftheequivalenciesB/L=V/X=D/N,B/L=W/Y=D/N,whichtheyhadnotbeenexplicitlytaughtbutthattheycouldpotentiallyderive.ThetestscoreswiththeWoYopairwereall,exceptinonebird,significantlyabovechance,butwiththeVoXopairtheywereonlysignificantfortwobirds(figure6.3b).ItseemsthatsomebirdshadlearnedtheA/K=V/XlesswellthantheC/M=W/Yequivalencies.WethentestedwhetherthebirdshadderivedtheuntaughtequivalenciesV/X=W/Yusingtheleadingpairs/trailingpairsV+-X-+/W0Y0andW+-Y-+/V0X0.Thetestscoreswereallclosetochanceexceptthoseofonebirdthatachievedsig•nificantscores.AfterasubsequenttrainingoftheV/X=W/Yequivalencies,arepeatofthesametestrevealedsomeasymmetryinthesensethatalthoughtwobirdsscoredsignificantlyinbothtests,twobirdsweresuccessfulonlywiththeWoYotestpair,andonebirdfailedonbothtests.TheasymmetrywasprobablyduetothefactthatmostpigeonshadearlierlearnedtheequivalenciesC/M=W/YbutnotlearnedtheequivalenciesA/K=V/X(figure6.3b).ThusfourpigeonsofJitsumorietal."s(1999)so-calledsimilaritygroupendedupincorporatingdissimilarstimuliintotheequiva•lenceclasses{A,B,C,D,W},{K,L,M,N,Y},andoneformingtheprobablydis•jointclasses{A,B,C,D},{K,L,M,N},{V,W},{X,Y}.Obviouslytheequivalenceformationthroughmultiplereversalswasevenlydis•similarwithinandbetweenstimuli,althoughinprinciplewithinthecompetenciesofpigeons,itisamoredifficulttaskthanwhenthestimuliaresimilarwithintheintendedequivalenceclasses.Thisagreeswiththeresultsobtainedwithaseparate StimulusEquivalenciesThroughDiscriminationReversals115so-calleddissimilargroupoffourpigeons.Theyweretrainedandtestedinthesamewayasthesimilargroupduringthefirstphaseoftheexperiment,butdealtwiththeA,B,C,DandK,L,M,Nsetscomposedofthedissimilarpatternsshowninfigure6.2d,e.All,barone,passedthetestsfortheA/K=D/Nequivalencies(mean79percentcorrect),buttheyallexceptforonefailedthetestsfortheC/M=D/Nequivalencies(mean66percentcorrect).Whentestedforcommandoftheequiv•alenciesA/K=C/M,B/L=D/Nderivablebytransitivity,allthebirdsexceptonepassedthetestforthefirst(mean80percentcorrect),butallbirdsfailedthetestforthesecondequivalencies.Overall,thedissimilargrouppresentedapatternofresultsreminiscentofthatoftheso-calledsimilargroupofbirdswhenthesewereconfrontedwithdissimilarstimuli:theresultswerecharacterizedbyindividualities.AsdetailedbyJitsumorietal.(1999),threebirdsofthedissimilargroupendedupcommandingtheequivalenceclasses{A,B,C},{K,L,M},andonebirdcommandingonlytheequivalenceclasses{A,B},{K,L}.EpilogueJitsumorietal."s(2000)studyprovidesbetterevidenceforequivalenceclassforma•tioninpigeonsthanthatproducedbyseveralearlierreversalstudies.Apartfromtheimplementationofstrictreversalcriteria,thepigeonsmayhaveprofitedfromtheadditionaltrainingwithcrossedstimuluscombination(theA+-L-+,B+-K-+pairs,inadditiontothestandardA+-K-+B+-L-+pairs)andtheconcurrent/successiveteachingofseveralequivalencies(theequivalenciesA/K=B/L,C/M=D/N,A/K=B/L,D/N=C/MinsteadofonlytheequivalenciesA/K=B/L).ThereversaltransfereffectsrevealedarecomparabletothoseobtainedbyVaughan(1988).ThefindingssubstantiateSidman"s(1992)admissionthatmultiplereversalsmaypowerfullypromotetheformationofequivalenceclasses.Heconsideredreflex-ivity,symmetry,andtransitivitytobeessentialpropertiesofequivalenceclasses.WithinthemultiplereversalprocedureusedbyJitsumorietal.(2000),thereflexivitypropertyisnotmeaningfulandthesymmetrypropertyisinevitablytaught.Butthesymmetryandthetransitivitypropertyofderivedequivalencieswasvariouslytestedandverified.Becausethedemonstrationsweremostlyassociatedwithsizeableerrormargins,theequivalenciescommandedbythepigeonsareneverthelessbestviewedasconformingwithfuzzyratherthanformallogicsets(YagerandZadeh,1994).Similaritywithinstimuliclassesandthedissimilaritybetweenclassstimulusclassesappeartofosterequivalenceclassformation.Thisisnotsurprising,asearlierpigeonstudieshaveshownthatsuchrelationshipsbetweencategorystimulifacilitatesthe 116JuanD.Deliusetal.formationofopen-endedcategoriesthatcantransfertonovelexemplars(e.g.,AstleyandWasserman,1992;JitsumoriandYoshihara,1997;seealsoMackintosh,thisvolume).Indeed,thesuccessfulequivalenceclassformationwiththesimilarpatternsmightwellhaveimportantlyreliedonintraclassstimulusgeneralizationandinter-classdiscriminabilityofthememberstimuli.However,itwouldbeshort-sightedtoascribealltheeffectsobtainedtosolelythesefactors.Jitsumorietal.(2000)foundthatthepatternsthathadendedupbelongingtooneortheotherofthesimilarstimulusequivalenceclasses{A,B,C,D}and{K,L,M,N}werestilleasilydis•criminatedbythepigeons.Also,mostofthesimilaritygrouppigeonslearnedtoincludethedissimilarWandYstimuliintotheequivalenceclassesthattheyfinallydominated.Mostofthedissimilaritygrouppigeonslearnedtoformsomeequiva•lencieswithexclusivelydissimilarstimuli.ItmustalsoberememberedthatVaughan"s(1988)excellentequivalenceformationwasobtainedwithstimulithatwerejustassimilarwithinsetsasbetweensets.Itisthusstillpossiblethatasimilaritybetweenstimulifavorsequivalenceformationevenifnodissimilarityseparatesthemembersofthestimulussets.Thismayarisebecausethemultiplereversalprocedureimple•mentsadirectdiscriminationbetweenclassstimulibutonlypromotesanindirectgeneralizationwithinclassstimuli.Fromanethologicalviewpoint,thefitnessutilityoftheabilitytoclassdiversestimuli,items,oreventsaseatable,matable,threatening,shelter-spending,home-directing,andsoforthforanimalssuchasthepigeonisvirtuallybeyonddoubt.Thatnaturalstimulisharingsuchfunctionalpropertieswilloften,butnotalways,alsoshareperceptualsimilaritiesisquitepatent.Thatnaturalfunctionalclasseswillsometimesvaryovertimesoastorequirecontinuoustrackingbyrelearningissimi•larlyplausible.ClosetoVaughan"s(1988)experiment,onemaysuspectthattherec•ognitionofclassesofrealtreesthatbearfruitsandberriesatgiventimesoftheyearbutnotothersmightbedirectlyimportantforfrugivorousbirds.Perhapsthisislesstrueforthedomesticpigeon,butthepigeonmayhavetosolverelatedproblemswhensearchingforsitespromisingprofitableforaging,auspiciousnesting,orreassuringfamiliarity.Thus,perhapsthecompetenceforformingsuchfunctionalclassescouldbeessentialfortheirsurvivalandreproduction.Undoubtedlyasimilarargumentappliestootherbehaviorallyadvancedspecies.Itseemsprobablethatmostmammalswillshowcapableoffullequivalenceclassformationastheexperimentalproceduresareimproved(rats:Nakagawa,1992;Roberts,1996;dolphins:FersenandDelius,2000;sealions:SchustermanandKastak,1993;chimpanzees:YamamotoandAsano,1995).Whethertheexperimentalproceduresthathavebeenhithertousedintheattemptstodemonstratesuchclassformationareecologicallywelladjustedisnotcertain.The StimulusEquivalenciesThroughDiscriminationReversals117evolutionarycircumstancethatthebehavioralcompetenciesofananimalspeciesmustbetunedtothedemandsthatthesocioecologicalnichemadeonitsrecentancestorssuggeststhatitmaybeaworthwhileefforttoadaptthelaboratoryproce•duresclosertothenaturalconditions.Theabilitiesofpigeonsregardingtheforma•tionofequivalenceclassesmightbecontextspecificandwemaynotyethavestrucktheoptimalcognitivemodule(Shettleworth,thisvolume).Itmightbeworthtryingotherproceduresthatemploymorevariedresponserequirements,morevariedrein•forcementconsequences,andmoretemporallypersistentschedulesofreinforcementreversalsthanthoseusedinthestudiesreviewedhere.Indeed,ifitwaspossibletodevisemoreefficientprocedures,itmaybepossibletoextendthereversalparadigmtoanalogicalequivalenceproblemsencapsulatedbythestatement"akeyistoalockasanopeneristoacan"(Premack,1988).Supposeonewouldtrainpigeonssepa•ratelyontheequivalenciesA/K=B/Lusingfood/no-foodreinforcementreversalsandtheequivalenciesC/M=D/Nusingwater/no-waterreinforcementreversals.Wouldtheythenrevealreversaltransferacrossthesepairsofequivalencies?Suchatransferinturnwouldargueagainstanarrowmodularspecializationofthiscognitivecompetence.Withthepossibleincreasesintasksophistication,itwillbenecessaryatsomepointtotakeintoconsiderationthattheavianbrainsizeismoreconstrainedinitspro•gressive,anageneticevolutionthanisthebrainofmammals.Thisisbecausetheflightcapabilitygenerallyimposesalimitonthebody/weightindependent,allometricgrowthofbrainweight,thesecondaryflightlessnessofsomeavianfamiliesnotwith•standing(Jerison,1973).Ifinhumansvariationsinindividualintelligenceonlyemergewhenthisspeciesischallengedwiththeharderitemsofintelligencetests,onecanexpectthatindividualpigeonsmayreachthelimitsoftheirintellectastheyarefacedwithevermoredifficulttasks.Moreover,pigeonsarealmostcertainlynotthecleverestavianspecies;thestudyofequivalenceformationinparrotsandcorvidsmaywellturnouttobemorerevealing(Deliusetal.,2000).Indeed,someofthevocallabelingcompetenciesexhibitedbyanAfricangreyparrotmayalreadyconsti•tuteinformalevidenceofaremarkablecapacityforequivalenceformationinatleastonesuchspecies(Pepperberg,1996;seealsoManabeetal.,1995,budgerigars).Theseevolutionaryandcomparativeconsiderationsalsoobligeustoconsidertheneuralmechanismsthatmusthaveevolvedtoenabletheneurallymoreadvancedanimals,includingpigeons,tolearnequivalenceclasses.Thisabilitymustbebasedonsecondarynetworksofneuronsfollowingtheprimarylayersofthevisualsystemengagedinextractingthebasalfeaturesofvisualstimuli.Inmammalsonewouldobviouslyconsidertheparticipationofsecondaryprojectionareasofthevisualcortex.Inbirds,withtheirratherdifferentbrainmake-up,theissueislesscertain.But 118JuanD.Deliusetal.forwhatitisworth,neuroanatomistsnowconsiderlargerpartsoftheavianforebrain(especiallythestructuresknownasektostriatumandneostriatum)asequivalenttothemammalianneocortex,inthattheyprobablyfulfillsimilarinformationprocessingonthebasisofananalogous,evenifwithahistologicallylessdistinctstructuring(Veenman,1997).Asmallpartofthiscorticalmassmayevenbeequivalenttotheprefrontalcortexofmammals(Aldavert-Veraetal.,1999),whichisinturnreckonedtobethemainsubstrateofthesupramodulargeneralintelligenceinhumans.Artifi•cialneuralnetworks,whichtosomeextentareconceivedtomimictheconnectivitiesofcorticalstructures,havealreadyshowntobeamplycapableofcategorizationandconceptualization-likefeatscomparabletothoseproducedbypigeonsandindeed,humans(Gluck,1991;RoitblatandFersen,1992).Thesynapticweightsoftheinputsofintermediate(orhidden)layerunitsorclustersofunitswill,withsuitablesimu•latedcategorizationtraining,cometoadjustthemselvessothattheseunitsorclusterswillmoreorlessindiscriminatelyrespondtoanystimulusbelongingtothecategorytheyhappentorepresentwhenthesestimuliareappliedtotheinputlayer.Thenet•workmodelmayprovehelpfulinexplainingwhytheequivalenceeffectsobtainedinthevariousreversalandmatchingstudieshavebeenrelativelyvariable.Allthefac•torsthatwehavetentativelyidentifiedascontributingtothatvariability—stimulussimilarities,stimuluscomplexities,numberofstimuli,andnumberofreversals—areofthekindalreadysuspectedtoaffectconceptformationinsuchnetworks.Networkscomposedofatleastthreelayersofneuronlikeunitsarenecessarytoaccountfortheformationofperceptualconcepts,wheretheunitsoftheintermediate,hiddenlayerarethoseprincipallymediatingtheconnectivitycodingofconcepts(Watanabeetal.,1993).However,preliminaryexplorationsuggeststhatfour-layernetworksmightbemoreeffectiveinimplementingequivalenciesthroughreversals,theadditionallayerenablingtheformationofadistinctresponse/reinforcementswitchcircuitry.Pigeonswouldinanycasenotbeshortofneurallayersasthefirststageoftheirvisualsystemalone,thetectumopticum,alreadycontainssome15layers(Gunturkiin,1991).Hiddenunitsare,inanycase,onlyeffectivelyrecruitedintonodesorclustersifthetrainingstimulussetsbearpredisposingsimilarities/dissimilarities,arecomplexenough(manystimuli,manyfeatures),andifreversalsoccuroftenenoughforanconcept-conformactivationoftheseunitstoyieldapro•cessingadvantage.Otherwise,suchneuralnetworkshaveatendencytosettleonaby-rotecategorization,two-layernetworklikemannerofprocessingthatlackstheelementofconcept-defininginter-stimulusassociations.Networksthatoperateondefiniteinstrumentalconditioningprinciples(SiemannandDelius,1998b)wouldappeartobethemostpromisingsubstratesforexploringthemechanismsofequiva•lenceclassformationinasmuchastheycanreadilyincorporatetheeffectsofrein- StimulusEquivalenciesThroughDiscriminationReversals119forcementswitchessocentraltothemultiplesynchronousreversalprocedurethathasbeenthemainsubjectofthischapter.SummaryThischapterarguesthat,contrarytowidespreadopinion,pigeonsarecapableofformingequivalenceclassesofvisualstimuli.ThemultiplereinforcementreversalprocedureoriginallyintroducedbyVaughan(1988)andrecentlydevelopedbyJitsu-morietal.(2000)mightbemoreadvantageousinthisrespectthanthemorepopularsymbolicmatching-to-sampleprocedureadvocatedbySidman(1992).Thismaybemainlysobecausetheformermethodobviatesasymmetrystagethatisessentialtothelattermethod.Althoughweshowthatphysicalstimulussimilaritiesfacilitatetheformationofequivalencenetworks,ithasalsobeendemonstratedthatfunctionalsimilaritiesalonecanbesufficienttoyieldequivalenceassociations.Itissuggestedthatequivalenceformationbyreinforcementreversalmaybewithinthecapabilitiesofsimplemultilayerneuralnetworksandthatavianspossessnervoussystemsthatarecomplexenoughtoincorporatetherequisiteneuronalcircuitry.Fromanevolution•aryperspectivetherecanbelittledoubtthatequivalenceformationcapacitiesmustbeinhighdemandinaviannaturalhabitats.AcknowledgmentsWethanktheDeutscheForchungsgemeinschaftandtheJapaneseMinistryofEducationforsupport,Prof.I.Morgado-Bernal(Barcelona)forhospitality,andJ.Grante(Exeter)forcorrections.ReferencesAldavert-VeraL,Costa-MiserachsD,DivacI,DeliusJD(1999)Presumed"prefrontalcortex"lesionsinpigeons:Effectsonvisualdiscriminationperformance.BehavioralBrainResearch,102:165-170.AstleySL,WassermanEA(1992)Categoricaldiscriminationandgeneralizationinpigeons:Allnegativestimuliarenotcreatedequal.JournalofExperimentalPsychology:AnimalBehaviorProcesses,18:193-207.AstleySL,WassermanEA(1998)Noveltyandfunctionalequivalenceinsuperordinatecategorizationbypigeons.AnimalLearningandBehavior,26:125-138.BhattRS,WassermanEA(1989)Secondarygeneralizationandcategorizationinpigeons.JournaloftheExperimentalAnalysisofBehavior,52:213-224.BhattRS,WassermanEA,ReynoldsWF,KnaussKS(1988)Conceptualbehaviorinpigeons:Catego•rizationofbothfamiliarandnovelexamplesfromfourclassesofnaturalandartificialstimuli.JournalofExperimentalPsychology:AnimalBehaviorProcesses,14:219-234. 120JuanD.Deliusetal.D"AmatoMR,SalmonDP,LoukasE,TomieA(1985)Symmetryandtransitivityofconditionalrelationsinmonkeys(Cebusapella)andpigeons(Columbalivid).JournaloftheExperimentalAnalysisofBehavior,44:35-47.DeliusJD(1992)Categoricaldiscriminationofobjectsandpicturesbypigeons.AnimalLearningandBehavior,20:301-311.DeliusJD(1994)Comparativecognitionofidentity.In:Internationalperspectivesonpsychologicalscience:Leadingthemes(BertelsonP,EelenP,d"YdewalleG,ed),pp25-40.Hillsdale,NJ,Hove,UK:Erlbaum.DeliusJD,AmelingM,LeaSEG,StaddonJER(1995)Reinforcementconcordanceinducesandmaintainsstimulusassociationsinpigeons.PsychologicalRecord,45:283-297.DeliusJD,SiemannM,EmmertonJ,XiaL(2000,inpress)Cognitionsofbirdsasproductsofevolvedbrains.In:Brain,evolutionandcognition(RothG,WullimannMF,ed),NewYork,Wiley.DougherMJ,MarkhamMR(1996)Stimulusclassesandtheuntrainedacquisitionofstimulusfunctions.In:Stimulusclassformationinhumansandanimals(ZentallTR,SmeetsPM,ed),pp137-152.Amster•dam:Elsevier.FersenLv,DeliusJD(1989)Long-termretentionofmanyvisualpatternsbypigeons.Ethology,82:141-155.FersenLv,DeliusJD(2000)Transferofdiscriminativeresponsereassignmentsindolphins(Tursiopstrun-catus):Evidenceforconceptformationwithauditorystimuli.AnimalCognition,inpress.FersenLv,EmmertonJ,DeliusJD(1992)Unexpecteddiscriminationstrategyusedbypigeons.Behav•iouralProcesses,27:139-150.FersenLv,LeaSEG(1990)Categorydiscriminationbypigeonsusingfivepolymorphousfeatures.JournaloftheExperimentalAnalysisofBehavior,54:69-84.FodorJ(1998)Whenisadogadog?Nature,396:325-327.GluckMA(1991)Stimulusgeneralizationandrepresentationinadaptivenetworkmodelsofcategorylearning.PsychologicalScience,2:50-55.GunturkiinO(1991)Thefunctionalorganizationoftheavianvisualsystem.In:Neuralandbehaviouralplasticity:Theuseofthedomesticchickasamodel(AndrewRJ,ed),pp92-105.Oxford:OxfordUni•versityPress.HerrnsteinRJ(1990)Levelsofstimuluscontrol:Afunctionalapproach.Cognition,37:133-166.HerrnsteinRJ,LovelandDH(1964)Complexvisualconceptinthepigeon.Science,146:549-551.HerrnsteinRJ,VaughanW,MumfordDB,KosslynSM(1989)Teachingpigeonsanabstractrelationalrule:Insideness.PerceptionandPsychophysics,46:56-64.JerisonHJ(1973)Evolutionofthebrainandintelligence.NewYork:AcademicPress.JitsumoriM(1990)Noequivalencesthroughsymbolicmatchinginpigeons(unpublishedexperiment).JitsumoriM(1993)Categorydiscriminationofartificialpolymorphousstimulibasedonfeaturelearning.JournalofExperimentalPsychology:AnimalBehaviorProcesses,19:244-254.JitsumoriM,SiemannM,LehrM,DeliusJD(2000,subm.)Theformationandexpansionofequivalenceclassesinpigeons:Emergentstimulusrelationsthroughdiscriminationreversals.JitsumoriM,YoshiharaM(1997)Categoricaldiscriminationofhumanfacialexpressionsbypigeons:Atestofthelinearfeaturemodel.QuarterlyJournalofExperimentalPsychology,50B:253-268.KellerFS,SchoenfeldWN(1950)Principlesofpsychology.NewYork:Appleton-Century-Crofts.Kirpatrick-StegerK,WassermanEA(1996)Thewhatandthewhereofthepigeon"sprocessingofcomplexvisualstimuli.JournalofExperimentalPsychology:AnimalBehaviorProcesses,22:60-67.KunoH,KitadateT,IwamotoT(1994)Formationoftransitivityinconditionalmatchingtosamplebypigeons.JournaloftheExperimentalAnalysisofBehavior,62:399-408. StimulusEquivalenciesThroughDiscriminationReversals121LeaSEG(1984)Inwhatsensedopigeonslearnconcepts?In:Animalcognition(RoitblatHL,BeverT,TerraceHS,ed),pp263-277.Hillsdale:Erlbaum.LeaSEG,RyanCME,KirbyRM(1990)Instancetocategorygeneralizationfollowingpigeon"slearningofanartificialconceptdiscrimination(unpublishedmanuscript).LipkensR,KopPFM,MatthijsW(1988)Atestofsymmetryandtransitivityintheconditionaldiscrimi•nationperformancesofpigeons.JournaloftheExperimentalAnalysisofBehavior,49:395-409.LubowRE(1974)Higher-orderconceptformationinthepigeon.JournaloftheExperimentalAnalysisofBehavior,21:475-483.ManabeK,KawashimaT,StaddonJER(1995)Differentialvocalizationinbudgerigars:Towardsanexperimentalanalysisofnaming.JournaloftheExperimentalAnalysisofBehavior,63:111-126.NakagawaE(1992)Effectsofovertrainingonreversallearningbyratsinconcurrentandsinglediscrim•inations.QuarterlyJournalofExperimentalPsychology,44B:37-56.PepperbergIM(1996)CategoricalclassformationbyanAfricangreyparrot(Psittacuserithacus).In:Stimulusclassformationinhumansandanimals(ZentallTR,SmeetsPM,ed),pp71-90.Amsterdam:Elsevier.PremackD(1988)Mindswithandwithoutlanguage.In:Thoughtwithoutlanguage(WeiskrantzL,ed),pp46-65.Oxford:Clarendon.RobertsWA(1996)Stimulusgeneralizationandhierarchicalstructureincategorizationbyanimals.In:Stimulusclassformationinhumansandanimals(ZentalTR,SmeetsPM,ed),pp35-54.Amsterdam:Elsevier.RoitblatHL,FersenLV(1992)Comparativecognition:Representationandprocessesinlearningandmemory.AnnualReviewofPsychology,43:671-710.SchustermanRJ,KastakD(1993)ACaliforniasealion(Zalophuscalifornianus)iscapableofformingequivalencerelations.PsychologicalRecord,43:823-839.SidmanM(1992)Equivalencerelationsandbehavior:Aresearchstory.Boston:AuthorsCooperative.SidmanM,RauzinR,LazarR,CunninghamS,TailbyW,CarriganP(1982)Asearchforsymmetryintheconditionaldiscriminationofrhesusmonkeys,baboonsandchildren.JournaloftheExperimentalAnalysisofBehavior,37:23-34.SidmanM,WynneCK,MacguireRW,BarnesT(1989)Functionalclassesandequivalencerelations.JournaloftheExperimentalAnalysisofBehavior,52:261-274.SiemannM(1998)Equivalenciesbyreversalsinpigeons(unpublishedexperiment).SiemannM,DeliusJD(1998a)Inductionofstimulusassociationsbyreinforcementconcordancesinpigeons.In:Gottingenneurobiologyreport1998(EisnerN,WehnerR,ed),pp447.Stuttgart:Thieme.SiemannM,DeliusJD(1998b)Algebraiclearningandneuralnetworkmodelsfortransitiveandnon-transitiverespondinginhumansandanimals.EuropeanJournalofCognitivePsychology,10:307-334.SlomanSA,RipsLJ,ed(1998).Similarityandsymbolsinhumanthinking.Cambridge:MITPress.Valero-AguayoL,Luciano-SorianoMC(1996)Relacionesdeequivalencia:Unestudiodereplicationdelefectodelarelationsimetricasobrelatransitiva.ApuntesdePsicologia,37:25-39.VaughanW(1988)Formationofequivalencesetsinpigeons.JournalofExperimentalPsychology:AnimalBehaviorProcesses,14:36-42.VaughanW,GreeneSL(1984)Pigeons"visualmemorycapacity.JournalofExperimentalPsychology:AnimalBehaviorProcesses,10:256-271.VeenmanCL(1997)Pigeonbasalganglia:Insightsintotheneuroanatomyunderlyingtelencephalicsen•sorimotorprocessesinbirds.EuropeanJournalofMorphology,35:220-233.WatanabeS,LeaSEG,DittrichWH(1993)Whatcanwelearnfromexperimentsonpigeonconceptdis•crimination?In:Vision,brainandbehaviorinbirds(ZeiglerHP,BishofH-J,ed),pp351-376.Cambridge:MITPress. 122JuanD.Deliusetal.WatanabeS,SakamotoJ,WakitaM(1995)Pigeons"discriminationofpaintingsbyMonetandPicasso.JournaloftheExperimentalAnalysisofBehavior,63:165-174.WrightAA,CookRG,RiveraJJ,SandsSF,DeliusJD(1988)Conceptlearningbypigeons:Matching-to-samplewithtrial-uniquevideopicturestimuli.AnimalLearningandBehavior,16:436-444.XiaL,DeliusJD,SiemannM(1996)Amultistimulusintelligenceplatformforpigeonconditioning.BehaviorResearchMethods,Instruments,andComputers,28:49-54.YagerRR,ZadehLA,ed(1994).Fuzzysets,neuralnetworksandsoftcomputing.NewYork:VanNos-trandReinhold.YamamotoJ,AsanoT(1995)Stimulusequivalenceinachimpanzee.PsychologicalRecord,45:3-21.ZentallTR,SteirnJN,SherbourneLM,UrciuoliPJ(1991)Commoncodinginpigeonsassessedthroughpartialversustotalreversalsofmanytooneconditionalandsimplediscriminations.JournalofExperi•mentalPsychology:AnimalBehaviorProcesses,17:194-201.ZentallTR,UrciuoliPJ(1994)Emergentrelationsintheformationofstimulusclassesbypigeons.Psy•chologicalRecord,43:795-810. 7AbstractionandDiscriminationNicholasJ.MackintoshTheuseofwordsthenbeingtostandasoutwardmarksofourinternalideas,andthoseideasbeingtakenfromparticularthings,ifeveryparticularideathatwetakeinshouldhaveadistinctname,namesmustbeendless.Topreventthis,themindmakestheparticularideas,receivedfromparticularobjects,tobecomegeneral....Thisiscalledabstraction,wherebyideastakenfromparticularbeingsbecomegeneralrepresentativesofallofthesamekind;andtheirnamesgeneralnames,applicabletowhateverexistsconformabletosuchabstractideas....Thusthesamecolourbeingobservedinchalkorsnow,whichthemindyesterdayreceivedfrommilk,itconsidersthatappearancealone,makesitarepresentativeofallofthatkind;andhavinggivenitthenamewhiteness,itbythatsoundsignifiesthesamequalitywheresoevertobeimaginedormetwith;andthusuniversals,whetherideasorterms,aremade.—JohnLocke(1690;bkII,ch.11,9)Inthenextparagraphofhisessay,Lockefamouslywentontoassertthat"brutesabstractnot."Butitseemsclearenoughthat,inLocke"sdefinition,manyanimalsbehaveas;/theyabstractedgeneralideasfromparticularinstances(evenifonlyafewhavelearnedtoassociategeneralnameswiththem).ThusifIrewardedapigeonforpeckingatascreenwheneverawhiteobjectwasshownonitbutwithheldrewardwheneverablackobjectwasdisplayed,itisquitecertainthathavingbeentrainedwithavarietyofdifferentstimuli(milkbottles,piecesofchalk,afieldcoveredinsnow,awhitetriangle,versusalumpofcoal,araven,ablackcat,ablacktriangle),thepigeonwouldgeneralizeappropriatelytonovelwhiteandblackpictures.AlthoughLockewouldattributesuchgeneralizationtotheabstractionofgeneralideas,thereisasimplerexplanation:thepigeonhasdetectedaninvariantfeaturecommontoanumberofdifferentpictures,andassociatedthatfeaturewithreward.Inwhatfollows,Iarguethatthisabilitytodetectinvariantfeaturesinavariablesetofstimuliprovidesasufficientaccountbothofdiscriminationlearningandso-calledcategorizationlearning.Therepresentationalprocessinvolvesnothingmorethandecomposingcomplexstimuliintosetsofelementsorfeatures.Otherlevelsofrepresentationareindeedpossible,andavailabletoatleastsomeanimals:thebestevidenceforthiscomesfromthedemonstrationofresponsetorelationshipsbetweentwoormorestimuli.Thatthisisindeedadifferentlevelofrepresentationissuggestedbytheobservationthatnotallanimalsprovideevidenceofresponsetorelationships.DiscriminationandCategorizationWeknow,fromethologicalobservationandexperiment,thatananimal"sdiscrim•inativebehaviorisoftencontrolledbyonlyoneortwofeaturesofanotherwise 124NicholasJ.Mackintoshcomplexstimulus.Thenewlyhatchedherringgullchickthatpecksasenthusiasticallyataredknittingneedlewithawhitebandnearitstipasitdoesatalifelikemodelofaherringgull"shead,isshowingthatthemain(only?)featureoftheadultthatcontrolsthisbeggingresponseistheredspotatthetopofthebeakcontrastingwiththewhitehead.Thefactthatthechick"sbehaviorissotightlycontrolledbysuchasimplefea•tureofacomplexstimulusseemsalimitation(albeit,ofcourse,anormallyperfectlyviableone),ratherthanevidenceofabstractionorhumanintelligence.Similarly,thepigeon"sdiscriminativebehaviorandgeneralizationinmyhypotheticalexperimentiswellcapturedbyanelementary,single-layeredconnectionistnetwork,whichdecomposesobjectspresentedtoitintoconstituentfeaturesorelements.Ifthenet•workistrainedonthisdiscriminationitwilllearnit,andgeneralizetonovelstimuli,becausethesetofunitsactivatedwheneverwhitestimuliareshowntoitwillendupwithstrongerconnectionstotherewardunitthanwillthoseactivatedbyblackstimuli.ComparativepsychologistshaveoftengonealongwithLockeindescribingthediscriminativebehaviorandgeneralizationoftheirsubjectsasevidenceofpossessionofgeneralideasorconcepts.ThusFields(1932)trainedratstodiscriminateanequi•lateraltrianglefromacircle,andfoundgoodtransfer(orgeneralization)todifferentsizedtriangles,aswellastoisoscelesandright-angletriangles.Hedescribedhisresearchasastudyinconceptformation—"thedevelopmentoftheconceptoftri•angularitybythewhiterat."ItwaslefttoKarlLashley,withcharacteristicacumen,tonotethesimplerexplanation:Thefundamentalprocess,theidentificationofcommonpropertiesintwoormoreconstella•tionsofelements,seemstobealmostuniversalamonganimals,appearingwheneveradiffer•entialreactionisestablished.(Lashley,1938,p.163)Lashley"sargumentamountedtosayingthatthereisnodifferenceofprinciplebetweenthecasewhereananimalistrainedonadiscriminationbetweenasinglestimulusassociatedwithreward(S+)andanotherassociatedwithitsabsence(S-),andshowstransfertonewstimulithatarevariantsonthesetwo;andthecasewhereanimalsarerequiredtodiscriminatebetweentwoormorelargesetsofstimuli,sev•eraldozenpicturesoftrees,people,andpatchesofwaterservingastheS+set,andequalnumbersofotherpicturescontainingnotrees,orpeople,orwaterastheS-set,andagainshowgoodtransfertonewstimuli(Herrnstein,1984).Someexperi•menters,itistrue,havedescribedsuchresearchasthestudyofconceptlearningorcategoricalconceptformation(e.g.,Wassermanetal.,1988).ButHerrnsteinhimselfwasreasonablyclearthatcategorizationlearninginvolveslittlethatisnew: AbstractionandDiscrimination125Tocategorize,whichistodetectrecurrencesintheenvironmentdespitevariationsinlocalstimulusenergies,mustbesoenormousanevolutionaryadvantagethatitmaywellbeuniver•salamonglivingorganisms....Categorizationisjustobjectconstancy.(Herrnstein,1984,p.257)Itmaywellbedifficulttospecifythenatureofthecommonfeaturesorelementsthatpigeonsusetosolvesuchcategorizationproblems.Itissurelythecasethatnat•uralcategoriesareonlyrarelydennedbyasinglenecessaryorsufficientattribute,butmorecommonlybyconjunctionsordisjunctionsofcorrelatedfeatures(Herrnstein,1984).Butthesearenotsufficientreasonsforsupposingthatpigeonsaredoingany•thingmorecomplexthanassociatingalargenumberofpicturesand/orthefeaturestheycontainwithareward,andthenshowingtransfertonewpicturestotheextentthattheycontainfeaturespreviouslyassociatedwithareward.Severallinesofevi•dencestronglysuggestthatsuchananalysisissufficient.Forexample,pigeonsareremarkablyadeptatlearningvisualdiscriminationsbetweenseveralhundredpicturesarbitrarilydesignatedbytheexperimenterastheS+setandseveralhundredothersdesignatedasS-(vonFersenandDelius,1989).BecausethereisnocategoricalbasisforthedistinctionbetweentheS+andS-set,thereisnobasisforgeneralizationtonovelstimuli.Butgiventhisremarkableabilitytomemorizethewhollyarbitraryrewardassignmentsofhundredsofpictures,itseemsreasonabletosupposethat,iftrainedtoassociate40picturesoftreeswithareward,theywillhaveassociatedalargenumberof"treefeatures"withareward,andthatmostnovelpicturesoftreeswillbemorelikelytocontainoneormoreofthesefeaturesthanpicturescontainingnotrees.Itisalsothecasethatsomecatego•rizationsareverymucheasierforthepigeonthanothers,andsomeofthesediffer•encesseemexplicableintermsofthesalienceoftheiridentifyingfeatures.ThusRobertsandMazmanian(1988)foundthatpigeons,unlikepeople,learnedtodis•criminatepicturesofkingfishersfromotherbirdsmuchmorerapidlythanpicturesofbirdsfromotheranimals,orpicturesofanimalsfromnonanimals.Onlyinthefirstcasedidtheyinitiallyshowanyevidenceofsuccessfultransfertonewpictures.Acluetothereasonfortherelativeeaseofthekingfisher-otherbirddiscriminationisthatwhentheotherbirdswereselectedtobebrightlycolored,thediscriminationwassignificantlydisrupted.Moreoever,experimentsinwhichpigeonshavebeentrainedtodiscriminatebetweenmorecarefullyspecifiedsetsofstimuli(Brunswikfaces,HuberandLenz,1993;segmentsofcircles,Mackintosh,1995)suggestthatdiscriminativeperformanceandtransferareanorderlyfunctionofthenumberandvalueoffeaturesassociatedwithreward.Finally,thereislittledoubtthatthemostpowerfulandwidelyaccepted 126NicholasJ.Mackintoshaccountofthebehaviorofpeopletrainedtocategorizeartificialsetsofstimuliisasimpleassociativeone,whereexemplars,ortheirfeatures,areassociatedwithcate•gorymembership,andtransferisbasedongeneralizationfromstoredexemplarsorweightchangesinanelementaryassociativenetwork(NosofskyandKruschke,1992).Inspiteoftheirfrequentlyratherglibtalkofconceptualizationorconceptforma•tion,Isuspectthatfewofthosewhohaveexperimentedonpigeons"categoricaldis•criminationlearningwouldseriouslydisputethisanalysis.EvenWasserman,forexample,allowsthat"stimulussimilaritymighteffectivelymediatetransferfromtrainingtotestingstimuli"intheseexperiments(Wasserman,1997,p.127).More•over,thereisreasontobelievethatasimilaranalysisappliesequallywellnotonlywherepeopleareaskedtocategorizeartificiallyconstructedstimuli,butalsotothecaseofnonhumanprimatescategorizingmorenaturalstimuli(RobertsandMazmanian,1988).TomaselloandCall,whoarenotgenerallyslowtodrawasharpdistinctionbetweenthecognitivecapacitiesofprimatesandnonprimates,acknowl•edgethatTheabilitytomakesenseoftheperceptualworldbydiscriminatingfeaturesandcategorizingphenomenaisabasiccognitivecapacityofmanyanimalspecies.(TomaselloandCall,1997,p.133)SuperordinateConcepts?Animalscan,however,donotcategorizeonlyonthebasisofperceptualsimilarity.SuggestiveevidenceofcategorizationonthebasisofuseorfunctioncomesfromastudybySavage-Rumbaughetal.(1980),wheretwochimpanzees,ShermanandAustin,sortedvarioustypesoffoodandtoolsintothesetwocategories,andwereequallyadeptatsortingthelabels(lexigrams)associatedwithpreviouslyuncatego-rizeditems.Althoughtheymighthavebeenpartlyrelyingonthedifferenceinper•ceptualcharacteristicsbetweenthetwoclassesofobject—forexample,thematerialsofwhichfoodsandtoolsarecomposed,ratherthantheiruseorfunction(seeTom•aselloandCall,1997)—thereisnodoubtthatmanyanimals,includingpigeons,arecapableofcategorizingtogetherpicturesthatarenotinstancesofanynaturalcate•gory,andsurelyhavenofeaturesincommonthatdifferentiatethemfromasecondsetofpicturesthattheyplaceintoadifferentcategory.Forexample,AstleyandWasserman(1998)showedpigeonsfoursetsofpictures—cars,chairs,flowers,andpeople.Althoughthereisampleevidencethatpigeonsdiscriminateamongallfourclassesofstimuli(Wassermanetal.,1988),inthisstudytheywereinitiallyrequiredtosortthemintoonlytwoclasses,thatis,tomakeoneresponsetopicturesofcars AbstractionandDiscrimination127andflowers,andasecondresponsetopicturesofchairsandpeople.Theywerethentrainedtomaketwonewresponsestocarsandtochairs,andfinallytestedtoseeiftheywouldalsorespondinthisnewwaytoflowersandpeople.Theydid(albeitnotataveryhighlevelofaccuracy).Wasserman(1997)arguesthatthisprocedureestablishedtwosuperordinatecate•gories(oneofcarsandflowers,theotherofchairsandpeople)which,becausetheycouldnotbebasedonperceptualsimilarity,provideaclearcaseofconceptualization(seealsoLea,1984,wheremuchthesamehypotheticalexperimentisdescribedaspotentialproofofananimal"spossessionofaconcept).Butthemechanismsunder•lyingsuchcategorizationareasreadilyunderstoodinassociativeasinconceptualterms.ConsiderthefollowingexperimentbyVaughan(1988),wherethestimuliwere40pictures,quitearbitrarilydividedintotwosets,1and2,andthebirdswereorigi•nallyrewardedforrespondingtoSet1andnotforrespondingtoSet2.Aftertheyhadlearnedthisdiscrimination,therewardassignmentswerereversedandthen,everyfewsessions,repeatedlyreversed,foratotalofover100reversals.Atthebeginningofeachnewreversal,thebirdsshowedtheyrememberedthepreviousrewardassign•mentofthestimulibyrespondingbelowchance.After20ormorereversals,however,theylearnedeachnewreversalwithinasinglesession,andtowardtheendofthisfirstsessionwererespondingabovechancetopictureswhosenewlyreversedrewardassignmenttheyhadnotyetexperienced.Inotherwords,experienceofthereversedassignmentofsomeofthepicturesinSet1wassufficienttoenablethemtorespondabovechancetotheremainingpicturesinthatset.AllthatthepicturesinSet1sharedincommon,todistinguishthemfromthepic•turesinSet2,wastheirearliercommonhistoryofreinforcement.Butthiscommonhistorywassufficienttoestablish,inVaughan"sterminology,anequivalenceclass.Althoughtherehavebeenseveralfailurestofindevidenceofsuchequivalenceclassesinpigeoncategorizationexperiments(seeDelius"schapterinthisvolume),successdoesnotseemtoosurprising(e.g.,Zentalletal.,1993),andthemechanismunder•lyingsuchsuccessisreadilydocumentedinrathersimplerexperiments(Nakagawa,1992;Zentalletal.,1993).Nakagawa(1992)trainedratsconcurrentlyontwoquiteindependentvisualdiscriminationproblems,andthenrequiredthemtolearnthereversaleitherofbothdiscriminationsorofonlyone(withcontinuedretrainingontheother).If(butonlyif)theyhadreceivedextensiveovertrainingontheoriginaldiscriminations,theyfounditmucheasiertoreversebothratherthanonlyone.TheresultisverysimilartoVaughan"sinimplyingthatasufficientlylongcommonhis•toryofreinforcement(overtraininginNakagawa"sexperiments,repeatedreversalinVaughan"s)sharedbytwoormorestimuliestablishesafunctionalequivalencebe•tweenthem,suchthatthereinforcementofanewresponsetoonewillgeneralizeto 128NicholasJ.Mackintoshtheother(s).Themechanismunderlyingthiseffectisonelongfamiliartoassociativelearningtheory,namely,theprocessof"acquiredequivalence"orHull"smediatedgeneralization(seeHoneyandHall,1989).RelationalConceptsPerceptualsimilaritybasedoncommonfeatures,andmediatedgeneralizationbasedonacommonhistoryofassociation,maybesufficienttoaccountformanyinstancesoftransferofdiscriminationorcategorization.Idonotbelievethatineithercaseithelpstotalkofconceptualizationorabstraction.Arethereotherinstancesofdis•criminationorcategorizationthataremoreusefullycharacterizedintheseterms?Forsomeyearsnow,comparativepsychologists"mostpopularexampleofasup•posedlyabstractconcepthasbeenidentity,orsamenessordifference.Thiscertainlyraisesanewissue,forwearenowtalkingaboutarelationshipholdingbetweentwoormorestimuli,ratherthananyfeatureorattributeofasinglestimulusorstimulusarray.Doanimals(andifsowhichanimals)showevidenceofsensitivitytotherela•tionshipsholdingbetweentwoormorestimuli(andifso,whichrelationships)?AccordingtoKohler(1918),thephenomenonoftranspositionwasevidencethateventhehumblechickenrespondedtorelationshipssuchasbrighterordarker,largerorsmaller.Iftrainedtochoosethelargeroftwostimuli,andthengivenachoicebetweentheiroriginalS+andanevenlargerstimulus,thechickenschosethehithertounseenlargerstimulusinpreferencetotheirfamiliarS+.ButSpence"s(1937)classicanalysisdemonstrateshowtranspositioncanarisefromtheinteractionofgradientsofexcitationandinhibitionconditionedindependentlytoS+andS-.Givenanappropriatechoiceoftheshapesoftheseexcitatoryandinhibitorygradients,Spenceargues,itiseasytoshowthatthenetexcitatoryvalueofthelargeS+mightbelessthanthatoftheevenlargerstimulusinKohler"sexperiment.Spence"sanalysiswasvindicatedbythedemonstrationthattheempiricalpostdiscriminationgradientsobtainedfrompigeonstrainedonasuccessivewavelengthdiscriminationbetweenanS+of550nmandanS-of560nm,didnotpeakatS+,butratheratstimuliof530or540nm(Hanson,1959;thisistheso-called"peakshift").Morerecently,WillsandMackintosh(1998)providedempiricalevidenceforBlough"s(1975)re-interpretationofSpence"sinteractinggradientsintermsofcommonelements.Pigeons(andpeople)willshowapeakshiftalongawhollyartificial"dimension,"whereneighboringstimuliareconstructedofoverlappingsetsofarbitraryicons.Nevertheless,thereisevidencethattranspositionmaynotbeentirelyaccountedforintermsofSpence"sorBlough"snonrelationaltheories(seeGonzalezetal.,1954).Rileyandhiscolleagues(seeRiley,1968)haveshownthattheopportunityfor AbstractionandDiscrimination129simultaneouscomparisonoftwostimulidifferingin,say,brightness,bothmakesiteasierforratstolearnthediscriminationbetweenthem,andincreasestheincidenceoftranspositiononsubsequenttest.Theobviousinterpretationseemedtobethatsimultaneouscomparisonoftwostimulidifferinginbrightnessmadetherelationshipbetweenthem(oneisbrighter/darkerthantheother)moresalient.Buthere,atleast,asRileynotes,thereisanevenmoreobviousexplanationofsucheffectsintermsofafairlylow-levelsensoryprocess—thatofcontrast.Byandlarge,theperceivedbrightnessofanypartofthevisualfieldisadirectfunctionoftheamountoflightreflectedfromthatpartofthefield,andaninversefunctionoftheamountoflightreflectedfromneighboringpartsofthefield.IfS+andS-differinluminanceandareshownsidebyside,thepresenceoftheobjectivelylighterS+willdecreasetheperceivedbrightnessofS-,justasthepresenceoftheobjectivelydarkerS-willincreasetheperceivedbrightnessofS+.Contrastincreasestheperceiveddifferencebetweenthem,thusmakingdiscriminationbetweenthemeasiertolearnand,eveninSpence"sanalysis,couldincreasetheincidenceoftransposition.WillsandMackin•tosh(1999)showthatthismaybetheonlysenseinwhichpigeonsshowevidenceofrelationallearninginsuchexperiments.AsFetterman(1996)notes,whatmaybelooselydescribedasevidenceofsensitivitytoabstractrelationshipsmayoftenbemoreappropriatelyanalyzedasahard-wiredconsequenceofthestructureofpartic•ularsensorysystems.Severalotherattemptstofindevidenceofrelationallearninginpigeonshavebeenequallyunsuccessful.Pearce(1991)trainedpigeonstomakeoneresponsetotwoverticalbarsofthesameheightandanothertotwobarsdifferinginheight,andfoundnosignoflearninguntilthetwobarswereactuallytouchingoneanother—atwhichpointarathersimplesolutionbecomesavailable:makeoneresponsetoasin•glewidebar(withastraightedgeatthetop)andanothertoabarwithastepinitstopedge.BothAydin(1991)andWills(1996)failedtoreplicateinpigeonstheresultsofastudyofrelationallearninginratsbyLawrenceandDeRivera(1954),wheresubjectswererequiredtomakeoneresponsetoaverticalrectanglewhosetophalfwaslighterthanthelowerhalf,andanotherresponsewhenthetophalfwasdarkerthanthelower.SamenessandDifferenceIdonotknowwhethertherelationshipofsamenessordifferenceisanymoreabstractthantherelationshipofdarker/brighter.Itiscertainlynoeasiertoestablishthenatureoftheprocessesinvolvedinananimal"ssolutionofasame-differentdiscrimination. 130NicholasJ.MackintoshAcommonprocedureforstudyingsame-differentdiscriminationlearningistotrainanimalsonmatchingtosampleorodditydiscriminations,wheretheymustchoosebetweentwoalternativestimuli,AandB,oneofwhichisthesameas,theotherdifferentfrom,thesamplestimulus(Aonsometrials,Bonothers)shownatthebeginningofeachtrial.Inanalternativeprocedure,onlytwostimuliareshownoneachtrial.Oneresponseisrequiredwhenthetwostimuliarethesame(AAorBB)andasecondresponseisrequiredwhenthetwoaredifferent(AB).Animalsrangingfrompigeons,crows,parrots,rats,andracoons,tomonkeysandapeshaveallsolvedsuchproblems(seeRoitblattandFersen,1992;TomaselloandCall,1997,forreviews).Thequestionishowtheydoit.Thesolutionofasingleproblem,itseemsobviousenough,couldbebasedonspecificconditionalorconfigurallearning,with•outanyneedtoappealtotheanimal"sdetectionoftherelationshipbetweenthestimuli:performoneresponsetotheAAandBBconfigurations,anothertotheABandBAconfigurations;orlearnthatifthesampleisA,chooseAnotB,ifitisB,chooseBnotA.Justasincategorizationexperiments,elucidationofwhatanimalshavelearneddependsontheirperformanceontransfertests.Thereisnowampleevidencethatsomebirds(crows,rooks,jays,Wilsonetal.,1985;anAfricangreyparrot,Pepperberg,1987),aswellasmonkeysandapes(seeTomaselloandCall,1997),showexcellenttransfertowhollynovelstimuliaftertrainingonmatchingorodditydiscriminationswithnomorethanoneortwopairsofstimuli.Isthissufficienttoestablishthattheylearnedtousetherelationship(sameordifferent)betweensampleandchoicestimuliasthebasisfortheirsolution?Itseemsareasonableinterpretation,butthereisanalternative,firstnotedbyPremack(1983).Inthetypicalmatchingoroddityproblem,atrialstartswiththepresentationofthesamplealone,towhichtheanimalisrequiredtorespondbeforethechoicestimuliarepresented.ItfollowsthatifthesamplewasA,thechoicestimulus,A,willberelativelymorefamiliar(recentlyencountered)thanthealternativeB.Perhapstheproblemissolvedasarelativenovelty-familiaritydiscrimination.Suchasolutionwould,ofcourse,transferperfectlywelltonewstimuli,CandD,andsoon.Novelty-familiarityseemsaparticularlyplausiblebasisfortheresultsofanexperimentbyYoungetal.(1997).Onsometrialsthebirdswererequiredtopeckateachof16differentpicturesoriconsshownoneaftertheother,andthentomakea"different"response.Onothers,thesameicon(whichcouldbeanyoneofthe16)waspresented16timesinsuccession,withthebirdsrequiredtopeckateachpresentation,beforefinallymakinga"same"response.Thebirdslearnedthediscriminationandtransferredmoderatelywelltonewicons.Anovelty-familiaritysolutionseemsratherhardertoimagineforthetwo-stimulussame-differentdiscrimination(AAorBBversusABorBA),andcorvids,parrots, AbstractionandDiscrimination131andprimateshaveallsolvedandtransferredsuchdiscriminations.Itismostclearlyruledoutbytwovariantsofsuchproblems.Pepperberg(1987)trainedherparrot,Alex,withobjectsvaryingincolor,material,andshape.Histaskwastospecifytheattribute(s)thatdifferedbetweentwootherwisesimilarobjects—forexampleared,woodentriangleversusagreen,woodentriangle—oranattributethatwasthesameintwootherwisedissimilarobjects—forexample,ared,woodentriangleversusared,leathercircle.Hisperformanceontransferteststonovelobjectswasimpressivelyaccurate.Equallyconvincingistheevidenceofsecond-orderrelationallearninginsomeprimates.Premack"schimpanzee,Sarah,learnedandtransferredtonovelstimulithefollowingmatchingproblem:ifthesamplewastwoidenticalobjects(AA),thecorrectchoicestimuluswastheonewithtwootheridenticalobjects(XX)ratherthantheonewithtwodifferentobjects(YZ).Ifthesamplecontainedtwodissimilarobjects(BC),thenYZwascorrectratherthanXX.Premackfoundthatnotallchimpanzeeswouldtransferthissolutiontonovelstimuli,andsuggestedthat"lan•guage"training,ofthekindexperiencedbySarah,wasnecessarytoenablethemtodoso.Subsequentstudieshaveshownthatthisisclearlynottrue(e.g.,Thompsonetal.,1997).Thefactremainsthatonlychimpanzeeshavesolvedandtransferredthisproblem,andnotallchimpanzeeshavedoneso.Nevertheless,asGillanetal.(1981)show,itisonlyasmallstepfromthistogeneralizedanalogicalreasoning,whereSarahhasshownthatshecandetectwhetherawiderangeofrelationships(notjustsameordifferent)betweentwoobjects,AandA",isorisnotthesameastherela•tionshipbetweenBandB".CanPigeonsLearnaSame-DifferentDiscrimination?Ratherthananalyzefurtherthebasisforsuchrelationallearningbyprimatesandsomebirds,manycomparativepsychologistshavedevotedmoreenergytotheratherlessinterestingquestion—whetherpigeonsarecapableofrelationallearningatall.Inspiteofconfidentearlierclaimstothecontrary,itnowseemsreasonablyclear(e.g.,Wilson,etal.,1985;Wrightetal.,1988)thataftertrainingonasinglematchingorodditydiscriminationwithonlyonepairofstimuli(AandB),pigeonsshownoreli•ableevidenceoftransfertonovelstimuli.Butisthatsufficienttoprovethattheyareincapableoftransferringasame-differentrule?Notatall.AccordingtoDelius(1994),suchexperimentshaveprovedlittleornothing:"mostoftheevidencedenyingtheidentity-oddityconcepttopigeonsmustbedismissedinretrospectasbeingduetotheemploymentofpatentlyinadequatemethods"(p.37).Thesepatentlyinadequatemethods,itshouldbestressed,havebeenquitesufficienttoyielddecisiveevidenceoftransferinotherbirdsandprimates,andhaveshowngoodtransferincorvidsandsquirrelmonkeys,butnoneinpigeons,whenthevarious 132NicholasJ.Mackintoshspecieshavebeentrainedunderessentiallythesameconditions(Wilsonetal.,1985;D"Amatoetal.,1986).Theminimalconclusion,therefore,mustbethatitishardertoobtainevidenceofgeneralizedsame-differentdiscriminationinpigeonsthaninmanyotheranimals.Giventhepaucityofotherevidenceforrelationallearninginpigeons,thisshouldnotcomeasasurprise.Butthepigeon"sobstinacyisnomatchforsomecomparativepsychologists"faithinitsintellectualprowess.Twotypesofexperimenthaveapparentlyyieldedevidenceofsuccessfultransferofmatchingoddityorsame-differentdiscriminationsinpigeons.Thefirst(e.g.,Lombardietal.,1984;Wrightetal.,1988)hassimplygivenevenmoreextensivetrainingonanevenlargernumberofdifferentpairsofstimuli.Thesecond(e.g.,Cooketal.,1995;Wassermanetal.,1995)greatlyincreasesthenumberofstimuliinthearraysbetweenwhichbirdsareaskedtodiscriminate.Inthestandardsame-differentdiscrimination,animalslearntomakeoneresponsetoAAorBBandanothertoABorBA.InWassermanetal."sexperiment,thestimulusarrayscon•sistedof16iconsina4x4grid.IfthesearelabelledAtoP,therewere16"same"stimuliconsistingof16repetitionsofA,orofB,andsoon.The"different"stimuliweresimilar4x4arrays,butwitheachofthe16iconsappearingonce.Bytheendoftraining,pigeonswereabout85percentcorrect,andwhentestedwithanovelsetoficonsrespondedcorrectlyon70percentoftrials.Theproblemwiththesenewerproceduresisthat,atthesametimeastheyprovidebetterevidenceoftransfer,theyalsointroducethepossibilityofnewbasesforsuchtransfer.Wherebirdsaretrainedonaverylargenumberofdifferentmatchingorodditydiscriminations,itbecomesdifficulttoruleoutthepossibilitythattransferwillbebasedonthephysicalsimilaritybetweensupposedlynovelteststimuliandsomeofthestimuliusedintraining.ThetypeofstimuliusedbyCookandWassermanandtheircolleaguessuggestsadifferentpossibility.Anobviousdifferencebetweena4x4arrayofAsanda4x4arrayconsistingofthe16lettersAtoPeachoccurringonce,isthattheformerhasaregulartextureandthelatterdoesnot.ThiswouldapplytothetypeofstimuliusedbyCook,wherethesamedisplaysconsistedoftheregularrepe•titionofaparticulartypeofitem(letussay,As)andthedifferentdisplaysconsistedofalargenumberofAs,plusasingleodditem,aB,orasmallblockofodditems.BothCookandWassermanargueagainstanysuchinterpretationoftheirdataontheratherstrangegroundsthattheirbirdsdidnotshowperfecttransfertonovelstimulusarrays,andthereforemusthavelearnedsomethingabouttheactualphysicalprop•ertiesoftheirtrainingstimuli.Sotheymusthave,butitdoesnotseemdifficulttoacceptthattheycouldhavelearnedbothsomethingaboutthetexturaldifferencebetweensameanddifferentstimuli,andsomethingaboutthespecificindividualitemstheycontained. AbstractionandDiscrimination133Youngetal.(1997)attemptedtoruleoutsuchaninterpretationbyspreadingthe16iconsintheirsameanddifferentarraysoverarandomsubsetof25pos•siblelocationsina5x5grid,reasoningthatthe"decidedlydisorderly"and"notablyuntidy"arraysthusgeneratedwouldmeanthatthe"same"stimulinolongerhadamoreregulartexturethanthe"different"stimuli.Theystillobservedgoodtransfer(about75percentcorrect)toarraysofnovelicons.Butthefactremainsthatalthoughlessregularthanthe4x4"same"arraysusedbyWassermanetal.(1995),the"same"arraysinthisexperimentmustalwayshavecontainedclumps,orlongrowsorcolumns,ofregulartexture.Theobvioustestofthisanalysisistoreducetheabsolutenumberoficonsinthesameanddifferentdisplays:anarrayof16or12identicalicons,evenifspreadovera5x5grid,willinevitablyproduceareasofuniformtexture,butanarrayofonlytwoorfouriconsmaynot.Andindeed,whenYoungandWassermantrainedpigeonsconcurrentlyonsame-differentdis•criminationswitheither2,4,8,12,or16iconsineachdisplay,theirbirdsshowednosignofevenlearning(letalonetransferring)thesolutiontothetwo-orfour-icondisplays.Itisworthrecallingthatatwo-itemarray(AAversusAB)isthestan•dardsame-differentdiscriminationsolvedandtransferredbyrooks,parrots,andprimates.Enough,forthemoment,ofpigeons.Forwhatitisworth,myownreadingoftheevidenceisthatnoconvincingcasehasyetbeenmadethatpigeonsrespondtotherelationshipbetweentwoormorestimuli.Otherswill,nodoubt,disputethatcon•clusion.Theymayberight.Buttoprolongthediscussionfurtherwouldbetoinvestitwithmoreimportancethanitdeserves.Foronething,allIamsuggestingisthattheevidencepresentlyavailabledoesnotprovideaparticularlyconvincingcaseforrela•tionallearninginpigeons(asitdoes,say,forchimpanzees).Itwouldbemerelyfoolishtoassertthatpigeonsareincapableofrelationallearning.Myargumenthasbeendesignedtoestablishasinglepoint.Itispossibletodistin•guishamongthreeratherdistinctwaysinwhichanimalsrepresentasetofstimuli.First,theycanrepresentthemasarraysorconfigurationsoffeaturesorelementsdennedintermsoftheirownabsolutevalues(allowingforknownmechanismsofsensoryinteraction).Second,theycanrepresentthemintermsoftherelationshold•ingbetweentwoormorearrays—oneisbrighterordarkerthananother,twoothersarethesame,andsoon.Finally,theycancomparetherelationsholdingbetweentwoormorearrayswiththoseholdingbetweenothers:therelationshipbetweenAandA"canbecomparedtothatbetweenBandB".Atafirstapproximation,pigeons(andnodoubtmanyotheranimals)trainedonasinglediscriminationproblem(whetherdescribableasarelationaldiscriminationornot)provideanexampleofthefirsttypeofrepresentation;someotherbirdsandmostprimatesanexampleofthesecond;and 134NicholasJ.Mackintoshchimpanzeesgivenappropriatepriorexperience(althoughwedonotknowwhatisorisnotappropriate)anexampleofthethird.PeopleItis,ofcourse,temptingtodescribethesedifferentmodesofrepresentationasmoreorlessabstract,complex,advanced,orintelligent.Inpartasawayofresistingthattemptation,Ishallconcludebypresentingevidenceforallthreelevelsofrepresenta•tioninpeople.Nooneseriouslydoubtsthatadulthumansunderstandthenotionofbothfirst-orderandsecond-orderrelationshipsbetweenevents.Itis,perhaps,hardertoacceptthatelementaryassociativeanalysesalsoapplytothebehaviorofpeople.Butthesuccessofsuchanalysesintherealmofcategorizationlearningshouldnotbeforgotten.Inmanycategorizationexperiments,however,thereisoftennoabstractorrule-governedstructureinthestimulussettoallowanalternativemodeofsolution,andwherethereis,peoplemayindeedshowsensitivitytoit(e.g.,RegehrandBrooks,1993).Thusafairlynaturalpreconceptionofmanycognitivepsychologistshasbeenthatwhereabstractstructureorrelationshipsbetweenstimuliareavailabletoguidechoice,peoplewillnormallyusesuchinformationtoreducetheburdenonmemoryimposedbylearningthesolutiontoaproblembyrote(i.e.,bysimpleassociativelearning).Thepreconceptionisoftenjustified.Whenpeopleandpigeonsaretrainedoneitherasimplediscriminationbetweentwostimulidifferinginbrightness,oronacategorizationtaskbetweenaseriesofwedge-shapedstimuli,wherelong,thinwedgesbelongtoonecategoryandshortfatwedgestotheother,theybehavedifferentlytoteststimulithatfallbeyondtherangeofvaluestheysawduringtraining(Aitken,1996;seealsoMackintosh,1997;WillsandMackintosh,1998).Thepigeonsshowapeakshift:thatis,theyrespondmorerapidlytostimulibeyondS+(ortheS+set)inthedirectionawayfromS-;butastheteststimulimoveyetfurtherawayfromS+,theystopresponding.Thisisthebehaviorpredictedbyinteractinggradientsofgen•eralizationorbyanelementaryassociativeanalysis.Butpeopleshowtruerelationaltransposition:theycontinuetotreatstimuliasbelongingtotheS+categoryevenwhentheyarefarremovedfromS+.Itseemsthattheyhavelearnedthatshortfatwedgesbelongtoonecategoryandlongthinonestotheother;theshorterandfatter(orthelongerandthinner)thewedgeseenonatesttrial,themoresecurelyitcanbeassignedtoitscorrectcategory.Theyhavelabeled,andrespondedto,therelationshipbetweenthestimuliencounteredintraining.Butpeopledobehaveexactlylikepigeons,andshownothingbutapeakshift,iftrainedundercertaincircumstances.InWillsandMackintosh"sstudy,whenthe AbstractionandDiscrimination135stimulus"dimension"wascreatedbyconstructingstimuliofoverlappingsetsofarbitraryicons,bothpeopleandpigeonsdisplayedsimilarpeakshifts,respondingmorepositivelytoanovelstimuluslyingclosertoS+ontheartificialdimensionthantoS+itself,butlesspositivelytoastimulusevenfurtherremovedfromS+.Aitkenwasabletodemonstrateapeakshiftinpeopletrainedonthecategoriza•tiontaskbetweenlong,thinandshort,fatwedges.Herethesecretwastomoveinthedirectionofanimplicitlearningtaskbypresentingthewedge-shapedstimuliembeddedinagreatdealofother,irrelevantinformation,schedulingtrialsinrapidsuccessionsothatsubjectshadnotimetostopandanalyzethesituation,andturningtheirtaskintoareactiontimetask,wheretheyhadtorespondasrapidlyandaccu•ratelyaspossibletoatargetstimulusappearingontheleftorrightofthescreen.Thelocationinwhichthetargetwouldappearwas,infact,sometimespredictedbytheshapeofthewedgethat,alongwithotherirrelevantstimuli,appearedatthebegin•ningofsometrials.Fewsubjectseverrealizedthis—althoughtheirreactiontimeswerefasterontrialswhenapredictivewedgeappearedthanontrialswhenonedidnot.Inotherwords,theycouldbedescribedaslearningimplicitly.Andthen,whentestedwitharangeoflong,thinandshort,fatwedges,theyshowedaclassicpeakshift.Itakethislastexperimentasevidencethat,evenwherearelationshipbetweentwoormorestimuliisavailabletocontrolpeople"sbehavior,theymaynotalwaysdetectthatrelationship.Butthepreconceptionthatpeoplemustdetectanduserelationshipsbetweenevents,orabstractstructure,orrulestoguidetheirdecisionmaking,issostrongthatitisoftenacceptedbycognitivepsychologistsastheexplanationoftheirsubjects"behavior,evenwhenthosesubjectsdisclaimallknowledgeoftherelation•ships,rules,orstructure.Theirabstractionoftherulesorstructureisthensaidtobeunconsciousorimplicit.Thusafterexposuretostringsoflettersgeneratedbyanartificialgrammar,subjectstypicallyperformwellabovechancewhenaskedtocate•gorizeasgrammaticalorungrammaticalnewletterstringstheyhavenotseenbefore(e.g.,Reber,1989).Eventhoughsubjectsaretypicallyunabletoarticulatetherulesunderlyingtheproblemstheyhaveapparentlysolved,Rebercreditsthemwithimplicitknowledgeofthoserules.Butwhyisasimpleassociativeanalysisnotjustascapableofexplainingcatego•rizationlearninghereasitisinothercases?Theexposurephaseoftheexperimentteachessubjectsthatcertainletterstringsare"grammatical,"andtheirtaskonthetestistocategorizenewletterstringsasgrammaticalornongrammatical,onthebasisoftheirsimilaritytothisinitialset(PerruchetandPacteau,1990).Here,asinthecaseofanimalexperiments,ithasseemedthatonewayofdistinguishingbetweenasimpleassociativeandarule-basedaccount,wouldbetotestsubjectsnotonlywithnovel 136NicholasJ.Mackintoshstringsconstructedfromthesamesetoflettersusedintraining,butalsowithstringsconstructedaccordingtothesamerulesbutfromanentirelydifferentsetofletters.Theanswersuggestedbyawholeseriesofsuchexperimentsisthatpeopledoshowsometransferwhentestedwithentirelynewsetsofletters,buttheirperformanceissignificantlyworsethanwhentheyaretestedonnovelstringsconstructedfromthefamiliarset(seeWhittleseaandDorken,1993).Dependingonone"spointofview,itispossibletopointtoabove-chanceperformanceasclearevidenceofsomerulelearning,ortothesignificantdecrementasevidenceofratherlittlerulelearning.ArathermorefruitfulapproachisillustratedinoneofWhittleseaandDorken"sexperiments.Here,theyexposedsubjectstoletterstringsunderthreedifferentcon•ditions:inthestandardcondition,theywereaskedtomemorizethestrings;inasec•ondcondition,theyweretoldthattheletterstringswereadistractorfromtheirprimarytask;inathird,theywereaskedtocountthenumberofrepetitionsofaparticularitemineachstring.Thethreetypesofinstructionhadonlyasmalleffectontestperformancetonovelstringsconstructedfromtheoriginalletterset.Butsig•nificantdifferencesappearedwhensubjectsweretestedonanewsetofletters.Thestandard,memorizingconditionproducedthestandardresult:performancesig•nificantlyabovechance,butsignificantlyworsethanonthefamiliarsetofletters.Whentheoriginaltaskhadbeenpresentedasadistractor,subjectswereatchanceonthetestwiththenovelset.Butwhenaskedtocountrepetitions,theyperformedaswellonthenovelsetasonthefamiliarset.WhittleseaandDorken"sreasonableinterpretationwasthatunderdistractorconditionssubjectsencodedonlythesurfacefeaturesoftheletterstringswithoutnotingtherelationships(repetitions)betweenitems;intherepetitioncondition,theydidencodetheserelationships,whichbecamemoreimportantthanthesurfacefeatures;andinthestandardconditiontheydidabitofboth.Itisthussurelymisleadingtoimplythatparticipantsinartificialgrammarexperi•mentslearntheabstractrulesthatgeneratelegalstrings,whetherunconsciouslyorconsciously,letalonethattheydosoautomatically.Whattheylearndepends,asever,ontheirreadingofthetasksetthem.Variationsinwhattheylearnunderdif•ferentinstructionsarebetterdescribedasvariationsinhowtheyencodethestrings.WhenWhittleseaandDorken"ssubjectswereaskedtonoterepetitions,theydidnotsuddenlystartlearningtherulesoftheartificialgrammar;theylearnedabouttherelationsamongitems.Whatpsychologistsoftendescribeastheabstractionofrules(whetherbypeopleorotheranimals)canveryoftenbeunderstoodaslearningabouttherelationshipsbetweenevents.Inthepresentcase,ofcourse,appropriateperfor•manceonatransfertesttonovelsetsoflettersdependsonmorethannotingtherelationshipbetweentwoitemsinastring,itrequiresrespondingtorelationshipsbetweenrelationships.SeeingthatthestringMQVVVAMisinoneimportantsense AbstractionandDiscrimination137thesameasthestringXBEEEHXrequiresnotingthepositionofrepeatedandnon-repeateditemsineachstring,aswellasthenumberofrepetitions.Thereisthusampleevidencethat,underdifferentcircumstances,peoplewillrespondsimplytothesurfacefeatures(absoluteproperties)ofitemsinanarray,totherelationholdingbetweentwoormoreitemsinanarray,andtothesecond-orderrelationholdingbetweentherelationsbetweenitemsindifferentarrays.Theabilitytodetectandrespondtosuchrelationships,Ishouldargue,oftenprovidesabetteraccountofproblemsolvingandtransferthandoesappealtothenotionofabstractrules.TheEvolutionofCognitionWhatimplicationsdoestheanalysissketchedhereholdforthequestionoftheevo•lutionofcognition?Onepervasivethemecommontoseveralchaptersinthisbookhasbeentoseehowfara"generalprocess"approachtothestudyoflearningandcognitionremainsvalid,orwhethercognitionmustbecharacterizedasmodular.Theanswergiventothisquestiondependspartlyonthetypeofbehaviorstudied,andpartlyontherangeofprocessescountedascognitive.Cognitionismodularifitisdennedtoincludespecializedsystemsfor,say,deadreckoningorcircadiantiming;butevenShettleworthacknowledgesthatassociativelearningiswidespreadandhascertaingeneralpropertiesthatallowanimalstolearnaboutthecausalrelationshipsamongaverywidevarietyofevents.TheresearchIhavediscussedhasbeencon•cernedwithassociativelearning,andtothatextentitisaboutsomerathergeneralprocesses.Butmyargumenthasattemptedtoshowthatthepowerofassociativelearningtosolvedifferentkindsofproblemsvariesdramaticallywiththenatureofthestimulusrepresentationsthatserveasinputtotheassociativesystem.Thatisaconclusionthatshouldbeentirelycongenialtothemodulartheoristwhoseesper•ceptualsystemsasthearchetypalmodules.AsThorndike(1911)arguedsomeyearsago,thebasiclawsofassociativelearningarecommontoanimalsrangingfromcrabtomonkeytohumanbaby(althoughwenowknowthattheselawsareratherdifferentfromthoseThorndikeenvisaged).Differencesbetweencrabsandmonkeysweretobefoundinwhatstimulitheyasso•ciatedwithwhatconsequences—inmyterminology,inwhatlevelsofrepresentationareavailabletothem.Thisanalysisimpliesbothcontinuityanddiscontinuity.Thediscontinuityisbetweentherepresentationofstimulisimplyintermsoftheirindi•vidualphysicalcharacteristicsandtherepresentationoftherelationshipsbetweenstimuli.Theperceptionofrelationshipspermitsthegeneralizationofthesolutionlearnedtooneproblemtootherproblemsthatsharenophysicalfeaturesincommonwiththefirst.Intheend,itpermitsanalogicalreasoning. 138NicholasJ.MackintoshAccordingtoTomasello(seechapter9),"itistheunderstandingofrelationalcat•egoriesingeneralthatisthemajorskilldifferentiatingthecognitionofprimatesfromthatofothermammals."Buthegoesontoarguethatsuchrelationalunderstandinghasitsoriginsintheneedtounderstandthird-partysocialrelationships.WhileTomaselloandIagreeonthekeyimportanceofrelationalunderstanding,Iremainscepticalofhisclaimthatonlyprimatesarecapableofsuchunderstanding(theevidencefromcorvidsandparrotscitedaboveseemsreasonablyconvincingtome),andalsoofhisargumentthattheoriginsofsuchunderstandingaresocial.Notallcorvids,forexample,areparticularlysocial,andinourownworkwefoundnoevi•denceofdifferencesbetweenthosespeciesthatweresocialandthosethatwerenot(Mackintosh,1988).Myanalysis,howeverisalsoanargumentformentalcontinuity,andnotonlyinthesenseenvisagedbyThorndike.Thereisalsoevidenceofcontinuityinthebasiclevelofstimulusrepresentation.Althoughpeopleareobviouslycapableofperceivingandunderstandingrelationshipsamongeventsandrelationshipsbetweenrelation•ships,experimentalparadigmscanbedevisedthatrevealtheoperationofthesimplerrepresentationallevel.Cambridgestudents,whonormallyshowrelationaltransposi•tion,willbehavelikepigeonsandshowonlyapeakshiftiftrainedincertainwayswithcertainkindsofstimulus.Whenstimuliarepresentedasincidentaldistractorstothesubject"smaintask,peoplewillnotdetecttherelationshipsbetweenthemthattheywouldotherwiseusetorelievetheburdenofrotememorization.AlthoughCharlesDarwin(1871)proposedthatcomparativepsychologyshouldseekevidenceofmentalcontinuitybetweenhumansandotheranimals,Iarguethatheplacedtoomuchemphasisonfindingevidenceofprecursorsto,orelementsof,humanconceptualthought,reasoning,orevenlanguageinthebehaviorofotheranimals.Butthereisequal,perhapsevengreater,valueinfindingevidenceoftheoperationofbasic,generalprocessesinpeople.That,ofcourse,wasoneofthegoalsofanearliergenerationofbehaviorists.Butitisnonetheworseforthat.Ibelievethatitwillprovejustasfruitful,andcommandmorerespect,thenthestrategythatseeks,ratherbreathlesslyandwithmoreenthusiasmthancaution,tofindevidenceofabstractconceptsinpigeons.Cognitivepsychologistsfinditrathereasytopourwitheringscornonthatparticularendeavor(ChaterandHeyes,1994).ReferencesAitkenMRF(1996)Peakshiftinpigeonsandhumancategorisation.Ph.D.thesis,UniversityofCambridge.AstleySL,WassermanEA(1998)Noveltyandfunctionalequivalenceinsuperordinatecategorizationbypigeons.AnimalLearningandBehavior,26:125-138. AbstractionandDiscrimination139AydinA(1991)Relationlearninginpigeons.Master"sthesis,UniversityofWales.BloughDS(1975)Steadystatedataandaquantitativemodelofoperantgeneralizationanddiscrimination.JournalofExperimentalPsychology:AnimalBehaviorProcesses,1:3-21.ChaterN,HeyesC(1994)Animalconcepts:Contentanddiscontent.MindandLanguage,9:209-246.CookRG,CavotoKK,CavotoBR(1995)Same-differenttexturediscriminationandconceptlearningbypigeons.JournalofExperimentalPsychology:AnimalBehaviorProcesses,21:253-260.D"AmatoMR,SalmonDP,LoukasE,TomieA(1986)Processingofidentityandconditionalrelationsinmonkeys(Cebusapella)andpigeons(Columbalivia).AnimalLearningandBehavior,14:365-373.DarwinC(1871)Thedescentofmanandselectioninrelationtosex.London:JohnMurray.DeliusJD(1994)Comparativecognitionofidentity.In:Internationalperspectivesonpsychologicalscience:Leadingthemes(BertelsonP,EelenP,d"YdewalleG,ed),pp25-40.Hillsdale,NJ,Hove,UK:Erlbaum.FettermanJG(1996)Dimensionsofstimuluscomplexity.JournalofExperimentalPsychology:AnimalBehaviorProcesses,22:3-18.FieldsPE(1932)StudiesinconceptformationI:Thedevelopmentoftheconceptoftriangularitybythewhiterat.ComparativePsychologyMonographs,9:1-70.GillanDJ,PremackD,WoodruffG(1981)Reasoninginthechimpanzee:I.Analogicalreasoning.JournalofExperimentalPsychology:AnimalBehaviorProcesses,7:1-17.GonzalezRC,GentryGV,BittermanME(1954)Relationaldiscriminationofintermediatesizeinthechimpanzee.JournalofComparativeandPhysiologicalPsychology,47:385-388.HansonHM(1959)Effectofdiscriminationtrainingonstimulusgeneralization.JournalofExperimentalPsychology,58:321-334.HerrnsteinRJ(1984)Objects,categories,anddiscriminativestimuli.In:Animalcognition(RoitblatHT,BeverTG,TerraceHS,ed),pp233-262.Hillsdale,NJ:Erlbaum.HoneyRC,HallG(1989)Acquiredequivalenceanddistinctivenessofcues.JournalofExperimentalPsy•chology:AnimalBehaviorProcesses,15:338-346.HuberL,LenzR(1993)Atestofthelinearfeaturemodelofpolymorphousconceptdiscriminationwithpigeons.QuarterlyJournalofExperimentalPsychology,46B:1-18.KohlerW(1918)NachweiseinfacherStructurfunktionenbeimSchimpansenundbeimHaushuhn.AbhandlungenderKoniglichPreussischenAkademiederWissenschaften,Phys.Math.Klasse,2:1-100.Translatedandcondensedas"Simplestructuralfunctionsinthechimpanzeeandinthechicken."In:Asourcebookofgestaltpsychology(EllisWD,ed)pp.217-227.London:RoutledgeandKeganPaul.LashleyKS(1938)Themechanismofvision.XV.Preliminarystudiesoftherat"scapacityfordetailvision.JournalofGeneralPsychology,18:123-193.LawrenceDH,DeRiveraJ(1954)Evidenceforrelationaltransposition.JournalofComparativeandPhysiologicalPsychology,47:465-471.LeaSEG(1984)Inwhatsensedopigeonslearnconcepts?In:AnimalCognition(RoitblatHL,BeverT,TerraceHS,ed),pp263-277.Hillsdale,NJ:Erlbaum.LockeJ(1690)Anessayconcerninghumanunderstanding.1sted.(abridgedandeditedbyA.S.Pringle-Pattison)1924.Oxford:OxfordUniversityPress.LombardiCM,FachinelliCC,DeliusJD(1984)Oddityofvisualpatternsconceptualizedbypigeons.AnimalLearningandBehavior,12:2-6.MackintoshNJ(1988)Approachestothestudyofanimalintelligence.BritishJournalofPsychology,79:509-526.MackintoshNJ(1995)Categorizationbypeopleandpigeons.QuarterlyJournalofExperimentalPsy•chology,48B:193-214.MackintoshNJ(1997)Hasthewheelturnedfullcircle?Fiftyyearsoflearningtheory,1946-1996.Quar•terlyJournalofExperimentalPsychology,50A:879-898. 140NicholasJ.MackintoshNakagawaE(1992)Effectsofovertrainingonreversallearningbyratsinconcurrentandsinglediscrim•inations.QuarterlyJournalofExperimentalPsychology,44B:37-56.NosofskyR,KruschkeJK(1992)Investigationsofanexemplar-basedconnectionistmodelofcategorylearning.In:Thepsychologyoflearningandmotivation(MedinDL,ed),pp207-250.SanDiego:Aca•demicPress.PearceJM(1991)Theacquisitionofabstractandconcretecategoriesbypigeons.In:Currenttopicsinanimallearning:Brain,emotion,andcognition(DachowskiL,FlahertyC,ed),pp141-164.Hillsdale,NJ:Erlbaum.PepperbergIM(1987)Acquisitionofthesame/differentconceptbyanAfricanGreyparrot(Psittacuseri-thacus):Learningwithrespecttocategoriesofcolor,shape,andmaterial.AnimalLearningandBehavior,15:423-432.PerruchetP,PacteauC(1990)Syntheticgrammarlearning:Implicitruleabstractionorexplicitfragmen•taryknowledge.JournalofExperimentalPsychology:General,119:264-275.PremackD(1983)Thecodesofbeastandman.BehavioralandBrainSciences,6:125-167.ReberAS(1989)Implicitlearningandtacitknowledge.JournalofExperimentalPsychology:General,118:219-235.RegehrG,BrooksLR(1993)Perceptualmanifestationsofananalyticstructure:Thepriorityofholisticindividuation.JournalofExperimentalPsychology:General,122:92-114.RileyDA(1968)Discriminationlearning.Boston,MA:AllynandBacon.RobertsWA,MazmanianDS(1988)Conceptlearningatdifferentlevelsofabstractionbypigeons,mon•keys,andpeople.JournalofExperimentalPsychology:AnimalBehaviorProcesses,14:247-260.RoitblattHL,vonFersenL(1992)Comparativecognition:Representationandprocessesinlearningandmemory.AnnualReviewofPsychology,43:671-710.Savage-RumbaughES,RumbaughDM,SmithST,LawsonJ(1980)Reference—thelinguisticessential.Science,210:922-925.SpenceKW(1937)Thedifferentialresponsetostimulivaryingwithinasingledimension.PsychologicalReview,44:430-444.ThompsonRKR,OdenDL,BoysenST(1997)Language-naivechimpanzees(Pantroglodytes)judgerelationsbetweenrelationsinaconceptualmatching-to-sampletask.JournalofExperimentalPsychology:AnimalBehaviorProcesses,23:31-43.ThorndikeEL(1911)AnimalIntelligence:ExperimentalStudies.NewYork:Macmillan.TomaselloM,CallJ(1997)Primatecognition.Oxford:OxfordUniversityPress.VaughanW(1988)Formationofequivalencesetsinpigeons.JournalofExperimentalPsychology:AnimalBehaviorProcesses,14:36-42.vonFersenL,DeliusJD(1989)Long-termretentionofmanyvisualpatternsbypigeons.Ethology,82:141-155.WassermanEA(1997)Thescienceofanimalcognition:Past,present,andfuture.JournalofExperimentalPsychology:AnimalBehaviorProcesses,23:123-135.WassermanEA,HugartJA,Kirkpatrick-StegerK(1995)Pigeonsshowsame-differentconceptualizationaftertrainingwithcomplexvisualstimuli.JournalofExperimentalPsychology:AnimalBehaviorPro•cesses,21:248-252.WassermanEA,KiedingerRE,BhattRS(1988)Conceptualbehaviorinpigeons:Categories,sub•categories,andpseudocategories.JournalofExperimentalPsychology:AnimalBehaviorProcesses,14:235-246.WhittleseaBWA,DorkenMD(1993)Incidentally,thingsingeneralareparticularlydetermined:Anepi•sodicprocessingaccountofimplicitlearning.JournalofExperimentalPsychology:General,122:227-248.WillsSJ(1996)RelationalLearninginPigeons?Ph.D.thesis,UniversityofCambridge. AbstractionandDiscrimination141WillsSJ,MackintoshNJ(1998)PeakShiftonanartificialdimension.QuarterlyJournalofExperimentalPsychology,51B:1-31.WillsSJ,MackintoshNJ(1999)Relationallearninginpigeons?QuarterlyJournalofExperimentalPsy•chology,52B:31-52.WilsonB,MackintoshNJ,BoakesRA(1985)Transferofrelationalrulesinmatchingandodditylearningbypigeonsandcorvids.QuarterlyJournalofExperimentalPsychology,37B:313-332.WrightAA,CookRG,RiveraJJ,SandsSF,DeliusJD(1988)Conceptlearningbypigeons:Matching-to-samplewithtrial-uniquevideopicturestimuli.AnimalLearningandBehavior,16:436-444.YoungME,WassermanEA,DalrympleRM(1997)Memory-basedsame-differentconceptualizationbypigeons.PsychonomicBulletinandReview,4:552-558.ZentallTR,HoganDE(1974)Pigeonscanlearnidentityordifference,orboth.Science,191:408-409.ZentallTR,SherburneLM,SteirnJN(1993)Commoncodingandstimulusclassformationinpigeons.In:Animalcognition:AtributetoDonaldARiley(ZentallTR,ed),pp217-236.Hillsdale,NJ:Erlbaum. 8PrimateWorldsKimSterelnyPrimateMindReadersDononhumanprimatesunderstandanythingaboutother"sminds?Whatwouldshowsuchunderstanding?Whatisthesimplestmindcapableofrepresentingotherminds?When,why,andhowdidprimatesbecome"mindreaders,"awarenotjustofthelikelyfuturebehaviorsoftheirsocialpartners,butofthementalcausesofthosebehaviorsaswell?Thispaperseekstoanswersomeofthesequestionsinthecontextofthe"socialintelligencehypothesis."Thiscurrentlypopulartheorysuggeststhatcognitiveevolutionintheprimatelineagehasbeendrivenbyselectionforsocialskills.Atsomepointintime,primatesocietybecamecomplex.Thisincreaseincomplexitydroveselectionforincreasedindividualintelligence,andwiththatincreaseinintelligence,primatesocietybecameyetmorecomplex.Primatefitnessbecameincreasinglydependentontheskillsofsocialnavigation.Theresultwasafeedbackloopforeverhigherintelligenceadaptedtosolvingtheproblemsofacom•plexsociallife.Inturn,selectionforsocialintelligencebecameselectionformindreadingcapacities,fortheactionsofothersaretrackedbestthoughrepresentingthementalstatesthatgeneratethoseactions.1Inthejargonoftheliterature,suchani•malsareknownas"mindreaders,"incontrasttolowlybehaviorreaders,capableonlyofrepresentingothers"actualandpotentialbehaviors.Mindreadersdonotjustrepresentbutmetarepresent.Perhapsthroughtheinfluenceofthesocialintelligencehypothesis,thereisawidespreadexpectationthatthetransitionfrombehaviorreadingtomindreadingtookplacesomewhereintheevolutionofthegreatapeclade.Themainweightofexperimentalandfieldstudieshasbeentoprobeforevidenceofmindreadinginthisgroup.Theseinvestigationswillbemyfocus,too,butinconsideringthemitisimportantnottoslideintoseeingmindreadingasasurrogateforcognitivesophisti•cationingeneral.Weshouldresistthetemptationtosupposethatanysignsofcog•nitivesophisticationsignaltheshifttomindreading.Behaviorreadersneednotberestrictedtosimplereinforcementlearning.ThusDickinsonandBalleine(thisvol•ume)arguethatratsunderstandthecausalrelationsbetweenactsandoutcomes.Imitation,too,iscognitivelysophisticated,butdespiteclaimstothecontrary(see,forexample,Tomasello,thisvolume)itdoesnotshowmind-readingcapacities—orsoIargue(Sterelny,1998).Itispossibletobe(relatively)smartandbeunabletometa•represent.Itisaheterodoxclaimofthispaperthatitisalsopossibletometarepresentdespitebeing(relatively)simple. 144KimSterelnyInconsideringtheshiftfrombehaviorreadingtomindreading,Ishalldependonanideaabouttheadaptiveadvantagesofrepresentation.Soit"simportanttodevelopanaccountofthepresumedbenefitsofmetarepresentationalcapacities.Themostobviousisanenhancedcapacitytoanticipateothers"actionsinbothcooperativeandcompetitiveinteractions.Intheirfamouspaperontheevolutionofcommunication,KrebsandDawkinsassumethatthisisthecriticaladvantageofmindreading(KrebsandDawkins,1984).It"softenthoughtthatmindreadingisusefultoanticipatebehaviorinnovelcircumstances.Forexample,Tomasello(thisvolume)argues:thiskindofcognitionenablesorganismstosolveproblemsinespeciallycreative,flexible,andforesightfulways.Thus,inmanycasesintentional/causalunderstandingenablesanindividualtopredictandcontroleventsevenwhentheirusualantecedentisnotpresent—ifthereissomeothereventthatmayservetoinstigatethemediatingforce.Forexample,anindividualmightcreateanovelwaytodistractacompetitor,(p.173)Inthispicture,behaviorreaderslearnspecificenvironment/responserules.IfyoumeetFredatthebananafeeder,staywellclearforotherwisehewillbiteyou.Theserules,theideagoes,willleavethebehaviorreaderwithnobasisforpredictinganagent"sbehaviorinnovelcircumstances.WhatwillFredbelikeatthemangofeeder?Onthisview,asprimateenvironmentsbecomemoreunpredictable,ametarepresent-ingprimatehasskillsofsocialnavigationthatabehaviorreadingprimatelacks.Oneempiricalchallengeistoprovideindependentevidenceofincreasingenvironmentalheterogeneity.Thatisnoteasy,forheterogeneitydependsinpartonhowanimalscategorizetheirworld.Ifababooncharacterizesitsworldinconcrete,sensoryterms,itwilloftenfinditselfinaseeminglynovelenvironment.Ifitisequippedwithmoreabstractcategories,itwillmuchlessoftenbeinnovelenvironments.Thereareotherproblemsaswell.Theadvantageaccruesonlytoverysophisticatedmindreaders.ThusTomasello"sexampleturnsonnoveltyleavinganagent"sgoalsunchangedbutchangingthepotentialmeanstothosegoals.Newenvironmentswilloftenhavethatimpactonanagent"sbeliefandpreferencestructure.Butnotalways:noveltycanchangeanagent"spreferences.Soinpredictinghowanotheragentwillactinanovelenvironment,themindreadinganimalwillneedtoknowwhetherthenoveltycausesare-orderingofpreferencesorjustofinstrumentalbeliefs.Thisisveryfancymindreading.Thereisasecondpossibility.Readersmaybenefitbyanabilitytouseothersasinstrumentsthattellthemabouttheworld.Theyexploitmind-worldandbehavior-worldrelationstofindoutabouttheworldasitisnowratherthanagents"futurebehaviors.Dennettcoinedtheterm"aninformationgradient"todescribegroupsinwhichindividualsvaryagooddealinwhattheyknow(Dennett,1983).Aninfor- PrimateWorlds145mationgradientselectsforthecapacitytouseothersassourcesofinformationabouttheworld.Theideaofusingothersasinformationsourceshasbeendiscussedmostlyinthecontextofimitation,butuseofothersneednotbeascognitivelysophisticatedasimitation.Otherscanbesourcesofinformationaboutwhatintheenvironmentisimportant.Readingothers"motivationsisaplausibletakeoffpointfortheuseofothersasinformationstores.Hencethephenomenoninsociallearningknownasstimulusenhancement:aheightenedattentiontoothers"interests.Iftherehasbeenselectionformindreading,mindreadersmustbehavedifferentlyfrombehaviorreaders,andthefitnessbenefitsofmindreadingmustbeinthebehaviorofmindreaders.Moreover,abehavioraldifferencethatisonlyverysubtlydifferent,oronethatisveryrarelymanifested,wouldprobablynotproduceenoughbenefitstopayitsway.Mindreadingcapacitiesareprobablyquitecostly,atleastiftheyrequireneocorticalexpansion(seeDunbar,thisvolume),sothesebehavioraldifferencesshouldbestriking.Therehasbeenagooddealofobservationalandexperimentalworkintheprimatologicalcommunitytryingto(a)isolatethebehav•ioralsignatureofmindreading,and(b)verifyorfalsifytheexistenceofthatsignatureinnonhumanprimates.Theremaybenobehavioralmagicbulletthatestablishestheexistenceofmindreading.AsWhitenandDennettemphasize,thedifferencebetweenbehaviorreadingandmindreadingmaybeinaoverallpatternofcompetenceratherthaninaspecificskill,accessibletomindreadersbutnotbehaviorreaders.Wecannotassumethereareunambiguoussignsofmindreading.Inthenextsection,Isketchageneralaccountofrepresentationanditsfunction,andapplyittoaviewofprimatemetarepresentation.Inthefollowingsections,Iexploitthatgeneraltheorywithtwocasestudies,beforeconcludingwithageneralmoral.Thefirstcaseconcernsarelativelybasicfeatureofanother"smind:itsfocusofvisualattention.Thesecondconcernsamoresophisticatedandlessbehaviorallyovertfeature:anotheragent"sknowledge.RepresentingMentalStatesOurcapacitytounderstandandpredictthebehaviorofothersisusuallytakentodependonagripof"folkpsychology,"animplicittheorythatlinksanagent"sbeliefs,preferences,emotions,andtheliketohisbehaviors,thoughthisviewisnowmorecontroversialthanitoncewas(Sterelny,1997).Thisaccountofourabilitytounderstandotheragentshasgeneratedatendencytoidentifythequestion"Donon-humanprimatesrepresentthementalstatesofotheranimals?"with"Dononhumanprimateshaveatheoryofmind?"Onthislineofthought,mindreadershaveagrip 146KimSterelnyonsomethinglikefolkpsychology.Inherrecentscepticalreviewof"theprimatetheoryofmind"literature,Heyesembracesthissuggestion.Shewrites:ananimalwithatheoryofmindbelievesthatmentalstatesplayacausalroleingeneratingbehaviour,andinfersthepresenceofmentalstatesinothersbyobservationoftheirappearanceandbehaviourundervariouscircumstances.(Heyes,1998;p.102)Tothosewhointerpretthedebatethisway,ananimalpossessesconceptsofmentalstatesonlyifthatanimalhasappropriateexpectationsabouttheconnectionsamongbeliefs,preferences,andbehavior.Thisisthenaturalviewtotakeifconceptposses•sionistiedtoinferentialconnectionsbetweenconcepts.Thisisknownasthe"infer•entialrole"theoryofmeaning.Defendersofthisviewholdthatthemeaningofaconceptderivesinpartorwholefromthewebofinferentialconnectionsamongthebeliefsinwhichthatconceptfigures.Thus,forexample,whatmakesmytiger-conceptatiger-conceptaretheinferentialconnectionsfromtiger-beliefstobeliefsaboutanimals,predators,prey,largestripedcats,andsoon.Mytigerconceptisdennedbymyimplicittheoryoftigers—mytigerlore—asrevealedbythiswebofinference.Forthosewhoacceptaninferentialroletoatheoryofmeaning,itisverynaturaltoidentifythepossessionoftheconceptsofbelief,preference,andthelikewiththeacquisitionofsomethinglikefolkpsychology,asetofbeliefsabouttheconnectionsamongbeliefs,preferences,intentions,andbehavior.Inthisviewofconcepts,tohaveaconceptofbeliefistohavemasteredabelieftheory.However,thisviewofconceptpossessionisnotmandatory.Therearealternativeviewsinwhichconceptsareidentifiednotbytheirconnectionsintheinternalecon•omyofthemindbutbytheirrelationstotheexternalworld.Solet"sbrieflybackofftothemoregeneralproblem:whatisitforananimaltorepresentsomething?Ihavearguedthatanorganismrepresentsafeatureofitsworld,asdistinctfrommerelyrespondingtoit,ifitcantrackthatenvironmentalfeatureviamorethanonekindofproximalstimulus.AnanimalthatrepresentsXisananimalwithseveralindepen•dentchannelsofinformationabouttheX-ishfeaturesofitsenvironment.Itmulti-tracksX(Sterelny,1995,1999).Consideracontrastingcase.Arthropodsoftenhavebeautifullyingeniouswaysofdetectingrelevantfeaturesoftheirenvironment,buttheyareoftendependentonasingleproximalcue.Thusthehygienicbehaviourofantsandbees—theirdisposalofdeadnestmates—dependsonasinglecue,theoleicaciddecayproduces.Theyhavenothingequivalenttoperceptualconstancymecha•nisms,mechanismsthatwouldenablethemtotrackthelivelinessoftheirnestmatesindifferentways.Communicationamongantsintheirneststypicallydependsonveryspecificchemicalsignals. PrimateWorlds147Assuccessfulcoordinationinanestorahiveshows,cue-boundbehaviorcanbeveryefficient.Butcontrolsystemsbasedonspecificcuesarefragileinimportantways.Anorganismthatcantrackitsenvironmentonlythroughasingle,specificcueisverylimitedinitsabilitytousefeedbacktocontrolandmodulateitsbehavior,foritreliesonchangeinthatsinglecue.Moreover,organismswhosebehavioriscue-boundareunlikelytohavecapacitiesthatarerobustoverarangeofdifferentenvi•ronments,forenvironmentalshiftswilloftendisrupttheircues.Thecapacitytotrackfunctionallyrelevantfeaturesoftheenvironmentinmorethanonewayisrequiredforbehavioralcapacitiestoberobust.Indistinguishingbetweencue-boundorganismsandthosethattrackagivenfea•tureoftheirworldthroughmultipleclues,weneedtodistinguishbetweentheuseofmultiplecuesandvariationin,andgeneralizationof,asinglestimulus.Thisisstraightforwardwhereorganismsmultitracktheirworldthroughdistinctsensemodalities.Butthezebrathattracksthedegreeofthreatposedbyahyenafrombothitspostureanditsgazedirectionisusingtwocues,notone,eventhoughusingvisionforboth.Thedistinctionbetweentheuseoftwocuesandstimulusgeneralizationfromasinglecueisprobablyhardtodefineprecisely.Afterall,notwohyenaapproacheswillprojectontotheeyesofazebraexactlythesameretinalstimulus.Butmanyparticularcasesareclear.Forinstance,considerthemuchdiscussedphe•nomenonofmirrorself-recognition,theabilityofchimpsandperhapsothergreatapestorecognizetheirownimageinamirror.Heyesrightlyarguesthatmirrorself-recognitiondoesnotdemonstratepossessionofaconceptoftheself(Heyes,1994,1995).Shepointsoutthatanimalsthatfindtheirwaythroughphysicallyclutteredenvironmentshavetoadjusttheirbehaviortothepositionoftheirbodyinspace,sotheymusthavea"bodyconcept"ofsomekind.Informationabouttheirbodyisusedtocontroltheirbehavior.Mirrorself-recognitionisjustalessusualexampleofthesamephenomenon,andnomoreshowsself-consciousnessthandoesthebehaviorofanimblebullinachinashop.Butalthoughself-recognitiondoesnotdemonstrateself-consciousness,itdoesshowanabilitytotrackbodilyfeaturesusingunusualperceptualinputs.Soanimalscapableofmirrorself-recognitiondonothaveacue-boundbodyconcept.Theirbodyconceptisarealrepresentationoftheirbody,fortheycanuseunusualinformationchannelstoupdateit.Solet"sdeploythisrepresentation/detectiondistinctiontogetafixonrepresenta•tionalcapacitiesofprimates.Primatesingeneraladapttheirbehaviortothepsy•chologicalstatesofotherprimates.Chimps,forexample,responddifferentiallytochimpsthataremotivatedtoattackthem.Thatis,theyoftenrecognizetheclues,thethreatsthatsignaltheimminenceofattack.Theirownbehaviorisadaptedto 148KimSterelnybehaviorscausedbyadistinctivepsychologicalstate.Theyrecognizeandrespondtosetsofbehaviorsthatarecuesto,becauseconsequencesof,particularpsychologicalstates.Sointhisveryminimalsenseprimatestrackthepsychologicalstatesofotherprimatesjustasantstrackthepropertyofbeinganestmate.Theyrespond,oftenappropriately,tothethreatofattackthroughaflowofinformationaboutthemoti•vationalstatesofthepotentialaggressortothemindoftherespondingprimate.Further,theyactaccordingtobehaviorrulesthatspecifytheirappropriateaction,giventhatstateoftheotheragent.However,trackingmentalstatesisonething;representingratherthandetectingthem,another.Weneedtoknowhowbonobos,forinstance,categorizethebehaviorofotherbonobos.Sosuppose:1.onebonoboalwaysreadsactionsa,b,c,d,e...asactionsofthesametype;2.actionsa,b,c,d,eareinfactalwaysgeneratedbyadistinctivementalstateQ;let"ssupposeQisanger.Atthispoint,thebonoboisatleastdetectingthementalstatesofanother,foritsresponsecovarieswithanger.Itsbehaviorisadaptedtothatfeatureoftheenviron•ment.Ourbonoboisrepresentingratherthanmerelydetectingthementalstatesofothersif:3.a,b,c,d,edonothaveanysinglesimplesensorycueincommon.Thereisrecentworksuggestingthatchimpstrackvisualattentionbyasimplecue,"facevisible."Bonobos,let"ssuppose,donottrackangerlikethis.Bodyposture,facialexpression,andvocalizationscanallindependentlyfeedintothe"placatethreat"behaviorrule.Iftheangerbehaviorsthatthebonobocategorizestogethersharenosingledis•tinctivesensorycue,thebonoboisnotstimulusboundwithrespecttoanger,butcantrackitviaavarietyofitsmanifestations.Thisconclusionisstrengthenedasthebonobo"scapacitytotrackangerapproachescompleteness:4.Notonlyarea,b,c,d,eproducedbyanger.Thebonoborespondsinthesamewaytoallormostofthebehaviorsangertypicallyproduces.Puttingthistogether,then,aprimaterespondstothementalstateofanotherifitcantrack—thatis,itrespondsdistinctivelywithsomereliability—tosomesuiteofbehaviorsthatareactuallycausedbysomespecificmentalstate;angerorfear,forexample.If,asdeWaalsuggests,abonoboappeasesangrybehaviorbytradingsexforpeace(deWaal,1989)itistrackinganger.Weinvestigatewhetherthisisangerdetectionorangerrepresentationthroughexperimentallyprobingtherobustnessof PrimateWorlds149thetracking.Inparticular,itisdetectionratherthanrepresentationifthiscapacityiscue-bound.Atthesametime,wecanexperimentallyinvestigatethesophisticationofthistrackingbyprobingthebreadthoftheresponsestoanger.Doestheanger-readeradapttoangrybehaviordifferentlyinenvironmentsthatcausethatbehaviortobeexpresseddifferently?Doesitresponddifferentlyifthephysicalorsocialenvironmentisdifferentinimportantways?Doesitrespondtoangerdifferentlydependingonitsrecognitionofothermentalstatesoftheanimal?Oristhe"angerbehavior"rulesimple:run!Sothispicturedefinestwoseparateexperimentalinvestigations.Weinvestigateaprimate"scapacitytorepresentamentalstatebyinvestigatingtherobustnessofitsabilitytotrackthatmentalstate.Robustnessisthevarietyofobser•vationalcuesitusesintracking.Wecanalsoinvestigatethebreadthofitsresponsetotracking.Breadthistheextenttowhichthetracker"sexpectationsabout,andresponsesto,theagent"sbehaviorareappropriatelymodifiedbywhatelsethetrackernotices.Wecanthenthinkofananimal"ssocialintelligencedevelopingviatwosortsofbehaviorrules.Recognitionruleslinkareadertotrackedmentalstates.Wesearchfortheanimal"srecognitionrulesbyfixingasfaraswecanthereader"senvironment,butvaryingbehavioralcuesofasingleunderlyingcognitivestate,toseewhetherthereadergivesthesameresponsetothesedifferentcues.Ihavebeenarguingthatananimalismindreadingonlyifithas,forsomementalstates,abatteryofrecognitionrules.Outputrulesgovernresponsestothestatesareadercantrack.Weprobeananimal"soutputrulesbyfixingthereader"scue,varyingtheenvironment,andtestingfordifferentresponses.Inaforthcomingpaper,Iarguethatthecapacitytorepresentevolvesfromthecapacitytodetectwhenorganismsliveininformationallytranslucentenvironments.Environmentsaretranslucenttoanorganismtotheextentthatecologicallyrelevantfeaturesoftheirenvironmentmapincomplex,one-manywaysontotheirtrans-ducibleworld.Iffood,shelter,predators,mates,friends,andfoemapincomplexwaysontodetectablephysicalsignals,cue-drivenorganisms"behaviorwilloftenmisfire.Ineedtofactortranslucenceintothedistinctionbetweenmultipleandcue-boundtrackingoftheenvironment.Inaveryminimalsense,ananimalcapableofformingaconditionedassociationiscapableofmultiplytrackingafeatureofitsenvironment.Pavlov"snotoriousdoglearnedtotrackthearrivaloffoodnotjustthroughthesightofitonitswaybutthroughthesoundofabellaswell.ButthisisnotaformofmultipletrackinginthesenseIhavebeendescribing,forthedog,asitsresponseshows,hasnoabilitytousetheflowofinformationdownonechanneltocheckthereliabilityoftheotheratthatsametime.Multipletracking,whichitadaptsorganismstothetranslucenceproblem,mustinvolvesomeformofintegrationor 150KimSterelnycross-channelchecking.Itisalsoimportanttodistinguishbetweentheuseofmultiplecuesandchangeinasinglecueovertime.Thefactthatananimalexhibitsplasticityovertimeinitsresponsivenesstoafeatureofitsworlddoesnotshowthatittracksthatfeaturemultiplyatatime.Sothereisadifferencebetween:(1)addingextratriggersofagivenaction;2(2)developingthroughlearninganewandmorecomplextriggerconsistingofagestaltwhoseelementsarenotindividuallysalient;and(3)beingabletouseanumberofindependent,butcross-checked,informationchannels.Only(3)adaptsananimaltotheproblemoftranslucence.Theimportanceofcross-talkbetweenchannelsisimplicitinWhiten"sdiscussionofdeceptionanditsunmasking,forexample.Deceptionmaybeunmaskedbyacom•binationofcuesandhencebyreaders"attemptstotrackmotivationthroughseveralcues(Whiten,1996).Forexample,theagent"sknownhistorymightcuethefactthatitsactualbehaviorwillbeatvariancewithitsadvertizedbehavior.Vervetscanlearntodiscountunreliableconspecific"salarmcalls.Aclearerexampleoftheadvantageofcross-talkdependsonthefactthatadvertisingmight"leak"—tell-talecuesmightundercutananimal"sadvertising.DeWaalhassomenice,thoughanecdotal,evidenceofchimpstryingtosuppressleakage;forexample,tryingtosuppresssignsofanxietyinconfrontations.Theenvironmentmight"leak"aswell.Thatis,theagentmightbeactinginawaythattheobserverknowsisinappropriatefortheactualenviron•ment.Forinstance,itmaybeactingasifitcanseesomethingtheagentknowsisnotthere.3Soinmyview,representationistiedtothisstrongsenseofmultitracking,andmultipletrackingisanadaptationtotheproblemposedbytranslucentenvironments.Manyversionsofthesocialintelligencehypothesisarebestseenasargumentsthatthesocialenvironmentsofprimates—orparticularprimates—areinformationallytranslucent.Thesehypothesesimplythatoftentherelationshipbetweenbehavioralcueandinnerstatewillbecomplex.Nosinglecueinisolationfromothersistobereliedon.Forexample,greatapesneedtotracksomeinnerstates,andtheycantrackthemonlybyrepresentingthem.Theyneedtoexploitamultiplicityofsignsofinnerstates,ratherthanasingleproximalsurrogate.AsPovinelliandCantpointoutintheirdiscussionoforangutans,thesheersizeofgreatapesisimportant,foritleadstophysicallylessstereotypedbehavior(PovinelliandCant,1995).Thatinitselfmeansthattheinnercausesofthosebehaviorsarehardertoreadoffbehavioralcues.Ifananimalmovesthroughitsenvironmentinahighlystereotypedway,aparticularposturalpositionmightwellbeanadequatesinglecueforflight/attack.Ifmotionisunstereotyped,andfunctionallydennedbehaviorpatternscross-classifybehaviorsdennedbymotorpatterns,behavioralcuestomotivationalstateswillbelessstraightforward. PrimateWorlds151Thesocialandcognitivecomplexityofprimatesexacerbatestheproblemsoftheobserver.Forexample,ifTomasello(thisvolume)isrightinthinkingonlyprimatestracktherelationsbetweenthirdpartiesandadjusttheirbehavioraccordingly,pri•matesocialbehaviorwillbelesspredictableonthebasisofsinglecuesthanthesocialbehaviorofotheranimals.Thegreaterthenumberoffactorsthatjointlydetermineanagent"sbehavior,themoreanobserverneedstonoticeifheorsheistopredictthatagent"saction.Shifting-balancesocialorganizations—organizationsthatareunstablemixesofcooperativeandcompetitiveinteractions—maysimilarlyexacer•batethepredictionproblem.Insomemonkeysocietiesrankisinheritedthroughthemother.Gibbonpopulationsmaybecomposedofhighlycooperativefamiliesbasedonamonogamouspair.Insuchsocieties,itisperhapsintheinterestsofallpartiestomakebehaviorreadingaseasyaspossible.Butthemoredynamicthemixtureofcooperationandcompetition,andthemoreimportantconcealmentanddecep•tionbecomes,thenthelesstransparenttheenvironmentbecomes.Inthisview,sheergroupsizewouldnotbeinitselfsignificant.Forthoughsizemightimposememorydemands,itisatransparent,nottranslucent,featureoftheenvironment.Themainpointisnotwhethertheseplausibilityconsiderationsarecorrect.Ratheritisthatthispictureofrepresentationleadstoamoreempiricallytractableaccountofthedistinctionbetweenbehaviorreadingandmindreading.MysuggestionhereconvergeswiththoseofWhiten(1996)andSober(1998).Theybothtreatmindreadingaspositingahiddenvariableconnectingananimal"scurrentandfuturebehavior,thoughtheirviewsarenotbasedonthisaccountofthedetection/representationdis•tinction.Theyarguethattheadvantageofroutingpredictionthroughahiddenvari•ableiscodingefficiency.Thereisnobehaviorthatonlyanagentemployingahiddenvariableanalysiscanpredict.Butasthepredictivedemandsonaagentrises,itbecomesevermoreefficienttopredictonthebasisofassignmentsofbeliefandmotivationtootheragents.Thisiswheretheirsuggestionsandminecometogether.Thecodingefficiencyderivesfromthefactthattherearedistinctovertsignsoftheinnermentalstate,andindistinctenvironmentsthatinnerstatewillgeneratedifferentbehaviors.Ifyoucantracktheinnerstate,youneednottrackseparatelyeachwaystimulusandbehaviorarelinked.ACaseStudy:VisualAttentionandImplicitKnowledgeofOtherMindsExperimentalworktestingforafullyfledgedgreatapetheoryofmindhasgeneratedresultsthatareatbestequivocal.Oneresponsehasbeentodevelopandtestlessambitiousideasaboutwhatapesknowaboutothers"minds.Animportantexample 152KimSterelnyofthishasbeentheinvestigationoftheextenttowhichchimpsunderstandvisualattention.Thisisimportantbecauserepresentingattentiondoesnotseemtobeaparticularlydemandingproblem—attentionhasanovertbehavioralsignature.Yetexperimentalresultssuggestthatchimps"griponvisualattentionissurprisinglylim•ited.Insofaraschimpsunderstandinghiding,stalking,andtheliketheymusthaveatleastacapacitytotrackthevisualattentionofothers.Theycanfollowthedirectionofgazetoitstarget(Whiten,1997p.164).ButPovinelliandEddyranaseriesofexperimentsthatseemtoshowboththattheircapacitytotrackvisualattentionmaybecue-bound,andtheirgripontheimpactofattentiononbehaviormightbeequallyimpoverished.Intheseexperiments,chimpswerefirsttrainedtousetheirnaturalbegginggesturetobegfoodfromatrainer,movingeithertotherightorleftsideofanenclosure(dependingonthetrainer"slocation),andbeggingthroughaholeinaclearwall.Theexperimentstestedforgeneralattentionbyhavingtestsinwhichonetrainerofferedfood,theotheravaluelessobject.Chimpshadnotroubleaskingforfood.Probetrialswerealternatedwithstandardtrialstoconfirmacontinuingmoti•vationtoaskforfood.Thereseemsnoreasontosupposethattheexperimentalsetupposedanyspecialproblemtothechimps.Intheprobetrials,thechimpswereofferedthechoicebetweenseekingfoodfromatrainerthatwasattendingtothem,andonesthatwerenot.Anumberofdifferent"attention-defeaters"weretested.Inattentivetrainersvariouslyhadbucketsovertheirhead,wereblindfolded,satfacingaway,orsatwiththeirhandsovertheireyes.Inthosecasesinwhichdistractionmighthavebeenaproblem,theattentivetrainershadmatched"distracters."Theysatwithbucketsontheirshoulder,withablindfoldaroundtheirmouth,andwithhandsovertheirears,respectively(PovinelliandEddy,1996).Thestrikingresultisthatthechimpsperformedatchanceonalltasksexceptthoseinvolvingthepairingoftrainersfacingawayorfacingforward.Eventhisseemedtodependonaverycrudecue.Resultsdroppedbackdowntochancewhentheexperi•mentwasredonewithbothsittingfacingaway,butonelookingbackovertheshoulder.PovinelliandEddydidtesttheideathatthereweretoomanydistracters,simplyoverwhelmingthechimps,thoughreallythisonlylooksplausiblewithbuck•etedandblindfoldedtrainers.Theyredesignedtheexperimentssothechimpswerefirstfamiliarizedwithtrainerswithoddaccoutrements,butthisredesigndidnotchangetheessentialresults.Thechimpsdideventuallylearntobegfromattendingtrainers,butthisseemstobebecausetheyextractedtheruleofattendingtothetrainerwhosefacewasvisible.Thustheyreturnedtochanceperformanceiftheyhadtodiscriminatebetweentrainerswitheyesopenandoneswiththeireyesshut.Theexperimentalupshotreadatfacevalueisthateitherchimpscandetectbutnotrep•resentvisualattention,ortheyhaveverylimitedmeansofusingtheirrepresentation PrimateWorlds153ofvisualattentiontotrackbehavior,orboth.Povinellihimselfseemstointerprethisexperimentsasshowingalimitedchimpgraspoftheattention-to-behaviorlink.Thatis,theyhaveimpoverishedoutputrules.Butit"sequallypossiblethattheirproblemlieswiththeappearance-to-attentionlink.Theyrelyoncrudesignalsofattention,ratherthanfailingtounderstandthebehavioralimportofattention.Inmyview,theseresultsareinput/outputambiguous.Inreadingabouttheexper•iment,Iwantedtoknowifthechimpsmadenoises,gestures,orotherattention-gainingactivities.Theserigiddirectionsofgazewouldcertainlybeveryanomalousinthechimps"naturallife.However,thatmightexplainwhychimpscansurvivebeingcue-boundwithrespecttovisualattention.Thecue"Iseeanagent"sface"doestrackattentionjustbecausedirectionofgazeisnotrigid.Soonerorlater,mostlysooner,youwillbespotted.Attentionmaybeatransparentfeatureoftheirworld,andhencenotonethattheyneedtotrackusingmanycues.Evenifthatisso,theymaybewellequippedtomanipulateandusevisualattention.Ithinkthereareexperimentalresultsthatlendsomesupporttotheideathatchimpstrackattentionfragilely,byasimplecue,buttheirresponsebreadthisnotimpoverished.Oncetheytrackattentiontheyknowwhattodowithit.Inparticular,theyknowhowtomaintain(andgain)it.Gomezreportsaseriesofexperimentsalongtheselines(Gomez,1996).Povinelli"schimpswereaskedtochoosebetweenattentiveandinattentivetrainers.Gomez"schimpswereaskedtomakethetrainersattentive,toattractattention.Gomez"strainersvariouslyhadtheirbacktochimp,hadtheireyesclosed,hadtheirheadpeeringintoacorneroftheenclosure,andhadtheireyeslookingoverthetopofthechimp.Whilefindingthe"eyesclosed"tasktough,Gomez"schimpsperformedmuchbetterthanPovinelli"s.Gomezhimselfinterpretstheseissuesviaadistinctionbetweenimplicitandexplicitknowledge.Theintuitiveideaisthatimplicitinformationismorecontextbound,andperhapsmorelikeknowing-howthanknowing-that,thanisexplicitinformation.Knowledgebecomesmoreexplicitthemoreitcanbeexportedfromveryparticularcontextsandactivities.Heinterpretshisworkasshowingthatthoughchimpsandgorillasmightnothaveexplicitknowledgeaboutotherprimate"smentallives,theydohaveimplicitknow-howaboutthem.Inparticular,theyhaveknow-howaboutbehaviorallyovertmentalstateslikeseeing.However,thedistinctionbetweenimplicitandexplicitrepresentationismoreaplaceholderforatheoryofrepresentationthanatheoryinitself.Theseissuesaremoreempiricallytractableifwedisaggregatetheimplicit/explicitdistinctionintotrackingrobustnessandresponsebreadth.Consider,forexample,responsebreadth.Idoubtwhetheranybehaviorisabsolutelyunconditional,forthemotivationalstateofthereaderwillnormallymatter.Buttheremaybesomethatareindependentofwhat 154KimSterelnyelsetheagentrepresentsabouthis/herenvironment.Sothebreadthofthereader"sresponsewilldependon(1)therangeofotherfeaturesoftheagentthereadertracks,and(2)theextenttowhichresponsedependsnotjustontheotheragentbuttherestoftheenvironment.Theeffectofthepresenceofsupportersofonepartyortheother;thephysicalgeographyoftheinteraction;thevalueofaresource(insituationsofconflict)andthelikeareallrelevanttooutputbreadth.(3)Astheexampleofvisualattentionshows,responsebreadthcanalsodependonthecapacitytomanageinter•action.Canareaderdirectattentioninawiderangeofsituationsandrathersubtly(assomeofthe"tacticaldeception"anecdotessuggest)oronlycrudely,bytouchingandscreaming?Totheextentthatananimal"sresponsedependsonvariationinits(nonsocial)environment,weneedtoconsiderthewaysanimal"scategorizetheirenvironment.Arethesecategoriesconcreteandsensory,oraretheysometimesfunctionalandab•stract(seeMacintosh,thisvolume;Bateson,thisvolume)?Oneimportantinstanceofthisdebateabouthowanimalsseetheirworldcentersoncausation,onthecausalconnectionbetweenevents.RumbaughandHillix(thisvolume)arguethatsomeprimatesunderstandsomecausalchains.Dickinsonandalliesarguethatevenratscangetit(DickinsonandShanks,1995;DickinsonandBalleine,thisvolume),whereasTomasello(thisvolume)arguesthattrueunderstandingofcausalrelationsisuniquetohumancognition.Thedistinctionbetweenrobusttrackingandresponsebreadthishelpfulintryingtodiscerntherealdisagreementshere.DickinsonandBalleinearguethatratsunder•standcausality.Rats,theyargue,areintentionalagentsbecausetheirbehaviorisgovernedthroughaninteractionoftheirpreferencesandtheirinstrumentalbeliefs.Theseinstrumentalbeliefsarecausalbeliefs.DickinsonandBalleinearguethatratsunderstandthecausalrelationbetweentheiractsandtheoutcomestheyexperience.Tomasello,incontrast,arguesthatnonhumanprimatesdonotunderstandthecausalrelationsbetweeneventsintheirenvironments.Unlessratsaresmarterthanmonkeys,itseemstheycannotbothberight.OnanalysisitturnsoutthatDickinsonandBalleineareconcernedwithrecognitionrules,whereasTomasellofocusesonresponsebreadth.DickinsonandBalleineareconcernedwithtracking,withwhatrelationshipintheirenvironmentratstrack.TheirexperimentalstrategyistopresentratswitharangeofenvironmentsinwhichtherelationshipbetweenAandBissometimescausal,some•timesprobabilistic,sometimesmerelycontiguous(soBfollowsAoften,butitfollowsnot-Aaswell).Theythenuseasimplebehavior(thewillingnessoftherattopressalever)totestwhetherratsaretrackingacausalrelationshipbetweenAandB,orwhethertheyaretrackingsomeotherrelation(perhapstemporalcontinuity)thatoverlapsthecausalone.Theyarenotprimarilyconcernedwiththecuesthrough PrimateWorlds155whichratstrackcausalconnection;stilllessaretheyconcernedwithshowingthatratscanusethisknowledgeflexibly,overarangeoftasks.Tomasello,bycontrast,isconcernedwithresponsebreadth.Heknowsthatprimateskeeptrackofrelationsamongdifferenteventsintheirworld,buthedoubtsthattheyunderstandcausalrelationsbecausetheycannotdomuchwiththeirknowledge.ThushecitestheexperimentsofVisalberghi,experimentsthatshowthatneithermonkeysnorapescanselectanappropriatetoolforasimplephysicalmanipulationwithoutextensivetrialanderrorlearning.Insum,thedistinctionbetweentrackingrobustnessandresponsebreadthisause•fulanalytictoolinprobingthedebatesaboutvisualattentionandmanyassociatedissues.Inparticular,ithelpsustodisaggregatethesedebates.Itisimportanttoseethatthereisnoapriorireasontoexpectthesefeaturesofmindreadingtobelinked.Detectioncanbeseveredfrombreadth:anagentmaytrackwillingness-to-playthroughasinglecue,theplayface,yetmighthavearichsetofbehaviorrulesabouttheconsequencesofthisstate.Equally,asfarasIcansee,thedifferentaspectsofbreadthmightvaryindependentlyofoneanother.Itwouldthenbeaninterestingempiricaldiscoverytoshowthattheywereinfactlinked.Equally,anevolutionaryordevelopmentalhypothesisthatpredictedlinkage,too,wouldbeimportant.Detectinga"TheoryofMind":WhattheExperimentsShowIhaveurgedthatthecapacitytorepresentothers"mindsisnottiedtothepossessionofatheoryoftheconnectionsamongbelief,preference,andbehavior.Ifurtherarguedthatthereareplausibilityconsiderationsthatsuggestthatprimatesdorepre•sent,andnotmerelydetect,others"mentalstates.Butthestrongerclaim—thatpri•mateshavesomethinglikeatheoryofotherminds—isinterestingandimportantinitsownright.Whatwouldshowthatananimalhadsomethinglikeatheoryofmind?BothWhiten(1996)andHeyes(1998)haveflirtedwiththeideathatrolereversal,per•spectivetaking,andsimilarexperimentsmightbeadiagnostictestforatheoryofmind.Inasimpleformofrolereversalexperiments,chimpsaretrainedinoneoftworoles.Theymusteitherindicateabinwithfoodtoanignoranthumanwhothensharesit,ortoacceptadvicefromaknowledgeablehuman.Povinellireportedthatthreeoffourchimpsshowedrolereversal:havingbeentrainedinonerole,theyper•formedintheother.Whitenagreesthatthereareproblemswiththeoriginalexperi•ment—perhapslearningthenewrolewasmerelyaccelerated—butheisinclinedtotheviewthatifchimpsgrasp"rolereversal,"itshowsmindreading.Thechimpdoes 156KimSterelnynotmerelycopythebehavioroftheagentwhoserolehasbeenoccupied.Rolereversalisnotjustdelayedimitation,atleastnotifimitationisjustmotor-patternimitation(Whiten,1996).Heyesisscepticalaboutallactualexperimentsallegedtoshowaprimatetheoryofmind.Butsheisinclinedtobeconcessiveonthe"roletaking"experimentsoftheoriginalPremackandWoodruffpaper(PremackandWoodruff,1978).Inthatpaper,PremackandWoodruffreportedachimp,Sarah,whochosephotographsthatrep•resentedsolutionstoaseriesofproblemsotheragentsfaced.Theideaisthattosolvethistask,Sarahhadtohavetheconceptofaplanandtoattributeplanstootheragents;shehadtoidentifytheintendedfinalgoalofaseriesofactions.HeyespointsoutthatSarah"ssuccessesarenotdecisive,foreachindividualcasemaybeexplainedaway.ButsheconcedesthereisnounitarykilljoyhypothesisexplainingawaySarah"sperformances.Idonotthinktheseexperiments,evenifredonetoavoidpossibleartifactsofthekindHeyesdiscusses,showanythinglikeafolkpsychology.Instead,IthinktheycanbeexplainedthroughByrne"sideaofabehavioralprogram.Indiscussinggorillaimitation,hearguesthattherewasevidencethatgreatapeshavetheabilitytolearnfromothersnotbymimickingeachchunkofmotorbehavior,butbyunderstandingabehavioralprogram.Theprogramistheoverallorganizationandsequencingoftheactsthatjointlycomposeaskill(Byrne,1995,1997).Gorillas,forexample,frequentlyeatthistlesandotherratherawkwardplants,sotheyoftenneedtodoagooddealofmanualprocessingoftheirfoodbeforetheycaneatit.Thisprocessingisquitecom•plex,andinvolvesadivisionoflaborbetweenthehandsthatchangesthroughdif•ferentstagesoftheprocessing.Ifsomeskillsdependonbehavioralprograms,imitationcaninvolvecopyingthatprogramratherthanaspecificmotorpattern.Thoughthereisplentyofanecdotalevidenceofgreatapeimitation,experimentalevidenceforimitationissurprisinglythin(Byrne,1995;Russon,1997).Butifgreatapesarecapableofimitation,itisimpressivebecauseitenablestheobservertoextracttheprogramfromthemotorbehavior.Heyes(personalcommunication)isverydubiousabouttheadequacyofthisre-analysis,becauseshedoubtsthatthenotionofabehavioralprogramhasbeenspecifiedinwaysthatmakeittestable.Inherview,Byrnehasnocriteriafordis•tinguishingimitationofabehavioralprogramfromaninaccurateorpartialimper•sonationofit.Hecannotdistinguishbetweenprogramlevelimitationandimitationofthedetailedsequenceofbehaviorthatisnoterror-free.Heyes"challengeiswelltaken,butatleastinprinciplethereisanempiricaldistinctionbetweenprogramlevelimitationandmereinaccurateimitation. PrimateWorlds157First,ifsociallearningconsistedoftheinaccurateimitationofafullbehavioralroutine,wewouldnotexpecterrorstohavethesamepatternacrossdifferentsubjects.Theyoughttobedistributedatrandom.Notsoifsociallearningisprogramlevelimitation.Differentsubjectsshouldbeinvariantwithrespecttothecomponentsofaskillidentifiedinthebehavioralprogram.If,say,gorillashavethecapacityforpro•gramlevellearningandshowthiscapacityin,say,learningfromamodelhowtoeatartichokes,wewouldexpect,onthesametask,differentsubjects—becausetheyseg•mentthetaskinthesameway—toresemblethemodelinthesamerespects.Notsoiftheerrorsarejustnoise.Second,imitationisexperimentallydistinguishedfromotherformsofsociallearningthroughanexperimentaldesigninwhichagivenoutcomecanbeachievedinmorethanoneway.Then,iftheaudienceaccomplishesthetaskusingthemodel"stechnique,wehavereasontothinkthattheyarelearningaboutmeans,notjustends,fromthemodel.Thisisknownasthe"twoaction"test.Ifwenestonetwo-actiontestinsideanother,wecandistinguishprogramlevelimitationfrommerelyinaccurateimitation.Weneedataskthat(1)permitssolutionthroughmorethanonebehavioralprogram;thatis,thetaskcanbedecomposedintosubtasksindifferentways.(2)Therearedifferentbutequallyadequatewaysofperformingthesubtasks.Eachsubtaskisatwo-actiontestitself.Forexample,Igatherthatchimpssometimesusetheirfeetratherthantheirhandsforatask.Sosupposethemimicshareswiththemodelasegmentationoftheoveralltask,butusesitsfeetratherthanitshands.Becausevariationsinthecomponentactivitieswouldnotbeerrorsormistakes,thatwouldbeprogramimitationratherthanerror-riddenimpersonation.Thereisnoerrorhere.Thechimphaslearnedfromthemodeladecompositionofthetaskintocomponents.Wehaveoverallsimilarityintheirchoiceofmeansbutnotsimilarityofdetailedphysicalmotions;thereisnoimitationatthesubtasklevel.Ihavenoideawhetherexperimentalevidencewouldsupporttheclaimthatthegreatapeshavethecapacitytorepresentbehavioralprograms.ButIthinkthisideaiswelldennedenoughtobeanalternativehypothesistotheoryofmindexplanationsoftheirfancierbehavioralcapacities.If,forexample,chimpsshowrolereversal,itisanimpressivecognitiveachievement.Itshowsanabilitytoabstractawayfromthemotordetailstoanoverallprogramofaction.Itislikelearningsocialroles,orlearninggameroles.ChimpsmaybeabletolearntherulesoftheGiving/Sharing/Pointinggame.It"sanotherbehavioralprogram.Theyrecognizewhatthegameis,andwhattheirroleisinit,throughthetrainingexperience.Ifchimpsocialbehaviorisstructuredinvariouswayswithoutbeingstereotyped,itwouldnotbesurprisingiftheycoulddothis.Collectivehunting,andchimpwar-making,ifthatturnsouttobe 158KimSterelnypartoftheirstandardrepertoire,maybenaturalexamplesofcoordinatedunster-eotypedactsinwhichdifferentplayersplaydifferentroles,butnotalwaysthesamerole.Heyescertainlydoesnotthinkanyexperimentsofarperformedisstrongenoughtoestablishthataprimateismindreading.Forexample,sheagreesthatthekilljoyresultsofthePovinelliandEddyattentionmonitoringexperimentsarebadnewsfortheideathatchimpsunderstandtheroleofvisualattentioninbehavior.Buthercrucialclaimisthattheexperimentisnotstrongenoughtodeliverapositiveresult.Itcannotconfirmthatchimpsunderstandtheimportanceofwhatanagentseesforwhatitcando.Shearguesthatsimplediscriminationtechniques...cantelluswhichobservablecueschimpanzeesusewhendecidingwhotoapproachforfood,buttheycannottelluswhythechimpanzeesusedthosecues.(Heyes,1998;p108)Imaginethatthedatawereascleanlypositiveastheycouldhavebeen.Westillwouldhavebeenleftwithtwohypotheses.Oneisaccordingtowhichchimpanzeesknowthatoneshouldbegfromthosewithvisibleeyes,becauseonlythosewhoseeyesyoucanseecanseeyou;theotherhypothesissuggeststhatchimpsonlyhaveatendencytobegfrompeoplewithvisibleeyes,andwhilethechimpanzeemayevenknowthatbeggingfrompeoplewithvisibleeyesismorelikelytoleadtoreward,theydonotexplainthiscontingencytothemselvesinmentalterms,orinanyotherway.(p108)Thisisright.Butthereisananalogousproblemfortheexperimentaldesignsheproposes.Sheaddsbellsandwhistlestothedesignofthe"knower/guesser"experi•mentsofPovinelli.Inthese"perspectivetaking"experiments,chimpsweretestedfortheirgripontheconnectionbetweenknowingandseeing.Inthefirststageoftheexperiment,thechimpisinaroomwithtwoexperimenters.Onebaitsabin(theknower)aftertheotherhaslefttheroom.Theabsenttrainer(theguesser)returnsandbothtrainerspointatacontainer.InPovinelli"sexperimenthalfthechimpslearnedtofollowtheknower"sadvice.Butthecritical,secondstageofthisexperimenttestedfortherobustnessofthiseffectbyseeingwhetherthechimpscontinuetoaskfortherightadvicewhentheguesser"sadviceisrenderedobviouslyuselessbyanothermeans—inparticular,byhavingabagovertheheadwhilethebinwasbeingbaited(Povinellietal.,1990).Heyespointsoutthattheoriginalexperimentaldesignisnotquitesensitiveenoughtotestwhethertakingtheknower"sadvicereallydependsongettingtheconnectionbetweenseeingandknowing.Butshethinksthisistherightkindofexperimentbecauseitenablesustozeroinonexactlywhatthechimpsnoticeinasituation.In PrimateWorlds159thetrainingregime—stageoneoftheexperiment—acausalcueandasensorycuecovary.Leavingtheroomisthesensorycuethatcovarieswiththecausalcueofbeingunabletoseewhichofthebinsisbaited.Thesensorycueofstayingintheroomcovarieswiththecausalcueofbeingabletoseethebaitedbin.Intheprobetrials,bothstayintheroom,andtheguesserhasabagoverherhead.InHeyes"view,thisexperimentaldesignfailstoexcludethepossibilityoftransferonthebasisofmerelysensoryrulesofthumb.Perhapschimpswerejusttracking"visiblehead."Inthatcasetheywillstillsucceedintheprobetrials,butnotbyunderstandingtheknowing/seeingconnection.Shethussuggestsanexperimentaldesigninvolvingtheuseoftransparentandopaqueglassestominimizethechancesofchimpsusingasensoryruleofthumbtotrackseeing.Heyes"experimentaldesignwillhelpustellwhetherchimpsarerepresentingratherthandetectingseeing.Herdesignwouldtelluswhetherchimpscanusemultiplecuesfortrackingthepropertiesofignoranceandknowledge.Butthisappearsnottobeherworry.Shedoesnotwanttoprobetherecognitionrulesforseeing.Rather,sheisinterestedinthechimps"understandingofthecausalimportanceofseeing.Butthisisanissueofbreadth,oftheoutputrules.Forthis,IdonotseehowherexperimentaldesignisanyadvanceonthatofPovinelli.Noexperimentaldesigninthisfamilyprobesforbreadthoftracking,forabroadbandofcapacitiestoactonseeing.Oncechimpstrackseeing,tosucceedintheseexperimentstheyneedonlytomasterfairlysimple,unconditionaloutputrules.Theyrepresentseeing,butperhapswithoutgreatbreadth.Forexample,inoneofPremack"sexperiments,onechimpremovedtheblindfoldthatpreventedatrainerfollowinghertoabinshewantedopened.Butsheneedonlyunderstandthatthereisaconnectionbetweenunobstructedlineofsightandthecapacitytofollowher.Sheneednotunderstandthatconnectionismediatedthroughacovertinnerstate,orthatitisimportantforawholeswagofotherpoten•tialinteractions.Similarly,achimpcouldsucceedintheseexperimentsbyfollowingtherule"begforfoodfromthosewhosawit"withouthavinganytheoryoftheinnercausesofbehaviororappreciationoftheroleofseeinginnumerousothercontexts.Theyonlyneedtotrackandrespondtovisualcontact.SoalthoughHeyes"sexperi•mentaldesigncanhelpustriangulateonseeingasthecriticalcuethatchimpsrepre•sentanduse,Idonotseehowitcouldplayacriticalroleinhelpingusidentifyarecognizablethoughrudimentaryfolkpsychologyinchimpminds.Nothinginthisargumentdownplaysthevalueofexperimentallyprobingchimpworlds.Rather,thepointisthatnosingleexperimentbyitselfrevealsbroadbehav•ioralcompetenceontheprimates"part.Moreover,thereisnoprivilegedbreadth,theattainmentofwhichwecouldreasonablyregardascriteriaforhavingfolkpsychol•ogy.OurevidenceherewillbeofthesortWhitenandDennettrecommend.Themore 160KimSterelnycircumstancesinwhichchimpscanexploitandusetherelevanceoflineofsight,themoretheycountasunderstandingthatlineofsiteismediatedthroughthemind.Itakethispointtobemethodologicalratherthanmetaphysical.IfFodorisright,thereissomeuniquepieceofcognitivearchitecturethatconstituteshavingfolkpsychol•ogy.Evenso,itsexperimentalsignaturewillbefuzzy.SoIthinkthisisthebestwecando:achimpanzeehasmoreofatheoryofmind(1)themorementalstatesofothersitcantrack;(2)themoreoftheseitcanrepresentratherthandetect;(3)thebroaderitsbehavioralcompetencesare,withrespecttothosetrackedstates.Butthisisgoodenoughtogoonwith.SummaryInmyview,anorganismrepresentsafeatureofitsworld,asdistinctfrommerelyrespondingtoit,ifitcantrackthatenvironmentalfeatureviamorethanonekindofproximalstimulus.Arthropodsoftenrespondtotheworldinadaptivelycomplexwaysbutviaasingleinformationchannel.Theyare(often)cue-boundwithrespecttothefeaturesoftheworldthatmattertothem,andonlydetectandrespondtotheirworld.Incontrast,animalscapableofmirrorself-recognitionshowanabilitytotrackbodilyfeaturesusingunusualperceptualinputs.Iusethisrepresentation/detectiondistinctiontogetafixonrepresentationalcapacitiesofprimates.Aprimaterespondstothementalstateofanotherifitcantrack—thatis,responddistinctivelywithsomereliability—tosomesuiteofbehaviorsthatareactuallycausedbysomespecificmentalstate.If,forexample,abonoboappeasesangrybehaviorbytradingsexforpeace,it"strackinganger.Weinvestigatewhetherthisisangerdetectionorangerrepresentationthroughexperimentallyprobingtherobustnessofthetracking.Atthesametime,wecanexperimentallyinvestigatethesophisticationofthistrackingbyprobingthebreadthoftheresponsestoanger.Doestheanger-readeradapttoangrybehaviordifferentlyinenvironmentsthatcausethatbehaviortobeexpresseddifferently?Sothispicturedefinestwosep•arateexperimentalinvestigations.Weinvestigateaprimate"scapacitytorepresentamentalstatebyinvestigatingtherobustnessofitsabilitytotrackthatmentalstate.Robustnessisthevarietyofobservationalcuesitusesintracking.Wecanalsoinvestigatethebreadthofitsresponsetotracking.Breadthistheextenttowhichthetracker"sexpectationsabouttheagent"sbehaviororitsappropriateresponseareappropriatelymodifiedbytheagent"senvironmentandtheothermentalstatesthemindreadertracks.Ananimal"ssocialintelligencedevelopsviatwosortsofbehaviorrules.Recogni•tionruleslinkareadertotrackedmentalstates.Ananimalismindreadingonlyifit PrimateWorlds161has,forsomementalstates,abatteryofrecognitionrules.Outputrulesgovernresponsestothestatesareadercantrack.Weprobeananimal"soutputrulesbyfix•ingthereader"scue,varyingtheenvironment,andtestingfordifferentresponses.Notes1.ThecentralideaofthesocialintelligencehypothesisisduetoHumphrey(Humphrey,1976)andJolly(Jolly,1966),bothreprintedinByrneandWhiten,1988.Sincetheirearlyformulations,ithasbeendevel•opedinsomewhatdifferentwaysbyDunbar,Tomasello,Byrne,Whiten,andothers.2.Extratriggersmaywellhaveafunctionalsalienceoftheirown.Addingredundancyintobehavioralcontrolmaywellbeimportantiffalsenegativesaretobeavoided.3.ThereisaniftydiscussionofthispossibilityinHauser"sdiscussionoffoodcallingindomesticchickens.Malesproducefalsecallsagooddeallessoftenwhenfemalesarecloseandinapositiontoseewhetherthemalehasfood(Hauser,1997).Thisisprobablyacasebothofenvironmentalandbehavioralleaking.IfIremembermydaysasachickenherdcorrectly,callingisaccompaniedbyexcitedpecking.Sounlessthemalesaresmartenoughtopeckatnothing(whichIdoubt),theirbehavior,notjusttheircircumstances,willgivethemaway.Hensareable,ifHauserisright,tocross-checkasocialsignaloffood(malecalling)againstphysicalevidence.ReferencesByrneRW(1995)Thethinkingape:Evolutionaryoriginsofintelligence.Oxford:OxfordUniversityPress.ByrneRW(1997)Thetechnicalintelligencehypothesis:Anadditionalevolutionarystimulustointelli•gence?In:MachiavellianintelligenceII:Extensionsandevaluations(ByrneR,WhitenA,ed),pp289-311.Cambridge:CambridgeUniversityPress.ByrneRW,WhitenA(1988)Machiavellianintelligence.Socialexpertiseandtheevolutionofintellectinmonkeys,apes,andhumans.NewYork:OxfordUniversityPress.deWaalF(1989)Peacemakingamongprimates.Harvard:HarvardUniversityPress.DennettD(1983)Intentionalsystemsincognitiveethology:The"panglossianparadigm"defended.Be•havioralandBrainSciences,6:343-390.DickinsonA,ShanksD(1995)Instrumentalactionandcausalrepresentation.In:Causalcognition:Amultidisciplinarydebate(SperberD,PremackD,PremackAJ,ed),pp5-24.Oxford:ClarendonPress.GomezJ-C(1996)Non-humanprimatetheoriesof(non-humanprimate)minds:Someissuesconcerningtheoriginsofmind-reading.In:Theoriesoftheoriesofmind(CarmthersP,SmithP,ed),pp330-343.Cambridge:CambridgeUniversityPress.HauserM(1997)Mindingthebehaviorofdeception.In:MachiavellianintelligenceII:Extensionsandevaluations(ByrneR,WhitenA,ed),pp112-143.Cambridge:CambridgeUniversityPress.HeyesCM(1994)Reflectionsonself-recognitioninprimates.AnimalBehaviour,47:909-919.HeyesCM(1995)Self-recognitioninprimates:Furtherreflectionscreateahallofmirrors.AnimalBehav•iour,51:1533-1541.HeyesCM(1998)Theoryofmindinnonhumanprimates.BehavioralandBrainSciences,21:101-134.HumphreyNK(1976)Thesocialfunctionofintellect.In:Growingpointsinethology(BatesonPPG,HindeRA,ed),pp303-317.Cambridge:CambridgeUniversityPress.JollyA(1966)Lemursocialbehaviourandprimateintelligence.Science,153:501-506.KrebsJ,DawkinsR(1984)Animalsignals,mind-readingandmanipulation.In:Behaviouralecology:Anevolutionaryapproach(KrebsJR,DaviesNB,ed),pp380-402.Oxford:BlackwellScientific. 162KimSterelnyPovinelliD,CantJGH(1995)Arborealclamberingandtheevolutionofself-conception.QuarterlyReviewofBiology,70:393-421.PovinelliDJ,EddyTJ(1996)Whatyoungchimpanzeesknowaboutseeing.MonographsoftheSocietyforResearchinChildDevelopment,61:1-152.PovinelliDJ,NelsonKE,BoysenST(1990)Inferencesaboutguessingandknowingbychimpanzees(Pantroglodytes).JournalofComparativePsychology,104:203-210.PremackD,WoodruffG(1978)Doesthechimpanzeehaveatheoryofmind?BehavioralandBrainSciences,1:515-526.RussonAE(1997)Exploitingtheexpertiseofothers.In:MachiavellianintelligenceII:Extensionsandevaluations(ByrneR,WhitenA,ed),pp174-206.Oxford:OxfordUniversityPress.SoberE(1998)Blackboxinference—whenshouldinterveningvariablesbepostulated?BritishJournalforthePhilosophyofScience,49:469-498.SterelnyK(1995)Basicminds.PhilosophicalPerspectives,9:251-270.SterelnyK(1997)Navigatingthesocialworld:Simulationversustheory.PhilosophicalBooks,37:11-29.SterelnyK(1998)Intentionalagencyandthemetarepresentationalhypothesis.MindandLanguage,13:11-28.SterelnyK(1999)Situatedagents:Thedescentofdesire.In:Biologymeetspsychology:Constraints,con•jectures,connections(HardcastleV,ed),pp203-219.Cambridge:MITPress.WhitenA(1996)Whendoessmartbehaviour-readingbecomemindreading?In:Theoriesoftheoriesofmind(CarruthersP,SmithP,ed),pp277-292.Cambridge:CambridgeUniversityPress.WhitenA(1997)TheMachiavellianmindreader.In:MachiavellianintelligenceII:Extensionsandevalu•ations(WhitenA,ByrneRW,ed),pp144-172.Oxford:OxfordUniversityPress. 9TwoHypothesesAboutPrimateCognitionMichaelTomaselloIwouldliketoproposetwohypothesesaboutprimatecognition.Thefirsthastodowiththecognitionofprimatesasanorder,andhowitmaybedifferentfromthatofothermammals;thesecondhastodowiththecognitionofhumanbeingsasaspecies,andhowitmaybedifferentfromthatofotherprimates.Bothoftheproposeddistinctionsrepresentnovelbiologicaladaptationsthatprimarilyconcernprocessesofsocialcognition.Theuniquelyprimatesocial-cognitiveadaptationmakespossibleamorecomplexsociallife;theuniquelyhumansocial-cognitiveadaptationmakespossibleaculturallife.UniquelyPrimateCognitionThereisexcellentevidencethattheindividualsofmostspeciesofprimaterecognizeotherindividualsintheirowngroups.Ininteractingwiththeseindividuals,theyformsocialrelationshipswiththem.Inparticularsituationstheyarequitegoodatpre•dictingwhatotherindividualswilldo,baseduponbothgeneralcontextualinforma•tionandontheirpastexperiencewithparticularindividuals"behavioraltendencies.Theyformcomplexcoalitionsandallianceswithgroupmates,andcooperateandcompetewiththeminmanyotherkindsofimportantactivities.Mostprimateindi•vidualsfollowthegazeofconspecifics,andlearnfromthemsociallyinanumberofdifferentsituations.ThelargeamountofevidencesupportingtheexistenceoftheseprimatecognitiveskillsisextensivelyreviewedinTomaselloandCall(1994,1997).Theproblemforthequestionofprimateuniquenessisthatmanyothermamma•lianspeciesrecognizeindividualsintheirgroups,formrelationshipswiththem,and"read"theirbehaviorinvariousways(GreenandMarler,1979).Inaddition,anumberofnonprimatemammalsalsoformcoalitionsandalliances—thebest-studiedbeinglions,hyenas,anddolphins(seepapersinHarcourtanddeWaal,1992)—andalsocooperatewithoneanotherinotherwaysaswell(PackerandRuttan,1988).Butonedifferenceinthecoalitionsandalliancesofprimates,pointedoutbyHarcourt(1992),isthatprimatesseemtoselecttheirpartnersbasedontheirappropriateness,giventheopponentandcircumstance.Thus,whenfemalebaboonsengageinapro•tectedthreat,theydonotveryoftensolicitalow-rankingmaleinordertothreatenahigh-rankingmale,butrathertheopposite.Thisselectivitywithregardtotherelativerankofallyandfoesuggeststhepossibilitythatprimatesdonotjustunderstandtheirownsocialrelationships,buttheyalsounderstandsomethingoftherelationshipsthatthirdpartieshavewithoneanother. 166MichaelTomaselloFigure9.1Naturalisticevidenceofprimateunderstandingofthird-partysocialrelationshipsinvolvingdominance,kinship,and"friendship."(Notethatsomeofthesebehaviorshaveotherinterpretations;seeTomaselloandCall,1997.)Thereisstrongevidenceforthisunderstandingofthird-partysocialrelationshipsinanumberofdifferentprimatespecies(TomaselloandCall,1994).Observationalevidenceinvolvesatleastsevendifferentcommonbehaviors.Forexample,redirectedaggression,inwhichyouattackmebutIretaliatebyattackingyourrelative-indicatingmyknowledgeofthekinshiprelationbetweenthetwoofyou.Theothersixbehaviorsarelistedandbrieflydescribedinfigure9.1.Inaddition,thereareatleasttwopiecesofexperimentalevidence.First,CheneyandSeyfarth(1980)playedapreviouslyrecordedvocalizationofoneofthreejuvenilevervetmonkeystoitsmotherandtwootheradultfemaleswhoalsohadabsentoffspring.Eachfemalerespondedtoherownjuvenile"scallbylookingtowardthecall"ssource.Butwhenthecallofoneoftheotherfemales"offspringwasplayed,theylookedtotheappro•priatemother—oftenbeforeshehadmadeanyovertmovementorsignofrecogni•tion.Thisanticipatorybehaviorwouldseemtoindicatethatindividualsrecognizethethirdpartyrelationshipholdingbetweenparticularmothersandtheiroffspring.Sec- TwoHypothesesAboutPrimateCognition167ond,andperhapsmostimportantly,Dasser(1988a)foundevidencefortheunder•standingofcategoriesofsocialrelationships.Sherewardedlongtailmacaquesforchoosingvariouspicturesofonemother-offspringpairoverpicturesofpairsofunrelatedgroupmates.Subjectswerethenaskedtochoosebetweenpicturesofothermother-offspringpairsandotherunrelatedpairs.Subjectsconsistentlychosethemother-offspringpairs,demonstratingtheirabilitytoseeacategoricalrelationshipamonganumberofdifferentpairsofthistype(Dasser,1988b,alsoobtainedsimilarresultsinastudyofthesiblingrelationship).Becausetheywereconcernedwithcognitiveskillsthatdistinguishprimatesfromotheranimals,TomaselloandCall(1997)searchedtheliteratureonthesocialmam•malsthathavebeenmoststudied—especially,lions(e.g.,Packer,1994),elephants(e.g.,Moss,1988),hyenas(e.g,Zabeletal.,1992),anddolphins(e.g.,Connoretal.,1992)—andfoundnoneofthesevenpiecesofobservationalevidencefortheunder•standingofthirdpartysocialrelationshipsdepictedintable9.1;andtherearenoexperiments,either.Oneproblem,ofcourse,isthattheseresearchersmaynothavebeenlookingforthesebehaviors.Ingeneral,primatologistsaremoreinterestedincognitiveskillsthanarebehavioralbiologistsstudyingsocialmammals.Andso,althoughfutureresearchmaytelladifferentstory,fornowIthinktheabsenceofevidencefromsocialmammalsissignificant.Myhypothesisisthusthatalthoughmanyanimalsrecognizeindividualsandformrelationshipswiththem,onlyprimatesunderstandandformcategoriesofthirdpartysocialrelationships.Obviously,keep•ingtracknotonlyofone"sowndirectrelationships,butalsotherelationshipsthatholdamongothersinthegroup,makesforspecialcognitivecomplexitiesinthesocialdomain.Verydifferent,butneverthelessconverging,evidenceforthevalidityofthiscon•clusioncomesfromthedomainofphysicalcognition.Ascomparedwithothermammals,thereissomeevidencethatprimatesareespeciallyskillfulindealingwithrelationalcategoriesasmanifest,forexample,indiscriminationlearningproblemsinvolvingoddity,transitivity,andrelationmatching(Thomas,1980).Understandingthesecategoriesisclearlysimilartounderstandingthird-partysocialrelationshipsinthattheybothinvolvetheunderstandingofhowtwoexternalentitiesrelatetooneanother.Intheexperimentalstudieswithobjects,however,ittakesmanyhundredsoftrials,sometimesthousandsoftrials,forindividualstobeginmakingtheappropriatediscriminations.Thisisincontrasttotheunderstandingofcategoriesofthird-partysocialrelationshipsthatseemtocometothemsonaturally.FollowingHumphrey"s(1976)generallineofreasoning,therefore,onehypothesisisthatprimatesevolvedtheabilitytounderstandcategoriesofthird-partysocialrelationships,andinthelaboratorywemaysometimestapintothisskillusingphysicalratherthansocial 168MichaelTomaselloobjectsifwetrainindividualslongenough.Indeed,itisdifficulttothinkofspecificproblemsinthephysicalworldwithwhichtheunderstandingofrelationalcategorieswouldbeofdirecthelp,whereasinthesocialworldthereareallkindsofsituationsinwhichtheunderstandingofthird-partysocialrelationshipswouldimmediatelymakeformoreeffectivesocialaction.Overall,then,itistheunderstandingofrelationalcategoriesingeneralthatisthemajorskilldifferentiatingthecognitionofprimatesfromthatofothermammals(cf.Mackintosh,thisvolume).Thehypothesis,however,isthattheoriginalevolutionaryadaptationwasforunderstandingthethird-partysocialrelationshipsofconspeci-fics,andthismaybeextendedtophysicalobjectsonlywithspecialeffortsinspecialcircumstances.Afurtherhypothesisisthattheunderstandingofrelationalcategoriesisanevolutionaryprecursor—akindofhalfwayhouse—fortheuniquelyhumancognitiveabilitytounderstandtheintentionalrelationsthatconspecificshavetotheexternalworldandthecausalrelationsthatinanimateobjectsandeventshavewithoneanother.UniquelyHumanCognitionItiswidelybelievedthatnonhumanprimateshaveanunderstandingoftheinten-tionalityofconspecificsandthecausalityofinanimateobjectsandevents.Idonotbelievethattheydo,andIhavearguedandreviewedevidenceextensivelyforthisnegativeconclusion(Tomasello,1990,1994,1996;Tomaselloetal.,1993;TomaselloandCall,1994,1997).However,itshouldbemadeveryclearthatthisnegativecon•clusionabouttheunderstandingofintentionalityandcausalityisquitespecificanddelimited.Ibelievethatnonhumanprimatesdohaveanunderstandingofallkindsofcomplexphysicalandsocialconcepts,theypossessanduseallkindsofcognitiverepresentations,theyclearlydifferentiatebetweenanimateandinanimateobjects,andtheyemployintheirinteractionswiththeirenvironmentsallkindsofcomplexandinsightfulproblem-solvingstrategies(asreviewedabove).Itisjustthattheydonotviewtheworldintermsofthekindsofintermediateandoftenhidden"forces"—underlyingcauses,reasons,intentions,andexplanations—thataresoimportanttohumanthinking.Inthesocialrealm,theevidencefornonhumanprimateunderstandingoftheintentionalityofotheranimatebeingsisoftwosorts.First,therearetwoexperi•mentalstudiesthatpurporttoshowthatchimpanzeesunderstandothersasinten•tionalbeings.PremackandWoodruff(1978)hadthechimpanzeeSarahchoosepicturestocompletevideosequencesofintentionalhumanactions(e.g.,shehadto TwoHypothesesAboutPrimateCognition169chooseapictureofakeywhenthehumaninthevideowastryingtoexitalockeddoor).Hersuccessinthetaskledtotheinferencethatsheknewthehuman"sgoalinthedepictedactions.However,Savage-Rumbaughetal.(1978)producedsimilarresultsusingasstimulisimpleassociates;forexample,theirapesalsochoseapictureofakeywhenshownapictureofalockwithnohumanactionoccurringatall.ThisraisesthepossibilitythatwhatSarahwasdoingwassimpleassociativelearning.(Premack[1986]reportedthatinasubsequentstudyhecouldnottrainSarahtodis•criminatevideosofhumansengagedinintentionalvs.nonintentionalactions,andPovinellietal.[1998]werealsonotabletofindthisdiscriminationintheirsixchim•panzeesubjects.TheresultsofCallandTomasello[1998]aremixed.)TheotherstudyisthatofPovinellietal.(1990),whofoundthatchimpanzeespreferredtoaskforfoodfromapersonwhohadwitnesseditshidingoversomeonewhohadnotwitnesseditshiding—theinferencebeingthattheycoulddiscriminatea"knowledgeable"froman"ignorant"human.Theprobleminthiscaseisthattheapesinthisstudyonlylearnedtodothisovermanyscoresoftrialswithfeedbackontheiraccuracyaftereverytrial(Heyes,1994;Povinelli,1994),andthisisalsoaproblemforthestudyofWoodruffandPremack(1979),inwhichchimpanzeeslearnedaftermanytrialswithfeedbacktodirecthumanstotheboxwithoutfoodsotheycouldobtaintheonewithfood(whatsomecall"deception").Theproblemisthusthatthechimpanzeesinthesestudiesdidnotbringaknowledgeofothers"intentionalitytotheexperiment,butrathertheylearnedhowtobehavetogetwhattheywantedastheexperimentunfolded.Inastudyinwhichlearningduringtheexperimentwasallbutruledout,CallandTomasello(inpress)foundthatchim•panzeesshowednounderstandingofthefalsebeliefsofothers.Butbecausealloftheseexperimentsareartificialinvariousways,otherinves•tigatorshaveturnedtothenaturalbehaviorofchimpanzeesandothernonhumanprimatesforpositiveevidenceoftheunderstandingofintentionality.Theprobleminthiscaseisthatalmostallofthereportedobservationsareanecdotesthatlacktheappropriatecontrolobservationstoruleoutcompetingexplanations(ByrneandWhiten,1988).Buteveninreliable(replicable)casesitisnotclearwhatisgoingoncognitively.Forexample,deWaal(1986)observedafemalechimpanzeeonrepeatedoccasionsholdoutherhandtoanotherinanapparentappeasementgesture,butwhentheotherapproachedsheattackedhim.Thismightbeacaseofhumanlikedeception:theperpetratorwantedtheothertobelievethatshehadfriendlyintentionswheninfactshedidnot.Itisjustaslikely,however,thattheperpetratorwantedtheotherindividualtoapproachher(soshecouldattack),andsoperformedabehaviorthathadinthepastledconspecificstoapproachinothercontexts.Thisuseofanestablishedsocialbehaviorinanovelcontextisclearlyaveryintelligentandperhaps 170MichaelTomaselloinsightfulstrategyformanipulatingthebehaviorofothers,butitisnotclearthatitinvolvestheunderstandingandmanipulationoftheintentionalstatesofothers.Ishouldalsopointoutsomethingsthatnonhumanprimatesintheirnaturalhab•itatsdonotdo(apesraisedinsomethingresemblingahumanculturalenvironmentdosomeofthem—seediscussionbelow).Intheirnaturalhabitats,nonhumanprimates:•donotpointorgesturetooutsideobjectsforothers;•donotholdobjectsuptoshowthemtoothers;•donottrytobringotherstolocationssothattheycanobservethingsthere;•donotactivelyofferobjectstootherindividualsbyholdingthemout;•donotintentionallyteachthingstoothers.Theydonotdothesethings,inmyview,becausetheydonotunderstandthattheconspecifichasintentionalandmentalstatesthatcanpotentiallybeaffected.Sothemostplausiblehypothesisisthatnonhumanprimatesunderstandcon-specificsasanimatebeingscapableofspontaneousself-movement—indeed,thisisthebasisfortheirsocialunderstandingingeneralandtheirunderstandingofthird-partysocialrelationshipsinparticular—buttheydonotunderstandothersasinten•tionalagentsintheprocessofpursuinggoals.Nonhumanprimatesseeaconspecificmovingtowardfoodandmayinfer,basedonpastexperience,whatislikelytohap•pennext,andtheymayevenuseintelligentandinsightfulsocialstrategiestoaffectwhathappensnext.Buthumanbeingsseesomethingdifferent.Theyseeaconspecificastryingtoobtainthefoodasagoal,andtheycanattempttoaffectthisandotherintentionalstates,notjustbehavior.Thissomewhatsubtledifferenceofsocialper•ceptionandunderstandingwillbeexplicatedmorefullybelow,aswillsomeoftheprofoundconsequencesithasforallaspectsofsociallifeandcognition.Inthephysicalrealm—andwithspecificreferencetoprimateunderstandingofcausality—Visalberghihasrecentlyobservedsomelimitationsinprimates"skillsatadaptingtonovelforagingtasksinwhichsomeunderstandingofcausalityisrequired(VisalberghiandLimongelli,1996).Thebasictaskinvolvesthesubjectusingasticktopushfoodoutofacleartube.Inonesetoftasks,thetoolsarevaried,withsomebeingtooshort,ortoofat,ornotrigidenoughtoworkproperly.Thebasicideaisthatifanindividualunderstandsthephysicalcausalityinvolvedinhowthestickworkstoextractthefoodfromthetube—physicalforcetransferredfromselftosticktofood—itshouldbeabletopredictjustfromperceptualinspectionofatool,withoutextensivetrialanderror,whetherornotitwilleffecttherequiredcausalsequence.Althoughapesareabitmoreskillfulthancapuchinmonkeysatthistask,bothsue- TwoHypothesesAboutPrimateCognition171ceedwiththenoveltoolsonlyaftermuchtrialanderror.Inarecenttaskvariation,thesesamespeciesweregivenacleartubewithasmalltrapunderonepart.Ifsub•jectsappreciatethecausalforceofgravityandthephysicsofholesandsticksmovingobjects,theyshouldlearntoavoidthistrapastheyattempttopushthefoodthroughthetube(i.e.,theyshouldalwayspushthefoodouttheendawayfromthetrap).Butneithercapuchinsnorchimpanzeeslearnedtodothisquickly;forexample,allfourchimpanzeesubjectsbehavedatchancelevelsfor70ormoretrials.Inafinaltwist,aftertheanimalshadlearnedthroughtrialanderrortoavoidthetrap,thetubewasflippedover—sothatthetrapwasontopofthetubeandposingnodanger.Subjectsofbothspecies(thechimpanzeesinastudybyReaux,1995)stillpushedthefoodawayfromthetrap,notunderstandingitsnewharmlessstatus.Two-tothree-year-oldchildrenbehavemuchmoreflexiblyandadaptivelyonthesetubeproblems-seemingtounderstandsomethingofthecausalprinciplesatwork—fromtheveryearliesttrials(VisalberghiandLimongelli,1996).Theconclusionisthusthatnon-humanprimateshavemanycognitiveskillsinvolvingphysicalobjectsandevents;theyjustdonotperceiveorunderstandunderlyingcausesasmediatingthedynamicrelationsamongtheseobjectsandevents.Bywayofsummary,Iwouldliketobefullyexplicitaboutwhatdifferentiatesintentional/causalcognitionfromothertypesofcognition.First,thisformofthinkingconcernstherelationsamongexternalevents.Thus,manyanimalsunderstandthattheirownactionsproduceresultsintheworld,andtheyrepeattheactionsneededtoproducedesiredresults;thisisthesinequanonofbehavioralandcognitiveadapta•tions.Butonlyprimatesalsounderstandandcategorizesomethingoftherelationsamongexternalentities,evenintheabsenceoftheirownbehavioralinvolvement.But,inaddition,theunderstandingofintentionalityandcausalityconcerntheunderstandingofmediatingforcesintheseexternaleventsthatexplain"why"aparticularantecedent-consequentsequenceoccursasitdoes(typicallynotreadilyobservable),andthisunderstandingisuniquetohumans.Thus,forhumans,theweightofthefallingrock"forces"thelogtosplinter;thegoalofobtainingfood"forces"theorganismtolookunderthelog.Obviously,thewaytheseforcesworkareverydifferentinthecausalityofinanimateobjectsandtheintentionalityofani•matebeings,buttheoverallstructureofthereasoningprocessesinvolvedisofthesamegeneralnature.Thiscanbeclearlyseeninfigure9.2,whichdepictsonephysicalcausalsituation—differentphysicaleventsthatcreateaforcethatcausesafruittodrop—andonesocialcausalsituation—differentsocialeventsthatcreateapsycho•logicalstatethatcausesanindividualtoflee.Intermsofevolution,then,thehypothesisisthathumanbeingsbuiltdirectlyontheuniquelyprimatecognitiveadaptationforunderstandingexternalrelations;they 172MichaelTomaselloFigure9.2Graphicdepictionofonephysicalcausalevent(top)andonesocialcausalevent(bottom).Inbothcasesmanydifferentantecedenteventsmaycreatetheforcethatcausestheconsequentevent. TwoHypothesesAboutPrimateCognition173justaddedasmallbutimportanttwistintermsofmediatingforcessuchascausesandintentions.Thisscenariogainssomeofitsplausibilityfromthefactthatitpro•videsforcontinuitybetweenuniquelyprimateanduniquelyhumancognitiveadap•tations.Moreover,myhypothesisisthat,justasprimateunderstandingofrelationalcategoriesevolvedfirstinthesocialdomaintocomprehendthird-partysocialrela•tionships,humancausalunderstandingalsoevolvedfirstinthesocialdomaintocomprehendothersasintentionalagents.Thereiscurrentlynowayofknowingifthisistrue,ofcourse,butformanyofthepeopleoftheworld,whentheyareindoubtastothephysicalcauseofanevent,theyquiteofteninvokevarioustypesofanimisticordeisticforcestoexplainit;perhapsthisisthedefaultapproach.Myhypothesisisthattheuniquelyhumanabilitytounderstandexternaleventsintermsofmediatingintentional/causalforcesemergedfirstinhumanevolutiontoallowindividualstopredictandexplainthebehaviorofconspecificsandhassincebeentransportedtodealwiththebehaviorofinertobjects.Wehavenoideawhenthismighthaveoccurred,butonepossibilityisthatitwascharacteristicofmodernhumansastheyfirstevolvedsomewhereinAfricasome150-200,000yearsago—indeeditdefinesmodernhumanscognitively—andthismayevenexplainwhytheyoutcompetedotherhominidsastheymigratedallovertheglobe.Thecompetitiveadvantagesofintentional/causalthinkingaremainlytwo.First,thiskindofcognitionenablesorganismstosolveproblemsinespeciallycre•ative,flexible,andforesightfulways.Thus,inmanycasesintentional/causalunder•standingenablesanindividualtopredictandcontroleventsevenwhentheirusualantecedentisnotpresent—ifthereissomeothereventthatmayservetoinstigatethemediatingforce.Forexample,anindividualmightcreateanovelwaytodistractacompetitorawayfromsomethingoverwhichtheyarecompeting(e.g.,byplacingfoodintheoppositedirection),oranoveltoolforgeneratingtheforceneededtomoveanobstacle.Conversely,ifaneventoccursinacircumstanceinwhichthemediatingforceissomehowblocked,itcouldbepredictedthatitsusualconsequentwillnotfollow.Forexample,anindividualcouldblockthevisualaccessofacom•petitortotheobjectoftheircompetition,orcouldpreventastonefromrollingdownahillbyplacinganotherstoneunderit.Causalandintentionalunderstandingthushaveimmediateconsequencesforeffectiveaction,astheyopenupthepossibilityoffindingnovelwaystoeithermanipulateorsuppressmediatingforces.Thesecondadvantageofintentional/causalunderstandingderivesfromitspowerfultrans•formingroleinprocessesofsociallearning.Thatis,understandingthebehaviorofotherpeopleasintentionaldirectlyenablescertainverypowerfulformsofculturallearningandsociogenesis,andtheseformsofsociallearningaredirectlyresponsibleforthespecialformsofculturalinheritancecharacteristicofhumanbeings. 174MichaelTomaselloCulturalInheritanceInthemostgeneralbiologicalsenseoftheterm,culturalinheritanceisawidespreadphenomenonintheanimalkingdom(seeHeyesandGalef,1996,forarecentsurvey).Itsevolutionarysignificanceisimmenseas,essentially,itallowsforanotherchannelofinheritance,andonethatoperatesonamuchquickertimescale,withmuchmoreflexibility,thanbiologicalinheritance(BoeschandTomasello,1998).Therearemanydifferentformsofculturalinheritanceandtransmission,however.Forexample,individualsofaspeciesmaybeexposedtonewlearningexperiencesbecausetheystayphysicallyclosetoconspecificsbutstillnotlearnanythingfromthebehavioroftheconspecificsdirectly—aswhentheyoungsterstumblesupon,andsolearns,theloca•tionofwaterwhilefollowingitsmother.Alsoimportantareprocessesofstimulusenhancement,inwhichoneanimalisattractedtotheobjectwithwhichanotherisinteracting,andthenlearnsthingsonitsownaboutthatobject—aswhenachim•panzeeyoungsterisattractedtoastickitsmotherhasdiscarded,whichthensetsinmotioncertainindividuallearningexperiences.Insomecases,suchassomebirdspecies,therearealsoadaptivespecializationsforlearningabouttheactualbehaviorofconspecificsthroughsuchthingsasthemimickingofspecies-typicalvocalizations.Thiskindoflearningistrulysocialinthesensethatsomethingisbeinglearnedfromtheactualbehaviorofconspecifics,butitisgenerallyverynarrowlyspecializedwithnorepercussionsbeyondthesinglebehavior.Inthehumancase,however,itseemsthatsomethingmoreprofoundlysocialisatwork.Developinghumanbeingsdonotjustfollowtheirmothers,becomeattractedtoobjectstheirmothersaretouching,ormimicconspecificbehaviorinonenarrowdomain.Humanyoungstersacquirefromtheireldersandotherconspecificsallkindsofculturalbehaviors,skills,artifacts,symbols,conventions,values,attitudes,andbeliefs.Theyareabletodothisbecausethespecies"uniquecognitiveadaptationthatenableshumanstounderstandconspecificsasintentionalagents,likethemselves,potentiatessomeespeciallypowerfulformsofsociallearningknownasculturallearning(Tomaselloetal.,1993).ChimpanzeeCultureAlthoughthe"spread"ofpotatowashingasanovelbehaviorinonegroupofhumanprovisionedJapanesemacaquesiswellknown,itturnsoutthatthemostlikelyexpla•nationforthatbehaviorissociallyinfluencedindividuallearning.Oneindividualinventedthebehaviorbywalkingintothewaterwiththepotatoesthrowntoherbyhumans,andherrelativesandfriendsfollowedherintothewaterwiththeirpotatoesandinventedthebehaviorforthemselves—withperhapssomeprocessesofstimulus TwoHypothesesAboutPrimateCognition175enhancementoperativeaswell(Galef,1992).Amuchbetterspeciesforinvestigatingpossibleculturalprocessesinnonhumanprimatesischimpanzees(McGrew,1992;Wranghametal.,1994).Thebestknownexampleischimpanzeetooluse.Forexample,chimpanzeesintheGombeNationalPark(aswellasseveralothergroupselsewhere)fishfortermitesbyprobingtermitemoundswithsmall,thinsticks.InotherpartsofAfrica,chimpanzeessimplydestroytermitemoundswithlargesticksandattempttoscoopuptheinsectsbythehandful.FieldresearcherssuchasBoesch(1993)andMcGrew(1992)claimthatspecifictoolusepracticessuchastheseare"culturallytransmitted"amongtheindividualsofthevariouscommunities.Theproblemisthatitispossiblethatchim•panzeesinsomelocalitiesdestroytermitemoundswithlargesticksbecausethemoundsaresoftfromrain,whereasinotherlocalitiesthereislessrain,sothemoundsareharder,andthusthechimpanzeestherecannotusethisstrategy.Insuchacasetherewouldbegroupdifferencesofbehavior—superficiallyresemblinghumancul•turaldifferences—butwithnotypeofsociallearninginvolvedatall.Insuchcasesthe"culture"issimplyaresultofindividuallearningdrivenbythedifferentlocalecologiesofthedifferentpopulations(andsoitissometimescalled"environmentalshaping").Althoughenvironmentalshapingislikelyapartoftheexplanationsforgroupdif•ferencesofbehaviorforallspecies,experimentalstudieshavedemonstratedthatmorethanthisisgoingoninchimpanzeeculture.Tomasello(1996)reviewedalloftheexperimentalevidenceonchimpanzeeimitativelearningoftooluse(atotaloffivestudies)andconcludedthatchimpanzeesareverygoodatlearningaboutthedynamicaffordancesofobjectsthattheydiscoverthroughwatchingothersmanipu•latethem,buttheyarenotskillfulatlearningfromothersanewbehavioralstrategyperse.Forexample,ifamotherrollsoveralogandeatstheinsectsunderneath,herchildwillverylikelyfollowsuit.Thisissimplybecausethechildlearnedfromthemother"sactthatthereareinsectsunderthelog—afactshedidnotknowandverylikelywouldnothavediscoveredonherown.Butshedidnotlearnhowtorolloveralogortoeatinsects;thesearethingsshealreadyknewhowtodoorcouldlearnhowtodoonherown.(Thus,theyoungsterwouldhavelearnedthesamethingifthewind,ratherthanitsmother,hadcausedthelogtorolloverandexposetheants.)ThisiswhatIhavecalledemulationlearning,becauseitislearningthatfocusesontheenvironmentaleventsinvolved—thechangesofstateintheenvironmentthattheotherproduced—notonaconspecific"sbehaviororbehavioralstrategy(seealsoNagelletal.,1993).Chimpanzeesarethusveryintelligentandcreativeinusingtoolsandunderstand•ingchangesintheenvironmentbroughtaboutbythetooluseofothers,buttheydo 176MichaelTomasellonotseemtounderstandtheinstrumentalbehaviorofconspecificsinthesamewayasdohumans.Forhumans,thegoalorintentionofthedemonstratorisacentralpartofwhattheyperceive,andindeedthegoalisunderstoodassomethingseparatefromthevariousbehavioralmeansthatmaybeusedtoaccomplishthegoal.Observers"abilitytoseparategoalandmeansservestohighlightforthemthedemonstrator"smethodorstrategyoftooluseasanindependententity—thebehaviorusedinanattempttoaccomplishthegoal,giventhepossibilityofothermeansofaccomplishingit.Intheabsenceofthisabilitytounderstandgoalandbehavioralmeansasseparableintheactionsofothers,chimpanzeeobserversfocusonthechangesofstate(includingchangesofspatialposition)oftheobjectsinvolvedduringthedemonstration,withthemotionsofthedemonstratorbeing,ineffect,justothermotions.Theintentionalstatesofthedemonstrator,andthusherbehavioralmethodsasdistinctbehavioralentities,aresimplynotapartoftheirexperience.Theotherwell-knowncaseisthegesturalcommunicationofchimpanzees(Good-all,1986;Tomasello,1990).Inongoingstudiesofthegesturalsignallingofacaptivecolonyofchimpanzees,Tomaselloandcolleagueshaveaskedwhetheryoungstersacquiretheirgesturalsignalsbyimitativelearningorbyaprocessofontogeneticritualization(Tomaselloetal.,1985,1989,1994,1997).Inontogeneticritualizationacommunicatorysignaliscreatedbytwoorganismsshapingeachothers"behaviorinrepeatedinstancesofasocialinteraction.Forexample,aninfantmayinitiatenursingbygoingdirectlyforthemother"snipple,perhapsgrabbingandmovingherarmintheprocess.Insomefutureencounterthemothermightanticipatetheinfant"simpendingbehavioraleffortsatthefirsttouchofherarm,andsobecomereceptiveatthatpoint—leadingtheinfantonsomefutureoccasionstilltoabbreviateitsbehaviortoatouchonthearmwhilewaitingforareponse("arm-touch"asaso-calledintentionmovement).Notethatthereisnohintherethatoneindividualisseekingtorepro•ducethebehaviorofanother;thereisonlyreciprocalsocialinteractionoverrepeatedencountersthatresultseventuallyinacommunicativesignal.Thisispresumablythewaythatmosthumaninfantslearnthe"arms-over-head"gesturetorequestthatadultspickthemup(Locke,1978).Alloftheavailableevidencesuggeststhatontogeneticritualization,notimitativelearning,isresponsibleforchimpanzees"acquisitionofcommunicativegestures.First,thereareanumberofidiosyncraticsignalsthatareusedbyonlyoneindividual(seealsoGoodall,1986).Thesesignalscouldnothavebeenlearnedbyimitativeprocessesandsomusthavebeenindividuallyinventedandritualized.Second,longi•tudinalanalyseshaverevealedquiteclearly,bybothqualitativeandquantitativecomparisons,thatthereismuchindividualityintheuseofgestureswithmuchindi•vidualvariabilitybothwithinandacrossgenerations—suggestingsomethingother TwoHypothesesAboutPrimateCognition177thanimitativelearning.Itisalsoimportantthatthegesturesthataresharedbymanyyoungstersaregesturesthatarealsousedquitefrequentlybycaptiveyoungstersraisedinpeergroupswithnoopportunitytoobserveolderconspecifics.Finally,inanexperimentalstudy,Tomaselloetal.(1997)removedanindividualfromthegroupandtaughthertwodifferentarbitrarysignalsbymeansofwhichsheobtaineddesiredfoodfromahuman.Whenshewasthenreturnedtothegroupandusedthesesamegesturestoobtainfoodfromahumaninfullviewofothergroupmembers,therewasnotoneinstanceofanotherindividualreproducingeitherofthenewgestures.Myconclusionisthusthatchimpanzeeyoungstersacquirethemajority,ifnotthetotality,oftheirgesturesbyindividuallyritualizingthemwithoneanother.Theexplanationforthislearningprocessisanalogoustotheexplanationforemulationlearninginthecaseoftooluse.Likeemulationlearning,ontogeneticritualizationdoesnotrequireindividualstounderstandthebehaviorofothersasseparableintomeansandgoalsinthesamewayasdoesimitativelearning.Imitativelylearninganarm-touchasasolicitationfornursingwouldrequirethataninfantobserveanotherinfantusinganarm-touchandknowwhatgoalitwaspursuing(viz.,nursing)—sothatwhenithadthesamegoalitcouldusethesamebehavioralmeans(viz.,arm-touch).Ritualizinganarm-touch,ontheotherhand,onlyrequirestheinfanttoanticipatethefuturebehaviorofaconspecificinacontextinwhichit(theinfant)alreadyhasthegoalofnursing.OnceagainImustemphasizethatontogeneticritu•alizationisaveryintelligentandcreativesociallearningprocessthatisveryimpor•tantinallsocialspecies,includinghumans.Butitisnotalearningprocessbymeansofwhichindividualsattempttoreproducethebehavioralstrategiesofothers.Thesetwodomainsthusprovideuswithtwoverydifferentsourcesofevidenceaboutnonhumanprimatesociallearning.Inthecaseoftooluse,itisverylikelythatchimpanzeesacquirethetooluseskillstheyareexposedtobyaprocessofemulationlearning.Inthecaseofgesturalsignals,itisverylikelythattheyacquiretheircom•municativegesturesthroughaprocessofontogeneticritualization.Bothemulationlearningandontogeneticritualizationrequireskillsofcognitionandsociallearning,eachinitsownway,butneitherrequiresskillsofimitativelearninginwhichthelearnercomprehendsboththedemonstrator"sgoalandthestrategysheisusingtopursuethatgoal—andtheninsomewayalignsthisgoalandstrategywithherown.Indeed,emulationlearningandontogeneticritualizationarepreciselythekindsofsociallearningonewouldexpectoforganismsthatareveryintelligentandquicktolearn,butthatdonotunderstandothersasintentionalagentswithwhomtheycanalignthemselves.Itshouldalsobenotedthatwithrespecttotheothermainprocessinvolvedinculturaltransmission,teaching,onlythestudyofBoesch(1991)reportsevidenceofchimpanzeeteaching,andthisforonlytwoclearinstances(overmany 178MichaelTomaselloyearsofobservation).Thisisascomparedwithmuchintentionalteachinginallhumansocieties(KrugerandTomasello,1996).Thisdifferenceagainisplausiblyattributedtodifferencesofsocialcognition,asteachingrequiressomeunderstandingofwhatothersdoanddonotknow(CheneyandSeyfarth,1990).HumanCulturalEvolutionandtheRatchetEffectWemayconclude,then,thatwhereaschimpanzeesclearlycreateandmaintaincul•turaltraditionsverybroadlydenned,theserestondifferentprocessesofsocialcognitionandsociallearningthantheculturaltraditionsofhumanbeings.Insomecasesthisdifferenceofprocessmaynotleadtoanyconcretedifferencesofoutcomeinsocialorganization,informationtransmission,orcognition.Butinothercasesacrucialdifferenceemerges.Thismanifestsitselfinprocessesofculturalevolution,thatis,processesbywhichaculturaltraditionchangesovertimewithinapopulation.Onepossiblechangeisthataparticularculturaltraditioncouldsimplydieout,forexample,iftheenvironmentalfunctiondisappearedorthesocialstructureofthegroupchanged,orforanyofmanyotherreasons.Itmightalsohappenthatacul•turaltraditionstaysthesameoveralongperiodoftime,asitservesitsfunctionadequatelyandenvironmentalconditionsremainconstant.Butbeyonddyingoutandstayingthesame,someculturaltraditionschangeovertimeinwaysthatseemtobeadaptiveand,moreover,inwaysthatseemtoaccumu•latemodificationsmadebydifferentindividualsovertimeinthedirectionofgreatercomplexitysuchthatawiderrangeoffunctionsisencompassed—whatmaybecalledcumulativeculturalevolutionorthe"ratcheteffect."Forexample,thewaythathumanbeingshaveusedobjectsashammershasevolvedsignificantlyoverhumanhistory.Thisisshownintheartifactualrecordbyvarioushammerliketoolsthatgraduallywidenedtheirfunctionalsphereastheyweremodifiedagainandagaintomeetnovelexigencies,goingfromsimplestones,tocompositetoolscomposedofastonetiedtoastick,tovarioustypesofmodernmetalhammers,andevenmechanicalhammers(somewithnail-removingfunctionsaswell;Basalla,1988).Althoughwedonothavesuchadetailedartifactualrecord,itispresumablythecasethatsomecul•turalconventionsandrituals(e.g.,humanlanguagesandreligiousrituals)havebecomemorecomplexovertimeaswell,astheyweremodifiedtomeetnovelcom•municativeandsocialneeds.Thisprocessmaybemorecharacteristicofsomehumanculturesthanothers,orsometypesofactivitiesthanothers,butallcultureswouldseemtohaveatleastsomeartificatsproducedbytheratcheteffect.However,theredonotseemtobeanybehaviorsofotheranimalspecies,includingchimpanzees,thatshowcumulativeculturalevolution(BoeschandTomasello,1998). TwoHypothesesAboutPrimateCognition179Tomaselloetal.(1993)arguethatcumulativeculturalevolutiondependsonimi•tativelearning,andperhapsactiveinstructiononthepartofadults,andcannotbebroughtaboutbymeansof"weaker"formsofsociallearningsuchaslocalen•hancement,emulationlearning,ontogeneticritualization,oranyformofindividuallearning.Theargumentisthatcumulativeculturalevolutiondependsontwopro•cesses,innovationandimitation(possiblysupplementedbyinstruction),thatmusttakeplaceinadialecticalprocessovertimesothatonestepintheprocessenablesthenext.Thus,ifoneindividualchimpanzeeinventedamoreefficientwayoffishingfortermitesbyusingastickinanovelwaythatinducedmoretermitestocrawlontoit,youngsterswholearnedtofishviaemulationofthisindividualwouldnotreproducethisprecisevariantbecausetheywouldnotbefocusedontheinnovator"sbehavioraltechniques.Theywouldusetheirownmethodoffishingtoinducemoretermitesontothestick,andanyotherindividualswatchingthemwouldusetheirownmethodsaswell,andsothenovelstrategywouldsimplydieoutwiththeinventor.(Thisispre•ciselythehypothesisofKummerandGoodall,[1985],whobelievethatmanyactsofcreativeintelligenceonthepartofnonhumanprimatesgounobservedbyhumansbecausetheyarenotfaithfullypreservedinthegroup.)Ontheotherhand,ifobserverswerecapableofimitativelearning,theymightadopttheinnovator"snewstrategicvariantfortermitefishingmoreorlessfaithfully.Thiswouldputthemintoanewcognitivespace,sotospeak,inwhichtheycouldthinkaboutthetaskandhowtosolveitinsomethinglikethemanneroftheinnovator.Alloftheindividualswhohavedonethisaretheninaposition,possibly,toinventothervariantsthatbuildontheintitialone—whichothersthenmightadoptfaithfully,orevenbuildon,aswell.Themetaphoroftheratchetinthiscontextismeanttocapturethefactthatimitativelearning(withorwithoutactiveinstruction)providesthekindoffaithfultransmissionthatisnecessarytoholdthenovelvariantinplaceinthegroupsoastoprovideaplatformforfurtherinnovations,withtheinnovationsthemselvesvaryinginthedegreetowhichtheyareindividualorsocial/cooperative.Andsotheoverallconclusionisthathumanculturaltraditionsmaybemostreadilydistinguishedfromchimpanzeeculturaltraditions—aswellasthefewotherinstancesofcultureobservedinotherprimatespecies—bymeansoftheirtrajectoriesovertime.Humanculturaltraditionsaccumulatemodificationsovertime;thatis,theyhave"histories."Theydothisbecausetheculturallearningprocessesthatsup•portthemareofanespeciallypowerfulsort.Theirpowerderivesfromthefactthattheyaresupportedbytheuniquelyhumancognitiveadaptationforunderstandingothersasintentionalbeingsliketheself—whichcreatesformsofsociallearningthatactasaratchetbyfaithfullypreservingnewlyinnovatedstrategiesinasocialgroup 180MichaelTomasellountilthereisanotherinnovationthatreplacesit.Aspreviouslynoted,theseinnova•tionsmayvaryinthedegreetowhichtheyareindividualorsocial/cooperative,andindeedculturalinnovationsinthecaseofhumanscompriseacomplexmixofindi•vidualandsocialprocesses.ConclusionEvolutionaryfairytalesabouttheevolutionofhumancognitionarecurrentlyex•tremelypopular.Allofthem,includingthisone,arehighlyspeculative(cf.,Bitterman,thisvolume;Shettleworth,thisvolume).Butmostoftheseevolutionaryfairytalesdonottakeexplicitaccountofthefactthathumanbeingsandchimpanzeeshavebeenreproductivelyisolatedforonly6millionyearsorso—aboutthesametimethatmiceandratshavebeenreproductivelyisolated.Andsowecannotjustmultiplycognitivedifferencesbetweenhumansandotherprimatesatwillandpositageneticbasisforeachone.Therehasnotbeenenoughtimeforthat.Myproposalisthatthereisjustonemajorcognitivedifferencebetweenhumansandtheirnearestprimaterelatives—theunderstandingofothersasintentionalbeingsliketheself—andthatthisthenenablesaradicallynewformofculturalinheritance,whichchangestheprocessofcognitiveevolution.Thefactthatthisoneuniquelyhumancognitiveadaptationmaybeseenasanextensionoftheoneuniquelyprimatecognitiveadaptation—theunder•standingofrelationalcategoriesingeneral—isafurtheradvantageofthehypothesis.Andsomyevolutionaryfairytale,likeallevolutionaryfairytales,ismassivelyunderspecifiedbythedata.Wehavevirtuallynodirectevidenceofthelivesledbymanyofourimportantprimateancestors,includingboththecommonancestortochimpanzeesandhumanssome6millionyearsagoandthefirstmodernhumanssome200,000yearsago.Butthestudyofprimatebehaviorandcognitionisgraduallybuildinguptoapointwherewemayaskasharpersetofquestionsthanwashereto•forepossibleoftheminimalfossilandartifactualevidencethatwedohave.Thegoalistointegrateinformationgainedfromtheobservationalandexperimentalstudyoflivingprimateswiththatgainedfromthestudyoffossilandartifactualremainstomakeforevermoreconstrainedandaccuratetheoriesontheevolutionofprimate,includinghuman,cognition.SummaryTwohypothesesaboutprimatecognitionareproposed.First,itisproposedthatpri•mates,butnotothermammals,understandcategoriesofrelationsamongexternal TwoHypothesesAboutPrimateCognition181entities.Inthephysicaldomain,primateshavespecialskillsintaskssuchasoddity,transitivity,andrelationmatchingthatrequirefacilitywithrelationalcategories;inthesocialdomain,primateshavespecialskillsinunderstandingthethird-partysocialrelationshipsthatholdamongotherindividualsintheirgroups.Second,itispro•posedthathumans,butnototherprimates,understandthecausalandintentionalrelationsthatholdamongexternalentities.Inthephysicaldomainonlyhumansunderstandcausalforcesasmediatingtheconnectionbetweensequentiallyorderedevents;inthesocialdomainonlyhumansunderstandthebehaviorofothersasintentionallydirectedandcontrolledbydesiredoutcomes.Intentional/causalunder•standingofthistypeopensthewayforcumulativeculturalevolution(theratcheteffect),whichisresponsibleformanyofhumans"mostdistinctivecognitiveskillsandproducts.Boththeseuniquelyprimateandtheseuniquelyhumancognitiveskillsarehypothesizedtohavetheiroriginsinadaptationsfornegotiatingcomplexsocialinteractions.ReferencesBasallaG(1988)Theevolutionoftechnology.Cambridge:CambridgeUniversityPress.BoeschC(1991)Teachingamongwildchimpanzees.AnimalBehaviour,41:530-532.BoeschC(1993)Towardsanewimageofcultureinwildchimpanzees?BehavioralandBrainSciences,16:514-515.BoeschC,TomaselloM(1998)Chimpanzeeandhumanculture.CurrentAnthropology,39:591-604.ByrneRW,WhitenA(1988)Machiavellianintelligence.Socialexpertiseandtheevolutionofintellectinmonkeys,apes,andhumans.NewYork:OxfordUniversityPress.CallJ,TomaselloM(1998)Distinguishingintentionalfromaccidentalactionsinorangutans,chimpanzees,andhumanchildren.JournalofComparativePsychology,112:192-206.CallJ,TomaselloM(1999)Anonverbalfalsebelieftask:Theperformanceofchildrenandgreatapes.ChildDevelopment,70:381-395.CheneyDL,SeyfarthRM(1980)Vocalrecognitioninfree-rangingvervetmonkeys.AnimalBehaviour,28:362-367.CheneyDL,SeyfarthRM(1990)Howmonkeysseetheworld.Chicago:UniversityofChicagoPress.ConnorRC,SmolkerRA,RichardsAF(1992)Dolphinalliancesandcoalitions.In:Coalitionsandalli•ancesinhumanandnonhumananimals(HarcourtAH,deWaalFBM,ed),pp415-443.NewYork:OxfordUniversityPress.DasserV(1988a)AsocialconceptinJavamonkeys.AnimalBehaviour,36:225-230.DasserV(1988b)Mappingsocialconceptsinmonkeys.In:Machiavellianintelligence.Socialexpertiseandtheevolutionofintellectinmonkeys,apes,andhumans(ByrneRW,WhitenA,ed),pp85-93.NewYork:OxfordUniversityPress.deWaalF(1986)Deceptioninthenaturalcommunicationofchimpanzees.In:Deception:Perspectivesonhumanandnonhumandeceit(MitchellR,ThompsonN,ed),pp221-244.Albany,NY:SUNYpress.GalefJr,B.G.(1992)Thequestionofanimalculture.HumanNature,3:157-178. 182MichaelTomaselloGoodallJ(1986)ThechimpanzeesofGombe.Patternsofbehavior.Cambridge,MA:HarvardUniversityPress.GreenS,MarlerP(1979)Theanalysisofanimalcommunication.In:Handbookofbehavioralneuro•biology(MarlerP,VandenbergJ,ed),pp73-158.NewYork:Plenum.HarcourtAH(1992)Coalitionsandalliances:Areprimatesmorecomplexthannon-primates?In:Coali•tionsandalliancesinhumanandnonhumananimals(HarcourtAH,deWaalFBM,ed),pp445-471.NewYork:JohnWileyandSons.HarcourtAH,deWaalFBM(1992)Coalitionsandalliancesinhumansandotheranimals.NewYork:OxfordUniversityPress.HeyesCM(1994)Cues,convergenceandacurmudgeon:AreplytoPovinelli.AnimalBehaviour,48:242-244.HeyesCM,GalefJr,B.G.,ed(1996).Sociallearninginanimals.Therootsofculture.NewYork:AcademicPress.HumphreyNK(1976)Thesocialfunctionofintellect.In:Growingpointsinethology(BatesonPPG,HindeRA,ed),pp303-317.Cambridge:CambridgeUniversityPress.KrugerA,TomaselloM(1996)Culturallearningandlearningculture.In:Handbookofeducationandhumandevelopment:Newmodelsofteaching,learning,andschooling(OlsonD,ed),pp169-187.Oxford:Blackwell.KummerH,GoodallJ(1985)Conditionsofinnovativebehaviorinprimates.PhilosphicalTransactionsoftheRoyalSocietyB,308:203-214.LockeA(1978)Theemergenceoflanguage.In:Action,gesture,andsymbol:Theemergenceoflanguage(LockeA,ed),pp1-18.NewYork:AcademicPress.McGrewWC(1992)Chimpanzeematerialculture:Implicationsforhumanevolution.Cambridge:Cam•bridgeUniversityPress.MossCJ(1988)Elephantmemories.Boston:HoughtonMifflin.NagellK,OlguinK,TomaselloM(1993)Processesofsociallearninginthetooluseofchimpanzees(Pantroglodytes)andhumanchildren(Homosapiens).JournalofComparativePsychology,107:174-186.PackerC(1994)IntoAfrica.Chicago:UniversityofChicagoPress.PackerC,RuttanL(1988)Theevolutionofcooperativehunting.AmericanNaturalist,132:159-198.PovinelliDJ(1994)Comparativestudiesofanimalmentalstateattribution:AreplytoHeyes.AnimalBehaviour,48:239-241.PovinelliDJ,NelsonKE,BoysenST(1990)Inferencesaboutguessingandknowingbychimpanzees(Pantroglodytes).JournalofComparativePsychology,104:203-210.PovinelliD,PerillouxH,ReauxJ,BierschwaleD(1998)Youngchimpanzees"reactionstointentionalversusaccidentalandinadvertentactions.BehaviouralProcesses,42:205-218.PremackD(1986)Gavagai!Cambridge,MA:MITPress.PremackD,WoodruffG(1978)Doesthechimpanzeehaveatheoryofmind?BehavioralandBrainSciences,1:515-526.ReauxJ(1995).Explorationsofyoungchimpanzees"(Pantroglodytes)comprehensionofcause-effectrelationshipsintooluse.Mastersthesis,UniversityofSouthwesternLouisiana.Savage-RumbaughES,RumbaughDM,BoysenST(1978)Sarah"sproblemsincomprehension.Behav•ioralandBrainSciences,1:555-557.ThomasRK(1980)Evolutionofintelligence:Anapproachtoitsassessment.Brain,BehaviorandEvolu•tion,17:454-472.TomaselloM(1990)Culturaltransmissioninthetooluseandcommunicatorysignalingofchimpanzees?In:"Language"andintelligenceinprimates:Developmentalperspectives(ParkerS,GibsonK,ed),pp274-311.NewYork:CambridgeUniversityPress. TwoHypothesesAboutPrimateCognition183TomaselloM(1994)Thequestionofchimpanzeeculture.In:Chimpanzeecultures(WranghamRW,McGrewWC,deWaalFBM,HeltnePG,ed),pp301-317.Cambridge,Massachusetts:HarvardUniver•sityPress.TomaselloM(1996)Doapesape?In:Sociallearninginanimals:Therootsofculture(HeyesCM,GalefJr.BG,ed),pp319-346.NewYork:AcademicPress.TomaselloM,CallJ(1994)Socialcognitionofmonkeysandapes.YearbookofPhysicalAnthropology,37:273-305.TomaselloM,CallJ(1997)Primatecognition.Oxford:OxfordUniversityPress.TomaselloM,CallJ,NagellK,OlguinK,CarpenterM(1994)Thelearninganduseofgesturalsignalsbyyoungchimpanzees:Atrans-generationalstudy.Primates,35:137-154.TomaselloM,CallJ,WarrenJ,FrostT,CarpenterM,NagellK(1997)Theontogenyofchimpanzeegesturalsignals:Acomparisonacrossgroupsandgenerations.EvolutionofCommunication,1:223-253.TomaselloM,GeorgeB,KrugerA,FarrarJ,EvansE(1985)Thedevelopmentofgesturalcommunicationinyoungchimpanzees.JournalofHumanEvolution,14:175-186.TomaselloM,GustD,FrostGT(1989)Thedevelopmentofgesturalcommunicationinyoungchimpan•zees:Afollowup.Primates,30:35-50.TomaselloM,KrugerAC,RatnerHH(1993)Culturallearning.BehavioralandBrainSciences,16:495-552.VisalberghiE,LimongelliL(1996)Actingandunderstanding:Tooluserevisitedthroughthemindsofcapuchinmonkeys.In:Reachingintothought(RussonAE,BardKA,ParkerST,ed),pp57-79.Cam•bridge:CambridgeUniversityPress.WoodruffG,PremackD(1979)Intentionalcommunicationinthechimpanzee:Thedevelopmentofde•ception.Cognition,7:333-362.WranghamRW,McGrewWC,deWaalFBM,HeltnePG(1994)Chimpanzeecultures.Cambridge,Massachusetts:HarvardUniversityPress.ZabelCJ,GlickmanSE,FrankLG,WoodmanseeKB,KeppelG(1992)Coalitionformationinacolonyofprepubertalspottedhyenas.In:Coalitionsandalliancesinhumanandnonhumananimals(HarcourtAH,deWaalFBM,ed),pp113-135.NewYork:OxfordUniversityPress. 10CausalCognitionandGoal-DirectedActionAnthonyDickinsonandBernardW.BaUeineForstudentsoftherealevolutionofnaturalcognition,thevirtualevolutionofsyn•theticcognitionisinstructive.Inhisvideogame,Creatures,Grand(Grandetal.,1996)hascreatedsyntheticcreatures,Norns,thatrespectrealbiologicalprocesses.EachNornhasagenomeintheformofasingle,haploidchromosome,consistingof320interactinggenesthatarecrossedandsplicedatgeneboundariesduringrepro•duction.Thesegenescodenotforsuperficialcharacteristicsbutratherforunder•lyingphysiologicalandbiochemicalstructuresandprocesses,suchasthereactionsofmetabolism,hormonalsystems,immuneresponsestoenvironmentalantigens,andageing.Moreover,theirsurvivalinthepopulationofNornsisdeterminedbytheircontributiontoreproductivefitness.Nornsarealsoendowedwithsyntheticcognitioninthattheylearnabouttheirdemanding,ifvirtualenvironment.Theylearntoexploitresourcestoassuagetheirnutritionaldeficitsandstaveoffdeathbystarvation,toavoidpredationandinfec•tion,tocourtmates,andtoreproduce.Anditrequiresonlypassingfamiliaritywiththesecreaturestoexperiencetheirresistibledemandtoadoptan"intentionalstance"(Dennett,1987)totheirmorefloridemergentbehaviors—"cooperationinplayingwithaball,or"chase"scenesresultingfrom"unrequitedlove""(Grandetal.,1996)—letalonetheirimpressivecompetenceincontrollingtheirenvironmentinfaceofthevicissitudesoflife.Butthismanifestintentionalityisentirelyillusory:therearenorepresentationsofgoalsinthebrainsofNornsnoranyknowledgeofthecon•sequencesoftheiractions.TheNorn"sgenomealsocodesforthebasicstructureofitsbrain,whichconsistsoftwolobes.The"concept"lobecontains640neuronsthatreceive128sensoryinputsfromexternalstimuli,aswellasfrominternalonesgeneratedbytheNorn"sdrives.Thisnetworkofneuronsservestocreate"object"and"event"configuralneuronsforinputsthatco-occurbeforesupplyinganoutputtothe"decision"lobecontainingtheunitscontrollingtheNorn"sbehavior.Itistheplasticityofthesecon•nectionsbetweenthe"concept"and"decision"lobesthatenablesaNorntolearnandtherebytoadaptitsbehaviortoenvironmentalcontingencies.Importantly,thisplasticityiscontrolledbytheclassicstimulus-response(S-R)/reinforcementprocess:wheneveranoutputfromthe"concept"lobethatactivatesa"decision"neuronisfollowedbythedetectionofarewardstate,therespectiveconnectionisstrength•ened.Correspondingly,thedetectionofpunishmentweakenstheconnection.ThesereinforcementandpunishmentprocessesimplementThorndike"s(1911)positiveandnegative"LawsofEffect."ThespecificationofarewardingeventhasaprecedentinHull"s(1943)drive-reductiontheoryofreinforcement.EachNornis 186A.DickinsonandB.W.Balleineendowedwith13differentdrives,rangingfromhungerandpaintolonelinessandboredom,thatarecontinuouslymonitoredandadjustedinresponsetoenvironmen•talandphysiologicalinputs.ForaNorn,arewardissimplyaneventthatreducesacurrentlyactivedrive,whereaspunishmentarisesfromadriveincrease.AlthoughtheS-R/reinforcementprocessembodiedinaNornisasophisticatedimplementationofthestandardlearningalgorithmofneobehavioristpsychology,whatCreaturesillustratessoclearlyisthecapacityofsuchasimplemechanismofdocilitytoendowNornswiththeappearanceofbeingpurposiveandgoal-directedagentswhen,infact,theyarejusthabitmachinesperformingwhateverresponsehasbeenpreviouslyreinforcedinthepresenceofthecurrentstimulusinput.Theirbe•haviorisnotgovernedbyknowledgeofthepleasure,pain,boredom,orlonelinessproducedbydifferentactions,whateverthelabelsareattachedtotheirdrivestates.Allthatisnecessaryisthatfluctuationsoftheseinternalstatesprovidebothastim•ulusinputtoelicittheappropriateresponseandarewardorpunishmentsignaltochangethestrengthofS-Rconnections.Manyovertrained,well-practicedresponses,includingmuchofhumanbehavior,areprobablysimplehabitsmaintainedbytheS-R/reinforcementmechanism(e.g.,Dickinsonetal.,1995),and,indeed,theremaywellbemanyanimalsthatarenomorethanhabitmachineslikeNorns.Wealsoknow,however,thatsomerealcrea•turesarecapableoftruegoal-directedactions,actionsthatare,atthetimeoftheirexecution,controlledbytworepresentations:(1)arepresentationofthecausalnatureoftheinstrumentalrelationshipbetweenanactionanditsconsequenceoroutcome;and(2)arepresentationofthecurrentvalueoftheoutcome.Mediationbyanaction-outcomerepresentationwithcausalcontentisrequiredifwearetoregardtheactionasdirectedtowardtheoutcomeinanysensebeyondthepurelydescrip•tive.Furthermore,theactionmustbecontrolledatthetimeofitsexecutionbyarepresentationofthevalueoftheoutcomefortheagentifwearetoregardthebehaviorasdirectedtoagoal.Thus,underourpsychologicaldefinition,actionsmustbemediatedbyintentionalrepresentationstobetrulygoal-directedandpurposive.Inthischapter,wearguethatthehumancapacityforgoal-directedactionissharedwithotheranimals,specificallythehumblelaboratoryrat.Twomainsourcesofconvergingevidenceareofferedinsupportofthisthesis.Thefirstcomesfromstudiesofthesensitivityoftheinstrumentalbehaviorofratstoboththecurrentvalueofthegoalandthecausalrelationshipbetweenactionandoutcome;thesecondarisesfromthecompellingconcordancebetweenhumancausaljudgmentsandtherat"sgoal-directedactions. CausalCognitionandGoal-DirectedAction187GoalRevaluationThefunctionaladvantageconferredbygoal-directedactioncanbeillustratedbyconsideringthehypotheticalcaseofforagingbyanomnivorousanimal,suchasarat.Tosimplifythescenario,weassumethatthereareonlytwoavailablefoodsources,eachofwhichisdeficientinsomeessentialnutrientsbutwhichprovideabalanceddietwhenintermixed:oneisrichincarbohydratesandanotherinproteins.Accesstoeachsourcerequiresadifferent,learnedaction:forexample,huntinginthecaseofproteinsandgatheringinthecaseofcarbohydrates.Considernowaforagingepisodeinwhichtheratcomesacrossanunexpectedandabundantsourceofstarchonwhichitgorgestoanovelstateofcarbohydratesatiety.Followingthisepisodes,theforgingdecisionfacedbytheratiswhethernowtohuntortogather.Inastateofcarbohydratesatiety,asimple,functionalevaluationobviouslyfavorshuntingovergathering,andyetthisdecisionwoulddefeatahabit-boundNorn.Letusassumethatduringpreviousforagingepisodesbothactionshavebeenequallyreinforcedbytheirappropriatefoodrewardsinthepresenceofaninputfromthehungerdriveproducedbyadeficitinbothresources.Asaconsequence,thehunger-huntandhunger-gatherconnectionshaveequalstrength.Now,forthefirsttime,theNornexperiencesanovelstateofcarbohydratesatiety,astatethathasnopretrainedconnectionswitheitheraction,withtheresultthatthecreaturefindsitselfwithverylittleinclinationtoperformeitheractivityand,atbest,vacillatingbetweenhuntingandgathering.Intheabsenceofknowledgeofthecausalrelationshipbetweeneachactionandtheassociatedfood,theNorncannotchoosetohuntratherthangatheronthebasisofthefactthatproteinshouldnowhaveahighergoalvaluethanfurthercarbohydrateintake.BycontrastwiththeNorns,weknowthatratsdomaketheappropriatechoiceinthisforagingscenario.Forexample,we(BalleineandDickinson,1998a)maderatshungrybyrestrictingaccesstotheirgeneralmaintenancedietandthentrainedthemtopressaleverandpullachain—thelaboratoryanalogsofhuntingandgathering—withoneactionreinforcedbyprotein-richfoodpelletsandtheotherbyastarchso•lution.Thetwoactionsweretrainedinseparatesessionsandtheaction-foodassign•mentscounterbalancedacrossanimals.Althougheveryresponsewasreinforcedinitially,thereinforcementschedulewasgraduallychangedsothatattheendofin•strumentaltrainingeachresponsewasreinforcedwithaprobabilityof0.05.Inotherwords,onaverage,onlyeverytwentiethresponsewasrewardedwiththeappropriatefood,butthespecificnumberofresponsesrequiredforeachrewardvariedran•domly.Thisintermittentscheduleofreinforcementsimulatesthefactthatnotevery 188A.DickinsonandB.W.Balleineattemptathuntingorgatheringissuccessful.Wethensimplyprefedtheanimalsoneofthefoodstosatietyforonehourbeforegivingthemthechoicebetweenperform•ingthetwoactions.Duringthischoicetest,alleightratspreferredtheactiontrainedwiththenon-prefedfoodfromtheoutsetofthechoicetest.Itshouldbenotedthatnorewardsweregiveninthistest.Iftheactionsarerewardedduringsuchatest,anydifferenceinperformancecanreflectaselectivereductionofthereinforcementproducedbytheprefedfood.Intheabsenceofrewards,however,thechoiceofoneactionovertheothermustreflectknowledgeabouttherelationshipsbetweentheactionsandtheirassociatedfoodsacquiredduringtheinitialinstrumentaltrainingandthedeploy•mentofthisknowledgeinresponsetothedevaluationofoneoftheseoutcomesbyprefeeding.InstrumentalContingencyForaNorn,thecriticalfeatureofaninstrumentalcontingencyisthetemporalcon•tiguitybetweentheresponseandtheoutcome,forallthatisrequiredtostrengthentheresponseisthatitbefollowedbyaneffectivereinforcer.Asimplecontingentschedule(figure10.1A),suchasthatemployedinourspecificsatietyexperiment,confoundsthecausalrelationshipwithacontiguousone;notonlydoestheactioncausetheoutcome,butthetwoarealsopairedintime.Theclassicdemonstrationthatinstrumentalactionissensitivetocausalityunconfoundedbycontiguitycomesfromstudiesvaryingtheinstrumentalcontingencywhentheprobabilitythatanactionispairedwiththeoutcomeisheldconstant(Hammond,1980).Followingthespecificsatietytest,we(BalleineandDickinson,1998a)retrainedourratstoleverpressandchainpullinseparatesessionsonthescheduleinwhichtheactiondeliverstheappropriatefoodwithaprobabilityof0.05(figure10.1A).Thus,aspreviously,oneactionwaspairedwiththefoodpelletsandtheotheractionwiththestarchsolu•tion.Theinstrumentalcontingencywasthendegradedbypresentingoneofthefoodswiththesameprobabilitywithoutaresponse,sothattheseoutcomeswereunpairedwithbothresponses.Thus,foreachanimaltheunpairedoutcomewaseitherthefoodpelletsorthestarchsolution,andtheeffectoftheseunpairedoutcomesonper•formancecouldbecomparedwhentheoutcomepairedwiththeactionwaseitherthesameas(figure10.1B)ordifferentfromtheunpairedoutcome(figure10.1C).Theeffectofdegradingtheinstrumentalcontingencydependeduponwhetherthepairedandunpairedoutcomeswerethesameordifferent.Whenthepairedandunpairedfoodswereofthesametype(figure10.IB),thatis,bothwerepelletsor CausalCognitionandGoal-DirectedAction189Figure10.1Schematicrepresentationsofdifferentaction-outcomecontingencies.(A)Acontingentschedulewithnounpairedoutcomes;(B)anoncontingentscheduleinwhichtheunpairedoutcomesarethesameasthepairedoutcomes;(C)anoncontingentscheduleinwhichtheunpairedoutcomesaredifferentfromthepairedoutcomes;(D)anoncontingentscheduleinwhichtheunpairedoutcomesarethesameasthepairedoutcomebutprecededbyasignal.Inthefiguretheprobabilityofthepairedandunpairedoutcomesare0.5persec,whereasintheinstrumentalconditioningexperimentswithratstheprobabilitieswere0.05persec. 190A.DickinsonandB.W.Balleinebothwerethestarchsolution,theperformanceratesdeclined.Thisdeclineistobeexpectediftheratsaresensitivetonotjusttothecontiguousrelationshipbetweentheactionandoutcomebutalsotothecausaleffectoftheactiononthelikelihoodoftheoutcome.Underthisnoncontingentschedule,theactionhasnoeffectonthelikelihoodofanoutcomeastheprobabilitiesofpairedandunpairedoutcomesareidentical.Whetherornottheratpressestheleverorpullsthechain,anoutcomeoccursonaverageonceevery20seconds.Incontrasttotheactiontrainedwiththesamepairedandunpairedoutcomes,performancewassustainedwhenthepairedandunpairedfoodsweredifferent(figure10.1C;seealsoColwillandRescorla,1986).Alleightratsperformedtheactiontrainedwiththepairedfoodthatdifferedfromtheunpairedonemorethantheactiontrainedwiththesamepairedfood.Thesustainedperformancewithdifferentpairedandunpairedoutcomesprovidesfurtherevidencethattheratsaresensitivetothecausalrelationshipbetweenactionandoutcome.Whenthetwooutcomesarethesame,neitherthefrequencyoffoodnoritstypecanbecontrolledbytheratsandperformancedeclines.Bycontrast,thecompositionofthedietcanbealteredbyrespondingwhenthepairedandunpairedfoodsdiffer.Althoughtheoverallfre•quencyofthefoodisunaffectedbyresponding,higherratesofperformanceincreasetheproportionofthepairedfoodinthediet.Thefactthatouranimalssoughtvari•etybycontinuingtoperformtheactiontrainedwiththedifferentoutcomeindicatesthattheyaresensitivetothecausalrelationshipbetweenactionandanoutcomeandnotjusttotheircontiguity.Insummary,thespecificsatietytestdemonstratesthattheinstrumentalactionsoftheratarecontrolledbyarepresentationofthecurrentvalueoftheoutcomeandmediatedbyarepresentationoftheaction-outcomerelationship.Moreover,thefactthatratsperformaninstrumentalactionmorevigorouslywhentheycancontroleithertheabsoluteorrelativefrequenciesofaparticularoutcomeshowsthattheyaresensitivetothecausalpropertyofaninstrumentalcontingency.Takentogether,thesefindingssuggestthattheirinstrumentalbehaviorshouldberegardedasagoodcandidateforgoal-directedstatusunderapsychologicaldefinitionthatappealstointentionalrepresentations.Butwhethersuchbehaviorismorethanacandidateisimpossibletodeterminewithcertainty.Althoughthesensitivitytogoal-devaluationandtheinstrumentalcontingencyliesbeyondthescopeoftheS-R/reinforcementmechanismofNorns,thereareotherequallymechanisticsystemsthatmightexplainthegoal-directedcharacterofinstru•mentalactionwithoutappealingtointentionalrepresentations(Dickinson,1994).Theproblemis,ofcourse,that,otherthanspeechacts,behaviornevercarriesitsintentionalityonitssleeve,and,facedwiththisproblem,wehaveadoptedthestrat- CausalCognitionandGoal-DirectedAction191egyofseekingconvergingevidencetobolstertheattributionofcausalcontenttotherat"srepresentationofaninstrumentalcontingency.CausalJudgmentandInstrumentalActionOnetacticinthestrategyofbringingconvergingevidencetobearontherepresenta•tionalcontentmediatinginstrumentalactioncomparestheacquisitionofsuchactionsbyratswiththatofcausaljudgmentsinhumans.DickinsonandShanks(1995)documenttheconcordancebetweenthesetwoformsoflearning.Justastheacquisitionandterminallevelsofleverpressingbyratsisreducedbyimposingadelaybetweenthisactionandthedeliveryofafoodreward(e.g.,Dickinsonetal.,1992),soarejudgmentsofthecausaleffectivenessofasimilaractioninproducinganoutcomebyhumansubjects.Whenpeopleareaskedtoratethecausaleffectivenessofpressingthespacebaronacomputerkeyboardinproducingavisualdisplayonthemonitor,theirjudgmentsdeclinesystematicallywiththelengthofthedelaybe•tweentheactionandthisoutcome(e.g.,ShanksandDickinson,1991).Animalactionandhumancausaljudgmentsalsoexhibitcomparablesensitivitytothedegradationoftheinstrumentalcontingencybythedeliveryofunpairedout•comes.Recallthataninstrumentalactioncanberenderedcausallyineffectivebydeliveringunpairedoutcomeswiththesameprobabilityaspairedones(figure10.IB)andthattherateofleverpressingbyratsdeclinessystematicallyasthedifferenceintheprobabilityofthepairedandunpairedoutcomesisreduced.Thesameistrueofhumanperformancewhentheoutcomesaregivenanominalvalue(ShanksandDickinson,1991)and,moreimportantly,ofcausaljudgmentsoftheeffectivenessoftheiractions(e.g.,Wassermanetal.,1983;ShanksandDickinson,1991).Althoughtheconcordancebetweenanimalactionandhumancausaljudgmentsuggestsacommonacquisitionprocess,itisequallyexplicableintermsofindepen•dentbutconvergentprocesses.Withtheexceptionoftheaction-outcomedelay,theotherconcordancesarisefrombehavioralandcognitivepatternsthatapproximatetothenormativeadaptation.Anagentreallydoesexercisegreatercontrolovertheoutcomewhentheprobabilityofthepairedoneishigherthanthatoftheunpairedone.Consequently,evenifanimalactionandhumanjudgmentsaremediatedbydif•ferentpsychologicalprocesses,adaptiveselectionshouldensurethattheseprocessesconvergeonthesameoutputprofile.Aconvergenceaccountwarrantslesscredence,however,ifitcanbeshownthatanimalbehaviorandhumancognitionaresubjecttosimilarcausalillusions.Thereisnoreasonwhynaturalselectionshouldproduceconvergentmisrepresentation. 192A.DickinsonandB.W.BalleineDickinsonandShanks(1995)notethatsuchanillusioncanbegeneratedonanon-contingentschedule.DickinsonandCharnock(1985;seealsoHammondandWein•berg,1984)initiallytrainedratstoleverpressonthestandardcontingentschedule(figure10.1A)andthenswitchedtothenoncontingentscheduleinwhichthesameoutcomewasequallyprobableinanysecondwithandwithoutaresponse(figure10.IB).Wehavealreadynotedthatleverpressinghasnoeffectonthelikelihoodofanoutcomeonanoncontingentscheduleandleadstoalossofresponding.IntheCharnockandDickinsonstudy,however,eachpresentationofanunpairedoutcomewasprecededbyabriefvisualsignal(figure10.1C)foronegroupofrats.Thepres•enceofthissignalproducedamoresustainedlevelofrespondingcomparedtothatofanimalsinanothergroupforwhichthevisualstimuluswaspresentedwiththesameprobabilitybutrandomlyrelatedtothedeliveryofunpairedoutcomes.Althoughthissignalingeffectisanticipatedbycertaintheoriesoflearning(seeDickinsonandShanks,1995),theintuitiveexplanationisthatthesignalmarksthepresenceofapotentialcauseoftheunpairedoutcomes,therebydiscountingtheseoutcomesintheevaluationofthecontrolexertedbytheinstrumentalaction.Thiscontrolisillusory,however;ananimalhasnomorecontrolovertheoccurrenceofoutcomesunderthesignalconditionthanundertherandomone.Importantly,causaljudgmentsshowexactlythesamesignalingillusion.Shanks(1989)askedforratingsofthecausaleffectivenessofpressingthespacebarinproducingthevisualoutcomeunderanoncontingentschedule.Aconditioninwhichtheunpairedout•comeweresignaledconsistentlyyieldedhighercausalratingsthanoneinwhichthesignalstimulusoccurredrandomly.Thisbriefsurveyrevealsaconcordancebetweentheinstrumentalperformanceofratsandjudgmentsofthecausaleffectivenessofanactionbyhumans.Thisconcor•danceisobservednotonlyacrossmanipulationsthataffecttheobjectiverelationshipbetweenactionandoutcome,butalsounderconditionsthatinduceanillusionofcausalcontrol.Wetakethistobeatleastpresumptiveevidencethatgoal-directedactionsbyratsandcausalbeliefsbyhumansareacquiredbyacommonprocess.TheRepresentationofGoalsOurpsychologicaldefinitionofgoal-directedactionrequiresnotonlythattheactionbemediatedbyarepresentationofthecausalrelationshipbetweenactionandout•comeor,inotherwords,byacausalbelief,butalsobyarepresentationofthevalueoftheoutcome.Thisdefinitionis,ofcourse,nomorethantheclaimthatgoal-directedactionsasthosemediatedbytheinteractionofthebeliefsanddesiresoffolk CausalCognitionandGoal-DirectedAction193Figure10.2Schematicrepresentationoftheinteractionbetweenthecognitiveandmotivationalsystemsduringthedevaluationofafoodrewardbyincentivelearning.ThecontentsofthecognitivesystemillustrateinPROLOGthecausalbeliefanddesirefortheoutcomeacquiredduringinstrumentaltrainingandtheruleofpracticalinferencetogenerateanintentionfortheaction(seeHeyesandDickinson,1990),whereasthemotivationsystemcontainsanassociativeconnectionbetweenunitsactivatedbythefoodandgastricillnessacquiredduringaversiveconditioning.Presentationofthefoodoutcomeduringre-exposureleadstothecognizanceofthefoodatthesameasfeelingill,anexperiencethatremovesofthedesirefortheout•comefromthecognitivesystem.psychologythroughtheprocessofpracticalinference(HeyesandDickinson,1990).Cognitivetheoriesofthistype(e.g.,Tolman,1959;Bolles,1972)havealegacyincomparativepsychologythatisalmostasvenerableandenduringasthatoftheS-R/reinforcementtheoriesofneobehaviorism.FollowingHeyesandDickinson(1990),thecontentofaninstrumentalbeliefandadesireforfood(inthiscasecheese),andtheirinteractionthrougharuleofpracticalinferencetoderiveanintentiontoper•formtheinstrumentalaction,isillustratedintheCognitiveSystemoffigure10.2intheformofalittlePROLOGprogram.Wehavealreadyattemptedtoestablishgroundsforarguingthatleverpressingbyratscanbemediatedbyacausalbelief.WhatthePROLOGprogrammakesclear,however,isthatthisintentionalaccountofgoal-directedactionalsodependsupondemonstratingaroleforamotivationalrepresentationintheformofadesire. 194A.DickinsonandB.W.BalleineThemotivationalprocessesofNornsaresimpleandmechanisticinnature.Nutri•tionaldeficitsgiverisetoadrivestateofhungerthatthenprovidesaninputtotheneuralunitsofthe"conceptual"lobeandtherebyastimulusfortriggeringresponsesthathavebeenpreviouslyreinforcedwhenthecreaturewashungry.Sometheoristshaveattributedageneralinvigoratingandactivatingpropertytosuchdrivestates(e.g.,Hull,1943).Butwhateverthedetailsofadrivetheory,itisclearthatthepro•cessesinvokedbysuchatheorycanbedirectlyrealizedbyphysiologicalandneuralsystemsthatdetectthenutritionalstateofthebody;inprincipleatleast,thestimulusinputtothe"conceptual"lobecansimplybetheoutputofachemoreceptordetect•ing,forexample,thesugarconcentrationoftheblood.Bycontrast,theoriginofadesireisobscure,atleastadesirewiththerequisiterepresentationalpropertiestorationalizeanintentionalactionthroughaprocessofpracticalinference.Inthecaseofhumansocialandculturalgoals—prestige,justice,honor,andevenculturallydeterminedfoodpreferences—thereisnoproblem,atleastinprinciple,forthesedesirescanbeacquiredthroughtherepresentationalchannelsoflanguage.Butincaseofanimals,theexplanatorygulfbetweenadesireforagoalandasignalofitsbiologicalutilityisalarminglywide.Intheabsenceofaninterfacebetweentheseapparentlyincommensuratestatesandprocesses—oneintentionalincharacterandtheothermechanistic—thegroundingofdesiresinphysiologicallyrelevantprocessesfoundersandwithitacognitiveexplanationofgoal-directedaction.Ouranswertothisproblemisthatalldesiresandthereforegoalshavetobelearned,eventhosegroundedinthemostbasicbiologicalneeds.Toacquireadesireforcheese,wehavetoexperienceourownreactionstoamaturecheddarandlearnthroughexperiencehowthesereactionsareenhancedbygoodappetite(andapintofbeer)atlunchtimeorbythemildspecificsatietystateinducedbyadeliciouspud•ding(andsmoothclaret)atdinner.Werefertothisprocessasincentivelearning(DickinsonandBalleine,1994,1995).Accordingtoincentivelearningtheory,thephysiologicalstateinducedbyfooddeprivationdoesnotautomaticallyanddirectlygiverisetoadesireforaspecificfood,ortoanydesireforthatmatter;rather,thedesireisacquiredbyexperiencewithfoodinthisstate.Evidenceforincentivelearningcomesfromstudiesofshiftsinmotivationalstate.Considerthecaseofanondeprivedratthatistrainedtoleverpressforaparticular,mildlyattractivefoodandthentestedforthefirsttimewhenitisdeprivedofitsmaintenancediet,andishungry.Ifthenutritionaldeficitsinducedbythefooddep•rivationautomaticallyenhancethedesireforthefood,theratshouldpressmorevigorouslywhenhungrythanwhennondeprived.Surprisingly,nosuchenhancementisobservedunlesstheanimalhashadexperiencewiththefoodwhenhungry(Balleine, CausalCognitionandGoal-DirectedAction1951992),andthereforetheopportunitytolearnthatthefoodismoredesirablewhenhungrythanwhennondeprived.Wehavedemonstratedthatincentivelearningplaysaroleinthecontroloftheinstrumentalbehaviorofratsacrossavarietyofshiftsinmotivationalstate(seeDickinsonandBalleine,1994,forareview).Eventhedevalua•tionofafoodrewardbyspecificsatietyinvolvesincentivelearning.Prefeedingaspecificfoodtosatietydoesnotdirectlyreducethedesirabilityofthatfoodbutratherallowstheanimaltolearnthatthefoodislessattractivewhenintheparticularsatietystateinducedbyprolongedfeedingonthatveryfood(BalleineandDickinson,1998b).Theseincentivelearningstudiesdemonstratethatdesires,evenforcommoditiesofprimarybiologicalutility,aredisconnectedfromthephysiologicalmechanismsthatconferthisutility.Thecognitiveprocessescontrollinggoal-directedactionhavetobeinterfacedwiththesemechanismsthroughincentivelearningfordesirestoreflectthebasicneedsoftheanimal.Moreover,thisdisconnectionappliesnotonlytostatesinducedbynutritionaldeficits,butalsotononcognitiveformsoflearning.Forex•ample,ifaratbecomessickafteringestinganovelfoodordrink,itwillsubsequentlyavoidconsumingthatcommodity,aformofaversionlearningthatGarcia(1989)arguesisnoncognitiveorimplicitinnatureinthesensethatitdoesnotdependuponanintentionalrepresentationofthecausalorpredictiverelationshipbetweenthefoodandillness.Ratherthanshowinglearningthatthefoodpredictsillness,Garcia(1989)maintainsthattheaversionreflectsadirectchangeintheanimal"sreactionstothefooditselffromingestiveonestoreactionsreflectingdisgust,andinagreementwiththisaccountwehavefoundthatdevaluationofafoodrewardbyconditioninganaversionoftenrequiresincentivelearning.Itisworthconsideringgoaldevaluationbyaversionconditioninginmoredetailbecauseitprovidesamodelforanalyzingthemodulationofdesiresbyincentivelearning.We(BalleineandDickinson,1991)trainedratstopressaleverforanovelfoodinasinglesessionandthenimmediatelyinducedgastricillnessbyaninjectionoflithiumchloride.Thistreatmentisknowntoconditionanaversiontothefoodsothattheratswillnolongereatit.Thequestionatissue,however,waswhethertheaversiontreatmentalsodevaluesthefoodasagoalofleverpressing.Iftheaversionconditioningisanoncognitiveformoflearning,itshouldhavenodirectimpactonthedesireforthefoodandconsequentlyuponthesubsequentpropensitytoleverpress.Thisisexactlywhatwefound:whentestedintheabsenceofthefood,theseratspressedjustasmuchasanimalsthathadnotreceivedtheaversiontreatment.Toproduceadevaluationeffect,wehadtore-exposetheanimalstothefoodbetweenaversionconditioningandtesting.Figure10.2illustratestheprocessesthatweassumetakesplaceduringthisre-exposure.Asaresultoftheinstrumentaltraining,theratsacquireabeliefaboutthe 196A.DickinsonandB.W.Balleinecausaleffectivenessofleverpressinginproducingthefoodreward,inthiscasealumpofcheese.Moreover,byexperiencingthepleasureofcheeseduringtrainingtheyacquireadesireforit.Theresultoftheaversionconditioning,however,isnotrepresentedintentionally,butratherbringsaboutachangeinthemotivationalsystem.Thisisillustratedinfigure10.2bytheformationofanassociativeconnec•tionbetweenaunitactivatedbythefoodandonethatgeneratesthegastricillnessreactions.Intheabsenceofanyfurtherexperience,however,thecognitiveandmotivationalsystemscontinuetofunctioninisolationandtheaversionlatentinthemotivationalsystemhasnoimpactontheanimal"scognitiveevaluationofthefoodandthereforeonthecontrolofleverpressing.Re-exposuretothecheeseprovidesaninterfaceforthesetwosystems.Theratperceivesthecheeseintheformofanintentionalrepresentationwhileatthesametimedetectionofthisstimulusactivatestheillnessreactionsthroughtheassociativeconnectioninthemotivationalsystem.Thus,duringre-exposuretheratcognizesthecheeseatthesametimeasfeelingillwiththeconsequencethatthedesireforcheeseinremovedfromthecorpusofitsintentionalrepresentations.Weacknowledge,ofcourse,thatthisaccountisfartoobaroqueandelaborateanedificetostandonthefoundationsofincentivelearningalone.Theinterestingfeatureoftheaversionprocedure,however,isthatitprovidestheopportunitytestfurtherpredictionsofthedissociationbetweenthecognitiveandmotivationalsys•tems.Forexample,ifwecouldreducetheseverityofthegastricmalaiseduringre-exposureweshouldattenuatethedevaluationeffect.Thiswedidbyusingananti•emetic.Thedesignofthestudyisillustratedintable10.1(Balleineetal.,1995).Asinthepreviousexperiment,wegavetheratsasinglesessionofinstrumentaltraining,butinthiscasewetrainedthemtoperformtwoactions,leverpressingandchainpulling,fordifferentnovelrewards.Theanimalswerethirstyduringtraining,sotherewardswereasugarwaterandasaltsolution.Immediatelyfollowingthistrainingsession,allanimalswereinjectedwithlithiumchloridetoconditionaversionstoboththesugarandthesaline.Table10.1DesignofBalleine,Garner,andDickinson(1995)InstrumentalTrainingRe-exposureChoiceTest(operantchambers)(drinkingcages)(operantchambers)(Ai->Oi;A2->02)LiClOnd:d;Veh:02v;Ond:AivsA2A,leverpressingorchainpulling;O,sucrosesolutionorsalineoutcome;Ond,ondansetroninjection;Veh,vehicleinjection.LiCl,lithiumchlorideinjection. CausalCognitionandGoal-DirectedAction197Duringre-exposure,theratsweregivenaccesstobothsolutionsindrinkingcagesonseparatedays.Importantly,priortothere-exposuretoonesolution,weinjectedtheratswiththeanti-emeticondansetron,whichactsbyblockingtheserotonergicreceptorsinthebrain.Underthisregime,thesolutionre-exposedwithouttheanti•emeticisexperiencedinconjunctionwithnausea,which,accordingtoourincentivelearningaccount,producesalossofdesireforthiscommodity.Bycontrast,thenauseashouldbeattenuatedforthesolutionexperiencedunderondansetron,withtheresultthattheanimalsretainsagreaterdesireforthisreward.Thispredictionwasfulfilledinasubsequenttestofinstrumentalperformanceintheabsenceoftheoutcomes:theratsperformedtheactiontrainedwiththerewardre-exposedunderondansetronmorethantheactiontrainedwiththeoutcomere-experiencedwithouttheanti-emetic.Soitdoesappeartobetheaffectivereactionsexperiencedinconjunctionwitharewardthatdetermineitsdesirability.Thisbasicdesignalsoallowedustotestonefurtherandperhapsstrongerpredictionofthein•centivelearningaccount.Torecap,anessentialfeatureofthisaccountistheclaimthatdesiresareintentionalrepresentationsthathavenodirectchannelsofcommu•nicationwithbasicmotivationalmechanismsexceptthroughtheinterfaceoftheaffectivereactionsproducedbythesemechanismswhenactivatedbythegoalobject.Consequently,atthetimewhenanactionisperformedintheserviceofadesire,thestateofthemotivationalsystemshouldhavenoimpactuponperformanceofthataction,eventhoughitwasthissystemthatdeterminedthedesireatthetimeofin•centivelearning.Toexpressthisclaiminamoreconcreteform:toavoidperforminganactiontrainedwitharewardthattherathaspreviouslyexperiencedinconjunc•tionwithgastricmalaise,itdoesnotneedtofeelnauseousatthattimeofinstru•mentalperformance;theanimaljustknowsthatthisparticularoutcomeisnolongerdesirable.Wetestedthispredictionsimplybyattenuatinganynauseaatthetimeofinstru•mentaltestingwithondansetron.Ifsuchreactionsdirectlycontrolinstrumentalper•formance,theadministrationoftheanti-emeticshouldreducethedifferencebetweentheperformanceoftheactionstrainedwiththetwooutcomes,onere-exposedundertheondansetronandtheotherwithoutthedrug.Bycontrast,iftheassignmentoftherelativevaluesoftheoutcomesisfixedcognitivelyatthetimeofre-exposuretotheoutcomes,whetherornottheillnessreactionsareblockedduringtheinstrumentaltestshouldhavenoinfluenceonperformance.Ourtestresultsfavoredthislatteralternative.Therelativeperformanceofthetwoactionswasunaffectedbywhetherornottheratsweregiventheanti-emeticpriortotheinstrumentaltest.Thus,goal-directedactionsarecontrolledbydesiresforoutcomesthatarenotdirectlymodu•latedbytherevelantstatesofthemotivationalsystem. 198A.DickinsonandB.W.BalleineInsummary,wearguethatincentivelearningprovidestherequiredinterfacebe•tweentheintentionalprocessescontrollinggoal-directedactionandthemotivationalmechanismsthatgroundthevalueofprimarygoalsinbiologicallyrelevantvariables.Justascausalknowledgeabouttheeffectivenessofinstrumentalactionshastobelearned,sohavedesiresfortheoutcomesofthoseactions.Theoriginofbothinstru•mentalbeliefsanddesiresinexperienceaccordswiththeirrepresentationalstatus.AJustSoStoryofGoal-DirectedActionEverypsychologist—butperhapsnoteverybiologistwhoshouldknowbetter—canbeallowedjustoneJustSostoryabouttheevolutionoftheirfavoredbehavioralcompetence.Asourcredithasnotyetbeensquandered,inprintatleast,wetakethisopportunitytosquanderitnow.Ourstoryforintentionalcognitioningeneralandgoal-directedactioninparticularisasfollows.Whenneuralplasticityfirstevolvedtosupportlearning,itendowedanimalswithjustthosebehavioralcapacitiesthatweobservedinNorns.Inadditiontoinnateresponsesandtheirmodulationbyhabitu•ationandsensitization,theseanimalscouldalsolearntocontroltheirenvironmenttogainaccesstonecessaryresourcestosustainlifeandreproductionandtoavoidpredationanddamagebythesimpleS-R/reinforcementmechanisms.Amodelofsuchacreatureisthemarinesnail,Aplysiacalifornica,whoseconditioneddefensiveresponsesarethoughttobereinforcedbythepresynapticfacilitationofsensory-motorsynapsesbyaninputfromafacilitatingneuronactivatedbythereinforcer(Hawkinsetal.,1983).Andindeed,thisbasicmechanism,oronethatisfunctionallyverysimilar,hasbeenconservedinmorecomplexnervoussystemstoprovideabasisforsimplehabitlearning.Forexample,thegainofthespinalsensory-motorsyn•apsesmediatingthestretchreflexinprimatescanbemodulatedbyinstrumentalre•inforcement(CarpandWolpaw,1994).Aswehaverepeatedlypointedout,themainlimitationoftheS-R/reinforcementmechanismisitsinabilitytorespondtochangesintheutilityofreinforcerswithoutfurtherexperienceoftheinstrumentalcontingency.Asaconsequence,suchcreaturesareincapableofexploitingthebenefitsendowedbygoal-directedaction.Althoughattemptshavebeenmadetoconferatleasttheillusionofpurposivenessbyelabo•ratingthebasicS-R/reinforcementmechanismswithintheframeworkofboththeclassicneobehavioristtheory(e.g.,Hull,1952)andcontemporaryneuralcon-nectionism(e.g.,Donahoeetal.,1993),ourstrongclaimisthatthecapacityfortruegoal-directednesscannotbefoundeduponthisbasichabitmechanism.Purposiveactionrequiredtheevolutionofanentirelynewwayofcomputingbehavioralcon- CausalCognitionandGoal-DirectedAction199troluponneuralprocessingandplasticity,onethatsupportsintentionalrepresenta•tionsoftheinstrumentalcontingencyandgoalvalueandtheirinteractionthroughapracticalinferenceprocess.Inaccordwiththisevolutionarytrajectory,BalleineandDickinson(1998a)reportedthatsensitivitytotheinstrumentalcontingencyandgoalvaluedependsuponanintactprelimbicareaoftherat"sprefrontalcortex.Followinglesionsofthisarea,ratsareinsensitivetowhetherunpairedoutcomesarethesameasordifferentfromthepairedones;prefeedingdoesnothaveaselectiveimpactoftheresponsetrainedwiththeprefedoutcome.Infact,prelimbiclesionsappeartoconvertagoal-directedratintoanS-RNorn-likecreatureand,significantly,lesionsofthecorre•spondingareaofthehumancortexproducesprofounddeficitsintheplanningandexecutionofgoal-directedactions(e.g.,ShalliceandBurgess,1991).Wecharacterizethisinstrumentalrepresentationasencodingthecausalrelation•shipbetweenactionandoutcome;thisisnecessaryiftheresultingactionistobenotonlygeneratedbytheprocessofpracticalinference,butalsoarationalconsequenceofthisinference.Somemayobject,however,tothischaracterizationonthegroundsthatanycausalrepresentationworthitsnameshouldencodethespecificgenerativeprocess,beitmechanicalorsocial,bywhichanactioncausesitsoutcomeandshouldsupportinferencesandinsightsbasedonanunderstandingofthegenerativeprocess.Claimshavecertainlybeenmadeforcontent-specificcausalunderstandingbybothchimpanzees(seechaptersbyDunbarandbyRumbaughandHillix,butalsobyTomasello)andravens(seechapterbyHeinrich).Althoughnotdenyingaroleforcontent-specificmodelsinthecausalreasoning,atleastinthecaseofhumansandperhapsevengreatapesandcorvids,ourclaimisthatweandotheranimals,includ•ingtherat,canrepresentbasiccontent-freecausalpower.Althoughthisclaimraisesphilosophicalandpsychologicalissuesthatliewellbeyondthescopeofthepresentdiscussion,neverhaveanyofthenumeroushumansubjectsstudiedinourlabora•toryqueriedtherequesttoratethecausaleffectivenessoftheiractions,eveninthemostimpoverished,content-freeoperanttask,andintheabsenceofanyknowledgeoftheunderlyingcausalprocess.Whatwerefertoasacausalrepresentationissim•plytheknowledgethatinformedtheirjudgments.Ourclaimthenisthatananimalendowedwiththeappropriatecognitiveresourcescanrepresentasimplecontingencybetweeneventsascausal,aclaimthatliesfirmlywithintheHumeantradition.Moreover,justasPiagetsuggestedthatontogonyofcausalcognitionisfoundedontheexperienceoftheconsequencesofaction,sowearguethatthisformofcognitionevolvedintheserviceofgoal-directedaction.However,theburdenthatweplaceonourjustsostoryoftheevolutionofpurposive 200A.DickinsonandB.W.Balleineactionisnotsolelythatofreconfiguringneuralprocessingtosupportcausalcogni•tion,butalsothatofgeneratinganinterfacebetweenintentionalrepresentationsandtheneuralsignalsofmotivationallyrelevantphysiologicalstates.InthecaseofNornlikecreatures,motivationprovidesnoproblem,atleastinprinciple:theoutputofachemoreceptorsuppliesadirectinputtotheS-Randreinforcementmecha•nisms.Buthowsuchanoutputgivesrisetoadesireisfarfromclear.Ouransweristhatitcannotdirectly.Whatitcando,however,istogenerateahedonicoraffectiveexperience—pleasure,pain,disgust,andsuchlike—whichwhenexperiencedincon•junctionwiththeintentionalrepresentationoftheobjectoftheseaffectivereactionscreatesanappropriatedesire.Inotherwords,desiresaregroundedinouraffectivereactionstopotentialgoalobjects.Amodeloftheprocesswehaveinmindcanbeillustratedbytheorofacialreac•tionsofratstonoveltastingsolutions(GrillandBerridge,1985).Whenrats,likehumans,aregivensweetsolutionstodrinkforthefirsttime,theymakeanumberofingestiveconsummatoryresponses,suchasrhythmicmouthmovementsandtongueprotrusions.Bycontrast,bittersolutionselicitaversiveorrejectionreactions:headshakes,chinrubs,andfacewashing.Importantly,theseingestivereactionsaredirectlyresponsivetomotivationalmanipulations:theyareenhancedbyfooddepri•vation(Berridge,1991)andtransformedtoaversiveresponsesbytasteaversionconditioning(Berridgeetal.,1981).AlthoughtheseresponsesaredirectindicesofthehedonicevaluationofthetastesforGrillandBerridge(1985),weinterpretthemasnomorethanprecursorsofaffectthatshouldstilloccurinanimalsincapableofhedonicandaffectiveexperience.Therefore,Norn-likecreaturesshouldexhibitafullcomplementoftheappropriateconsummatoryreactions,butintheabsenceofanyaccompanyinghedonicoraffectiveexperience.AsMacphail(thisvolume)pointsoutwhendiscussingthe"functionoffeeling,"thereisnogoodreasonwhyaNornshouldexperienceorfeelatall;pleasure,pain,elation,andfearhavenofunctionalrolewithinthemechanismcontrollingitsadaptivebehavior.ItsS-Rhabitsensurethatitpressestheright"levers"andavoidstheright"stimuli"withouttheneedtofeelanyemotional,affective,orhedonicresponsetothereinforcersthatstrengthenthesehabits.Itisonlywiththeevolutionoftheintentionalcontrolofgoal-directedactionthatthereisafunctionforthefeelingsandaffectivereactionselicitedbymotivationallysignificantevents,thatofgroundingtheassignmentofvaluetotheoutcomesofactioninbiologicallyrelevantprocesses.Thus,theevolutionofthecognitivere•sourcesforgoal-directedactionhadtobeaccompaniedbytheco-evolutionofcapacityofconsummatoryreactions,ormorestrictlyneuralprocessescontrollingthem,tosupporttheappropriateaffectiveandhedonicexperiences.Wecannownolonger CausalCognitionandGoal-DirectedAction201postponelabelingtheinterfaceatwhichcognitionmeetsaffect:itissimple,non-reflexiveconsciousness.Underouraccount(BalleineandDickinson,1998c),thepri•maryfunctionofthisformofconsciousnessistoprovideaninterfacebetweenmotivationalandcognitiveprocessesthatallowsfortheconjointcognizanceofanintentionalgoalrepresentationandtheaffectiveexperienceproducedbythegoalobjectorstateandtherebyensuresthatdesiresaregroundedonbiologicallyimpor•tantstatesandevents.Ouraccountfortheevolution,mechanisms,andfunctionofaffectiveexperienceisclearlyakintothatsuggestedbyHumphrey(thisvolume)inhisanalysisofsensationanditsrelationtoperception.Humphreyarguesthatsensationarisesfromtheinter•nalizationofresponsestostimuliincomplexnervoussystemsthatwereoriginallydirectedtothesensoriuminsimpleanimals,andthatthis"privatization"ofsensoryresponsessupportsbothconscioussensoryexperienceandaninteractionwithcogni-tivelymediatedperception.Theorofacialresponsesoftherattogustatorystimuliprovideagoodmodelofsuchprivatizationinthattheseresponsearecontrolledbylowlevelsofsensoryprocessinginthebrainstem(GrillandBerridge,1985).Bycontrast,changesinthedesireforfoodsbroughtaboutbyspecificsatietyandincen•tivelearningappeartobeencodedatthehighestlevelofgustatoryprocessingintheinsularcortex(BalleineandDickinson,1998b),wheretheyarewellplacedtointer•actwithcausalbeliefsabouttheinstrumentalcontingencyencodedintheadjacentprefrontalcortex.Ouraccountoftheoriginsandfunctionofprimaryconsciousness,wehope,goessomewaytoansweringthefirstofMacphail"s(thisvolume)twoproblemsbyidenti•fyingafunctionforfeelingandaffect.Ourresponsetothesecond,"Whichanimalsareconscious?"isgrotesquelysimple.Inourpsychologicalclassificationofthefaunaofourbiospheretherearejusttwokindsofanimals:S-Rrobotsandcognitivecrea•turesthatarealsoendowedwithintentionalrepresentations,affectiveexperience,andtheabilitytointegratethetwoinconsciousness(thisisonebetterthansyntheticbiosphereswhere,tothebestofourknowledge,thereareonlyrobots,atleastasrealizedinsemicompletecreatures).Thetestsbywhichweassignmembershiptothetwoclassesisalsosimple.Tobeamemberofthecognitiveclass,theanimalsimplyhastobecapableoftruegoal-directedactionasassayedbysensitivitytogoalreval•uation,thecausalrelationshipbetweenactionandoutcome,andincentivelearning.Asassayedbythesecriteria,thepresentmembershipofthisclassisone:thelabora•toryrat.Weare,ofcourse,readytoadmittheprimatesbyproxy,andagoodcasecanalsobemadeforcorvids(Claytonetal.,thisvolume;Heinrich,thisvolume).Otherthanthese,tothebestofourknowledge,theelectioniswideopen. 202A.DickinsonandB.W.BalleineSummaryWehavearguedthatthegoal-directedactionsoftherataremediatedbyintentionalrepresentationsofthecausalrelationshipbetweenactionandoutcomeandofthevalueassignedtotheoutcome.Althoughthisthesisyieldsnocriticaltest,convergingevidencewasofferedinsupportoftheclaim.First,wedemonstratedthattheinstru•mentalactionsoftherataresensitivebothtothecurrentvalueoftheoutcomeandtothecausalrelationshipbetweentheactionandoutcomeinawaythatcannotbeexplainedbyS-R/reinforcementprocesses.Secondly,theconcordancebetweentheseactionsandhumancausaljudgmentssuggeststhatinstrumentalpeformanceisbaseduponacausalrepresentationoftheaction-outcomerelationship.Havingarguedthatatleastoneanimalrepresentsthecausalnatureofaction-outcomerelationship,wespeculatedthatthecapacityforsuchintentionalcontrolofbehaviorevolvedprimarilytosupportgoal-directedaction.Theevolutionofinten•tionalcontrolbringsinitswakeanumberofproblems,however,mostnotablythatofgroundingrepresentationsofthevalueofpotentialgoalsinbiologicallyrelevantstatesandvariables,thatis,inthebiologicalequivalentofthedrivestatesofNorns.Ouranswertothisproblemisthatthecapacityforgoal-directedactionrequiresnotonlytheevolutionofintentionalrepresentations,butalsotheco-evolutionofaninterfacebetweentheserepresentationsandtheanimal"sbiologicalresponsestothegoalobjects,events,orstates.Thisinterface,wesuggest,issimple,nonreflexivecon•sciousnessinwhichthebiologicalevaluationofapotentialgoalismanifestedasanaffectiveorhedonicresponseconjointlywithanintentionalrepresentationofthegoal.Theassignmentintentionalvalueisbaseduponthisconcurrentexperienceofthegoalrepresentationandtheassociatedaffect.ReferencesBalleineB(1992)Instrumentalperformancefollowingashiftinprimarymotivationdependsuponincen•tivelearning.JournalofExperimentalPsychology:AnimalBehaviorProcesses,18:236-250.BalleineB,DickinsonA(1991)Instrumentalperformancefollowingreinforcerdevaluationdependsuponincentivelearning.QuarterlyJournalofExperimentalPsychology,43:279-296.BalleineB,DickinsonA(1998a)Goal-directedinstrumentalaction:Contingencyandincentivelearningandtheircorticalsubstrates.Neuropharmacology,37:407-419.BalleineB,DickinsonA(1998b)Theroleofincentivelearningininstrumentaloutcomerevaluationbysensory-specificsatiety.AnimalLearningandBehavior,2:46-59.BalleineB,DickinsonA(1998c)Consciousness:Theinterfacebetweenaffectandcognition.In:Con•sciousnessandhumanidentity(CornwellJ,ed),pp57-85.Oxford:OxfordUniversityPress.BalleineB,GarnerC,DickinsonA(1995)Instrumentaloutcomedevaluationisattenuatedbytheanti•emeticondansetron.QuarterlyJournalofExperimentalPsychology,48B:235-251. CausalCognitionandGoal-DirectedAction203BerridgeKC(1991)Modulationoftasteaffectbyhunger,caloricsatiety,andsensory-specificsatietyintherat.Appetite,6:103-120.BerridgeKC,GrillHJ,NorgrenR(1981)Relationofconsummatoryresponsesandpreabsorptiveinsulinreleasetopayabilityandlearnedtasteaversions.JournalofComparativeandPhysiologicalPsychology,95:363-382.BollesRC(1972)Reinforcement,expectancy,andlearning.PsychologicalReview,79:394-409.CarpJS,WolpawJR(1994)MotoneuronplasticityunderlyingoperantlyconditioneddecreaseinprimateH-reflex.JournalofNeurophysiology,72:431-442.ColwillRM,RescorlaRA(1986)Associativestructuresininstrumentallearning.In:Thepsychologyoflearningandmotivation(BowerGH,ed),pp55-104.Orlando,FL:AcademicPress.DennettD(1987)Theintentionalstance.Cambridge,MA:MITPress.DickinsonA(1994)Instrumentalconditioning.In:Animalcognitionandlearning(MackintoshNJ,ed),pp45-79.London:AcademicPress.DickinsonA,BalleineB(1994)Motivationalcontrolofgoal-directedaction.AnimalLearningandBe•havior,22:1-18.DickinsonA,BalleineB(1995)Motivationalcontrolofinstrumentalaction.CurrentDirectionsinPsy•chologicalScience,4:162-167.DickinsonA,BalleineB,WattA,GonzalezF,BoakesRA(1995)Motivationalcontrolafterextendedinstrumentaltraining.AnimalLearningandBehavior,23:197-206.DickinsonA,CharnockDJ(1985)Contingencyeffectswithmaintainedinstrumentalreinforcement.QuarterlyJournalofExperimentalPsychology,37B:397-416.DickinsonA,ShanksD(1995)Instrumentalactionandcausalrepresentation.In:Causalcognition:Amultidisciplinarydebate(SperberD,PremackD,PremackAJ,ed),pp5-24.Oxford:ClarendonPress.DickinsonA,WattA,GriffithsWJH(1992)Free-operantacquisitionwithdelayedreinforcement.Quar•terlyJournalofExperimentalPsychology,45B:241-258.DonahoeJW,BurgosJE,PalmerDC(1993)Aselectionistapproachtoreinforcement.JournaloftheEx•perimentalAnalysisofBehavior,60:17-40.GarciaJ(1989)FoodforTolman:Cognitionandcathexisinconcert.In:Aversion,avoidanceandanxiety(ArcherT,NilssonL-G,ed),pp45-85.Hillsdale,NJ:Erlbaum.GrandS,CliffD,MalhotraA(1996).Creatures:Artificiallifeautonomoussoftwareagentsforhomeentertainment.TechnicalReport,UniversityofSussex.GrillHJ,BerridgeKC(1985)Tastereactivityasameasureoftheneuralcontrolofpalatability.In:Prog•ressinpsychobiologyandphysiologicalpsychology(SpragueJM,EpsteinAN,ed),pp1-61.SanDiego:AcademicPress.HammondLJ(1980)Theeffectsofcontingenciesuponappetitiveconditioningoffree-operantbehavior.JournaloftheExperimentalAnalysisofBehavior,34:297-304.HammondLJ,WeinbergM(1984)Signalingunearnedreinforcersremovessuppressionproducedbyazerocorrelationinanoperantparadigm.AnimalLearningandBehavior,12:371-374.HawkinsRD,AbramsTW,CarewTJ,KandelER(1983)AcellularmechanismofclassicalconditioninginAplysia:Activity-dependentamplificationofpresynapticfacilitation.Science,219:400-405.HeyesC,DickinsonA(1990)Theintentionalityofanimalaction.MindandLanguage,5:87-104.HullCL(1943)Principlesofbehavior.NewYork:Appleton-Century-Crofts.HullCL(1952)Abehaviorsystem.NewHaven,CT:YaleUniversityPress.ShalliceT,BurgessW(1991)Deficitsinstrategyapplicationfollowingfrontallobedamageinman.Brain,114:727-741.ShanksDR(1989)Selectionalprocessesincausalityjudgment.MemoryandCognition,17:27-34. 204A.DickinsonandB.W.BalleineShanksDR,DickinsonA(1991)Instrumentaljudgmentandperformanceundervariationsinaction-outcomecontingencyandcontiguity.MemoryandCognition,19:353-360.ThorndikeEL(1911)Animalintelligence:Experimentalstudies.NewYork:Macmillan.TolmanEC(1959)Principlesofpurposivebehavior.In:Psychology:Astudyofascience(KochS,ed),pp92-157.NewYork:McGraw-Hill.WassermanEA,ChatloshDL,NeunaberDJ(1983)Perceptionofcausalrelationsinhumans:Factorsaffectingjudgmentsofresponse-outcomecontingenciesunderfree-operantprocedures.LearningandMotivation,14:406-432. 11CausalReasoning,MentalRehearsal,andtheEvolutionofPrimateCognitionRobinI.M.DunbarDuringthelastdecade,muchhasbeenmadeoftheMachiavellianintelligence(orsocialbrain)hypothesis,whichclaimsthatprimateshavecognitiveabilitiesthatoutstripthoseofnonprimatesinoneparticularrespect,theabilitytohandleacom•plexsocialworld(ByrneandWhiten,1988;WhitenandByrne,1997;Brothers,1990).Wedonothaveanyrealunderstandingofthecognitivemechanismsthatareinvolvedinthis;indeed,wedonotevenknowforsurewhetherthedifferencesbetweenprimatesandnonprimatesarequalitativeorsimplyquantitative.However,considerableinteresthasfocusedontheroleofabilitiesliketheoryofmind(ToM)asakeyfeatureofsocialcognition(seeforexampleByrne,1995;TomaselloandCall,1997),atleastinsofarasitseemstodemarcatehominoids(humansandgreatapes)fromotherspecies(Dunbar,1998).Inthispaper,Ifirstreviewtheevidenceforthesocialbrainhypothesisandthenconsidertwootherrespectsinwhichthecognitiveabilitiesofprimatesmightdifferfromthoseofnonprimates(andmoreparticularlythoseinwhichhominoidsmightdifferfromotherprimates).Thesearetheabilitytoreasoncausallyandtheabilitytoengageinmentalrehearsal.Causalreasoningmustbeoneofthefundamentalbasesonwhichallcognitiveprocessesoperate.Dickinson(1980;DickinsonandBalleine,thisvolume)arguesthatinferentialreasoningcanbeshownexperimentallyinrats,probablybecausedeclara•tivecodingformstheverybasisonwhichananimal"sknowledgeabouttheworldisorganized(seealsoRumbaughandHillix,thisvolume).Beyondthis,mentalrehearsal(theabilitytopracticeorthinkthroughaprobleminthemind)providesacognitivemechanismbywhichtheimpactandsuccessofalternativestrategiescanbeevaluatedsothatthebehavioraltacticsusedtoimplementthechosenstrategycanbeexecutedmosteffectively.Thesetwocomponentscanbeseenasbeingespeciallyimportantinthewaythathumansgoabouttheireverydaybusiness,especially(butnotexclusively)inthesocialworld.Theyunderlietheprocessesofeverydaysocialinteraction,ourreflectionsonthenatureoftheworldinwhichwelive,andourattemptstoproduceculture.Howdootherspeciesofprimatescomparewithusintheseterms?TheSocialBrainHypothesisThegrowingconcensusofthepastdecadehasbeenthatprimatebrainevolutionhasbeendrivenprincipallybythedemandsofthesocialworldratherthanthedemandsofthephysical/environmentalworld(Brothers,1990;Byrne,1995;Barton 206RobinI.M.DunbarFigure11.1Regressionlinesformeansocialgroupsizeplottedagainstmeanneocortexratio(volumeofneocortexdividedbyvolumeoftherestofthebrain),forvariousmammalianorders.Eachpointisthemeanforasinglegenus.Theaxesarelog10-transformed.Thecarnivoreswithlargeneocorticesandgroupsofsizeonearemainlybears.andDunbar,1997;Dunbar,1998).Thisisnottosuggestthatremodelingofbrainpartstocopewithspecificphysicalworldproblemshasnothappened,butrathertopointoutthatprimatebrainsizehasincreasedacrossspecieslargelybecauseofchangesinthesizeoftheneocortex,andthatchangesinthesizeoftheneocortexcorrelatewithchangesinsocialvariablessuchasgroupsizeratherthanwithanyecologicalvariables(Dunbar,1992,1998).Inaddition,anumberofotherstudieshavedemonstratedthatneocortexsizeinprimatescorrelateswithatleastmeasuresofsocialskills:matingbehavior(Pawlowskietal,1998),groomingcliquesize(KudoandDunbar,inprep.),andsocialplay(Lewis,subm.).Twootherimportantobservationsariseoutofthiswork.Oneisthatsome(butnotall)othermammalslieonthesamegradientasprimatesintermsoftherela•tionshipbetweensocialgroupsizeandneocortexsize(figure11.1).Thisisespeciallytrueofcarnivores(DunbarandBever,1998),butmayalsobetrueofthecetaceans(Marino,1996;Tschudin,1998).Carnivoresdifferfromprimatesonlyinsofarastheydonotextendsofaralongthisgradient:primatessimplyhavebiggerbrains(neocortices)andbiggergroupsthanthosecharacteristicofcarnivores.Atthesametime,itisworthemphasizingherethatbothprosimianprimatesandinsectivoresseemtolieonamore"primitive"gradientcomparedtoanthropoidprimatesand CausalReasoning,MentalRehearsal,andPrimateCognition207Figure11.2Volumeoftheprimaryvisualarea(areaVI)plottedagainstthevolumeoftherestoftheneocortex(non-VI),forindividualanthropoidprimatespecies.Axesplotlogio-transformedvalues.(RedrawnfromJoffeandDunbar1997)carnivores,inthesensethattheirdistributionslietotheleftofthesetwoorders.Inthisrespect,prosimiansseemtosharewithinsectivoressomeprimitivetraitsreflect•ingareducedlevelofsociality.Thesecondpointconcernstheamountofneocortexvolumedevotedtovisualprocessing.Visionisclearlyveryimportanttothedailylivesofprimates;indeed,asagroup,theirlivesmaybesaidtobedominatedbyvision.Thisraisesanimportantquestion:istheincreaseinbrainsizesimplyduetotheneedtohavemoreaccurateandrefinedvisualprocessing,ordoesitreflecttheamountofwhatwemightrefertoastertiaryprocessing(thinkingabouttheoutputofthevisualsystem"sanalysesofprimaryvisualinput)?Unfortunately,thedatabasescurrentlyavailable(principallyStephanetal.,1981)donotdistinguishthesectorsoftheprimateneocortexinanygreatdetail.However,theydoprovidedataonthesizeoftheprimary(VI)visualareaforasampleofspecies.ThishasallowedustoexaminetherelationshipbetweenVIandtherestoftheneocortex(JoffeandDunbar,1997).Figure11.2showsthatthesizeofVIandtherestoftheneocortexarecorrelated.Visualandnon-Vlneocortexvolumeincreaseswithincreasingbrainsize.Inpartthisreflectsthefactthatbrainsizeincreaseswithincreasingbodysize,probablybecauseonlylarge-bodiedanimalsareabletosparesufficientenergytosupport 208RobinI.M.Dunbarlargerbrains.However,therelationshipbetweenthetwocomponentsisnotlinear:VIappearstoreachanasymptoticsizewiththelargestbodiedprimates(greatapesandhumans),whereastherestoftheneocortexcontinuestoincreaseinsize.ThisissignificantinviewofthefactthatitispreciselytheseareasintheprefrontalcortexthathaverecentlybeenshownbyPETscanstudiestobeactiveduringtheprocessingofToMtasksinhumansubjects(Frith,1996).JoffeandDunbar(1997)interpretthisresultasimplyingthat,afteracertainpoint,furtherincreasesinvisualacuityarenotgainedbycontinuedincreasesintheamountofcorticalprocessingareadevotedtoprimaryvisualanalysis.Thereafter,itpaysspeciestodevotewhateveradditionalgainstheymayhaveintotalbraincapacity(asaresultofincreasingbodysize)tothenonvisualareasofthecortex.Ofcourse,itisimpossibletobesurethatnoneoftheseareasplayaroleinvisualprocessing(weareunabletoexcludetheV2visualarea,forexample),butwecanatleastsaythat,whateverrolevisualprocessingplaysatthislevel,itisnotconcernedwithprimaryvisualanalysis.Itmayinvolveinterpretationofvisualpatterns,butnotthebasicprocessesofpatternrecognitionperse.Itis,Ithink,significantthatthepointatwhichthedistributioninfigure11.2startstoleveloffispreciselythatwhichcorrespondsintermsofbodysizetothegreatapes.Apesthusseemtomarkapointatwhichasignificantchangeoccursintheamountofcorticalprocessingcapacityavailableforhigherorderprocesses.Thisissignificantfortworeasons.First,ifwecarefullyexaminethedataonsocialgroupsizeversusneocortexsize,wediscoverthattherearefurthergradestobeidentifiedevenwithintheprimates:theapesseemtolieonthesamegradientasthemonkeys(andthecar•nivores)butaremovedfurthertotheright(Dunbar,1993).Humansappeartolieontheapelineratherthanthemonkeyline(Dunbar,1993).Inotherwords,apesseemtoneedmorecomputingpowertosupportagroupofagivensizethanmonkeysdo.Onereasonforthismightbethatapesengageinsomekindsof(social?)cognitionthatarenotfoundinmonkeys,andthattheseprocessesarecomputationallymuchmorecostlythanthestandardmechanismsusedbymonkeysandothermammals.Thesecondreasonwhythisfindingissignificantconcernspreciselythismecha•nism.Apesappeartobetheonlyspecies,asidefromhumans,forwhichthereisevencircumstantialevidencetosuggestthattheycanpassafalsebelieftask(Povinellietal.,1990;Povinelli,1994;O"Connell,1995).Afalsebelieftaskrequiresthesubjecttounderstandthatanotherindividualholdsabeliefthatthesubjectknows(oratleastbelieves)tobefalse.Aclassictestofthisusedonchildrenistheso-calledSmartieTest.(SmartiesarecandiessimilartoM&Msandsoldinsmalltubesapproximatelyfiveincheslong.)AchildisshownaSmartietubeandaskedwhatitcontains;everychildwillnaturallyrespondthatit CausalReasoning,MentalRehearsal,andPrimateCognition209containsSmarties.Thecapisthenremovedfromthetubeandthechildshownthatitactuallycontainspencils.Thechildisthenasked:ifyourbestfriendcomesinnow,whatwillhe/shethinkisinthetube?Childrenwhocandifferentiatetheirknowledgeoftheworld(thetubecontainspencils)fromtheknowledgeofanother(naive)indi•vidual(thetubecontainsSmarties)aredeemedtobedemonstratingthatheorsheisawarethatothersmayhavecontradictorybeliefsabouttheworld.Atthisstage,thechildissaidtohaveatheoryofmind(ToM).ChildrenarenotbornwithToM,butdevelopitmoreorlessspontaneouslyataboutfouryearsofage.Moreimportantly,perhaps,childrencontinuetodevelopmoresophisticatedversionsofthisabilitywithtime,suchthathumanadultscanaspiretoatleastfour(exceptionallyperhapssix)levelsofbelief-statereflexivity(Kindermanetal.,1998).Inotherwords,whereachildoffivecanfigureoutthat/believethatyouthinksomething"sthecase(twoordersofreflexivity),adultscanaspiretofour(IbelievethatyouthinkthatIwantyoutounderstandthatsomething"sthecase),andoccasionallymore.Thesearenowcommonlyreferredtoaslevelsofintentionality,followingDennett(1983).Althoughtheevidenceisstillequivocal(seeHeyes,1993,1998),thereisatleastprimafacieevidencetosuggestthatgreatapes(butprobablynotmonkeys)cansolvefalsebelieftasks(i.e.,aspiretosecondorderintentionality).Povinelli(Povinellietal.,1990,1991)hasshownthatgreatapes(butnotmonkeys)appeartobeabletodistinguishbetweenknowledgeandignoranceinhumantrainers.Inamoredirectattempttotestfalsebelief,O"Connell(1995)usedamechanicalanalogofafalsebelieftaskandfoundthatchimpanzeesperformedaboutaswellasfour-and-a-halfyearoldchildren(i.e.,thosethathavejustacquiredToM)andsignificantlybetterthanautisticadults(whoconventionallyfailfalsebelieftasks,eventhoughofnormalIQ)(figure11.3).TheoryofmindmaythusprovideakindofRubiconthatseparatestheapes(andhumans)offfromotherprimates.However,wearealongwayfromconclusivelyprovingthatevenapeshaveToM(Heyes,1998;TomaselloandCall,1997).CausalReasoningGiventheprecedingresults,thequestionarisesastowhethertheapesexhibitanyevidenceforotherkindsofspecialcognitiveabilitiesnotfoundinotherspecies.OverandabovetheargumentsdeployedbyDickinson(1980;DickinsonandBalleine,thisvolume)andothers(e.g.,RumbaughandHillix,thisvolume)forthespecies-generalityofinferentialreasoning,causalreasoningwouldseemtobeanimportant 210RobinI.M.DunbarFigure11.3Percentcorrectresponsesonthemechanicalanaloguetoafalsebelieftaskbynormalchildren,autisticadultsandchimpanzees.Thehorizontallineindicatesthelevelofrandomresponses(25percentcorrect).(redrawnfromO"Connell1995)processunderpinningthementalworldofanycognitivelyadvancedspecies.Hence,wemightreasonablyexpecttofindevidencefortheseabilitiestobewidespread,atleastamongthehighervertebrates.Nonetheless,VisalberghiandLimongelli(1994)haveshownthatcapuchinmonkeys,atleast,arepoorperformersontasksthatrequirecausalreasoning,whereasapesperformquitecompetently(Limongellietal.,1995).Onthebasisoftheseresults,theyconcludethatmonkeysprobablydonot,ingeneral,exhibitcompetencesinthisdomain,whereasapesdo.Thisapparentconflictofviewsmayreflectdifferencesinthekindsoftasksinves•tigated.VisalberghiandLimongelli(1994)testedmonkeysonataskthatrequiredthemtoreasonaboutphysicalrelationships(inferringthattheuseofarodtoreacharewarditemfromoneendofatubemightresultintherewardfallingintoatrap,whereasinsertedintotheotherenditwouldresultintherewardbecomingacces•sible).Kohler"s(1925)classicexperiments,ofcourse,suggestedthatchimpanzeescouldsolveproblemsofthisgeneralkind,sotheVisalberghi-Limongellifindingsmightsuggestanimportantdichotomybetweenapesandmonkeys.Incontrast,Dickinson(1980)considersproblemsmoredirectlybiologicalinnature—theabilitytoinferthatifanactionmakesoneill,thenastimuluspredicting CausalReasoning,MentalRehearsal,andPrimateCognition211aneventforwhichtheactionwouldbeappropriateshouldleadtoareductioninthefrequencywithwhichthatactionoccurs.IfAimpliesactionBandBimpliesC(becomingill),thendon"tdoBwhenAoccurs.Ratswereapparentlyabletosolveproblemsofthiskind.Dickinson"sintrepretationofclassicalconditioningexperi•ments,combinedwiththeresultsobtainedbyDomjamandWilson(1972)thatimplythatratshaveakindoftheoryofbiology,mightlendcredencetosomekindofdomain-specificreasoningability.Inotherwords,certainkindsofbiologicallyvalentphenomenamakeintuitivecausalsense,eventomammalsashumbleastherat,whereasotherkindsof(lessfamiliar?)physicalphenomenadonot.Oneimplicationoftheseresultsisclearlythatpresentinganimalswithpurelyphysicalproblems(asVisalberghiandLimongellidid)willleadtonegativeresults,whereaspresentingthemwithlogicallysimilarproblemsinabiologicallymoremeaningfuldomainwillresultinpositiveresults.Cheneyetal.(1995)carriedouttestsofbaboons"abilitytomakecausalinferencesinthesocialdomain(inferringwhowasthreateningwhomonthebasisofvocalizationsplayedbackfromahiddenspeaker).Theirfindingthatbaboonsdowellonthiskindoftasksupportsthesuggestionthattheremaybedomain-specificreasoningabilities.Wemightinferfromthesefindingsthatcausalreasoninginthebiologicaldomainisuniversallycharacteristicofmammals(andperhapsbirds),whereascausalreasoninginthesocialdomainmightbecharacteristicofprimates(orperhapsonlyapes),butnotothermammals.Thismightthuspromptustoask:(1)whetherapesdifferfrommonkeysinthisrespect,and(2)whetherprimatesingeneral(andapesinparticular)areanybetteratcausalreasoningtaskswhentheseareinthesocialdomainasopposedtothephysi•caldomain.Inanattempttoexplorethesequestions,O"Connell(1995)developedaseriesofexperimentaltestsbasedonLeslie"s(1982;LeslieandKeeble,1987)nowclassicstudyofcausalityinyounginfants.Lesliewasconcernedwiththequestionofwhethercausalityisanaturalcategoryinherentlypresentinyounginfantsbeforetheyhavehadanopportunitytolearnaboutsuchthingseitherthroughlanguageorbytrial-and-errorexperience.Leslie"stestinvolvedanhabituation-dishabituationformatinwhichfilmoftwocirclesmovingintoandoutofcontactwasplayedtoinfantsasyoungas27weeksofage.Inthetrainingtrials,onecirclemovesacrosstotouchtheothercircle,whichthenmovesawayasthoughpushed.Inthesubsequenttesttrial,thefilmisthesameexceptthatthecirclesdonotactuallymeet.Anunderstandingofcausalitywasinferredif,havinghabituatedtothetrainingfilm,theinfantsoncemoreattendedtothetestfilm,asthoughsurprisedbyan"unnatural"event.Itis,ofcourse,possibletoquibblewithLeslie"sconceptionofcausalityhere,aswellaswithhisexperimental 212RobinI.M.Dunbardesign.However,giventhedifficultyofdoingexperimentstotestsuchconcepts,ouraimshouldperhapsbealessambitiousone:ifweacceptLeslie"scriterionasatestofsomethinginhumans,wecanatleastaskwhethermonkeysandapesareabletoperformtothesamestandard.Ifthistestsanykindofcausalunderstandinginchildren,thenwehavetoascribethesamecognitiveabilitiestoanyprimatesthatcanpassthetest.WecanworryaboutjustwhatLesliewasactuallytestinglater.Totestprimatesonthistask,threeseparatesetsofvideoclipswereprepared.SeriesIwasidenticaltoLeslie"soriginalfilm.SeriesIIinvolvedasocialinteractionbetweenhumans:personAissittingonachaireatingabanana;personBsitsdownnexttopersonAandpushespersonAoffthechair,thentakesthebanana.Inthetestsequence,personAfallsofftheseatwithoutbeingtouchedbypersonBandthebananamovesacrosstopersonBwithoutbeingtouched.SeriesIIIinvolvedthehuntingsequencefromtheBBCfilmTooCloseforComfort,inwhichagroupoftheTaichimpanzeeschase,capture,anddismemberaredcolobusmonkey;inthetestsequence,thefilmwasplayedbackwardssothatthemonkeywasreassembledandescapedbackwardsfromitspursuers.Eachofthefilmclipswas20secondslong(30secondsinthecaseofSeriesIII),andwasrepeatedoverandoveragainthroughouta5-minutetestperiod.Subjectsweretestedindividuallyontheirown.Halftheanimalswereshownthenormal(contact)sequencefirst,andhalftheunnatural(nocontact)sequencefirst.Wherepossible,thoseanimalsthatwereshownthenatural(contact)sequenceofSeriesIfirstwereshowntheunnatural(nocontact)sequenceofSeriesIIfirst,andviceversa.Eachtrialbeganwhenthesubjectwaslookingatthevideoscreeninitscage.Thesubject"sbehaviorwasrecordedonvideo.Ineachseries,animalswereallowedtoviewthetrainingsequenceinsetsofthreerepetitions.Ameantimeoflookingatthescreenwasdeterminedforthefirstthreerepetitions,andthenthesequencereplayedinsetsofthreeuntilthesubjecthadproducedasetofthreelook•ingtimesthatwerelowerthanthemeanofthefirstthree.Thiswastakentoindicatehabituation.Thesecondtapeinthesameserieswasthenplayedthreetimesandtheamountoftimespentlookingatthescreenduringeachwasmeasured.SeriesIandIIweretestedon11commonchimpanzees,twobonobos(pygmychimpanzees),andfivespidermonkeyshousedatTwycrossZoo,EastMidlands(England).SeriesIIIwasshownonlytothechimpanzeesbecauseneitherbonobosnorspidermonkeysareknowntohuntinthewild.Forthechimpanzeesandbonobos,therewasnosignificantdifferencebetweenthedurationoflookingonthefirstorlasthabituationtrialsforthosethatsawthefilmsequencesintheordercontact/no-contactcomparedtothereverseonallthreeseries(ANOVAs,P>0.05).However,forbothsequences,thedurationoflookingwas CausalReasoning,MentalRehearsal,andPrimateCognition213Table11.1Durationofgazefixationonscreeninthehabituation-dishabituationexperimentoncausalityMeanDurationofLooking(sec)FilmOrder"TrialbChimpanzeesBonobosSpidermonkeys1C-NCHab7.06.09.3Dishab23.335.514.5N3241NC-CHab5.82Dishab8.31N612C-NCHab10.05.0Dishab35.726.0N312NC-CHab4.66.05.3Dishab6.458.88.3N5143C-NCHab17.5-Dishab56.0-N4-3NC-CHab11.6-Dishab43.4-N5-Listedvaluesarethedurationofthelasthabituationtrialandthefirstdishabituationtrial(seetextfordetails).aC,contact;NC,nocontactbHab,lasthabituationtrial;Dishab,firstdishabituationtrial;N,numberofsubjectssignificantlyloweronthelasthabituationtrialthanonthefirstone,demonstratinghabituation(SeriesI:Fhl6=21m,P<0.001;SeriesII:FhU=30.25,P<0.001;SeriesIII:Fl>u=6.71,P<0.025).Thus,thesubjectscanbeconsideredtohavehabituatedto"thetrainingsequences.FollowingLeslie(1982),thecrucialcomparisonisthefirstdishabituationtrialinthetestserieswiththelasthabituationtrialinthetrainingseries.Theresultsaresummarizedintable11.1.Again,withtheexceptionofSeriesI,therewerenodif•ferencesamongsubjectsshownthefilmsindifferentorders.Butfortwoofthethreeseries,thedurationoflookingduringthefirstdishabituationtrialwassignificantlylongerthanthatforthelasthabituationtrial(matched-pairsf-tests:SeriesI,tu=2.74,P=0.021;SeriesII,t9=2.07,P=0.068;SeriesIII,h=5.17,P=0.001;Fisher"sprocedureforcombiningprobabilitiesfromindependenttests:%2=26.92,df=6,P<0.001).ThemarginallysignificantresultforSeriesIIwasdueentirelytothepoorrateofresponseontheno-contact/contactsequence;threeofthefouranimalsshownthecontactfilmfirsthadlongerlookingdurationsduringthedishabituation 214RobinI.M.Dunbarthanduringthehabituationperiod.Thus,overall,thechimpanzeesshowedconvinc•ingevidenceofanunderstandingofcausality,atleastinsofarasLeslie"stestcanbeconsideredevidenceforthatinhumans.Incontrast,thespidermonkeysshowedverypoorresponserates.OnlythreeofthefivemonkeyshadlongerlookingtimesinthefirstdishabituationtrialonSeriesI(t-j=1.40,P=0.204),andonlytwoofthethreemonkeysshownSeriesIIhadlongerlookingtimesonthedishabituationtrial(t2=1.44,P=0.286).Theseresultsdonotshowanyevidenceofanunderstandingoftheunderlyingcausalityinthesesequences(Fisher"sprocedureforcombiningprobabilitiesfromindependenttests:%2=5.68,df=4,P>0.1).Insummary,weseemtohavesomeevidenceforanintuitiveappreciationofcau•salityinthechimpanzees,butnotinthemonkeys(atleastinsofarasNewWorldspidermonkeysarerepresentativeofthemonkeysasawhole),justasVisalberghiandcolleaguesclaimed.NoteonceagainthatSeriesIisaphysicalproblem,aswiththeVisalberghi-Limongellitests.Note,however,thatthechimpanzeesseemedtorespondsignificantlymorestronglytoSeriesIII(chimpanzeesocialevents)thantoSeriesI(physicalevents).AlthoughSeriesIIisnotionallysocial,inretrospecttheuseofhumansubjectsmakesitdifficulttointerpretitssignificancefromthemonkeys"pointofview:didtheyseethisasasocialproblemoraphysicalproblem?Thefactthatthechimpanzeesthemselvesrespondedonlyweaklytothisseries(theirresultswerebarelysignificant)lendssupporttothis.Perhapsthisparticularexperimentneedsrepeatingusingfootageofmonkeysifwearetohaveafairtestofmonkeys"abilitiesinthesocialdomain.MentalRehearsalIdeemmentalrehearsaltobeimportantbecauseitseemstounderliesomuchofwhathumansdo.Humansseemtospendagreatdealoftimemullingoverpossiblefutureactions,practicingthemmentallyorevaluatingthepossibleoutcomesofalternativeoptions.Wehavethereforeattemptedtodeterminewhethertheabilitytorehearsesomethingmentallyhasanyinfluenceonitsperformance.ThistestwasbasedonaseriesoffourpuzzleboxesofvaryingcomplexitythatwerepresentedtotwochimpanzeesatTwycrossbyO"Connell(1995)andtotwochimpanzeesattheWelshMountainZoo,sixorangutansatChesterZoo,and61primaryschoolchil•drenagedthreetosevenyearsbyMcAdam(1996).Thefourpuzzleboxesweregradedincomplexitybythenumberofoperationsthathadtobeexecutedtoobtainthefoodrewardinside(table11.2).Ineachcase,two CausalReasoning,MentalRehearsal,andPrimateCognition215Table11.2NumberofoperationsrequiredtoopenthepuzzleboxesusedtotestmentalrehearsalNumberofBoxoperationsOperationsA2Undocatch,openhatchB3Identifylargerhatch,opencatch,openhatchC4Recognizethathatchisbehindperspexpartition,liftpartition,opencatch,openhatchD5Recognizethatpartitionblocksaccesstoreward,liftpartition,tipboxwithpartitionliftedsorewardslidestofarend,undocatch,opendoorboxeswerepresentedtoeachsubjectwithoutpriorsight;theothertwowereleftoutsidetheanimal"scagefor24-48hoursbeforebeinggiventotheanimal,duringwhichtimeitcouldobservebutnottouchthebox.Inthecaseofthechildren,theconstraintsimposedbyclasstimemeantthattheprocessofmentalrehearsalhadtobespeededup.Ratherthanbeingallowedtoseetheboxesfor24hours,thechildrenwereallowedtostudytheboxesfromadistancefor20minutesbeforebeinggiventhem.Toensurethattheyengagedinmentalrehearsalduringthisperiod,thechil•drenwereaskedtodrawtheboxes.Thetimeeachsubjecttooktoopentheboxfromthemomentitwashandedtheboxwasrecorded.Theorderofpresentationofboxesandtheirappearanceintheprior-view/no-viewconditionswascounterbalancedasfaraspossible.However,foreachsubject,thetwopriorviewconditionswerealwayspresentedlast.Unfortunately,thetwochimpsattheWelshMountainZoodeclinedtotakepartintheexperiment,sothechimpanzeesampleisreducedtotwoanimals.Becauseoftheexperimentaldesign,bothoftheseanimalsreceivedtheboxesinthesameorderandunderthesameconditions.Wefirstusedthechildrentoprovideacomplexityrankingfortheboxesbasedonthetimetakentosolvethepuzzlesforthefirst(nopriorview)boxtheywerepresentedwith.Accordingtothiscriterion,theboxeswererankedintheorderATwounknownsources(2)>Onetrustedsource(T)>Oneunknownsource(1).Thedifferencebetweentheversionswashighlysignificant(ANOVA,df=3,SS=8.79,F=8.96,p=.00003).Amultiplecomparisonofmeansshowedthatthemajordifferencewasbetweentheeyewitnessaccountandotheraccounts,whichwerenotstatisticallydifferentfromeachother.Whentheprofessorwasbeingcaricatured,subjectsregardedthestoryasmoderatelycrediblebutwerecompletelyinsensitivetothequalityofinformation,withnosignificantdifferencesamongthefourversions.Allpaper-and-penciltestsinpsychologicalresearchmustbeinterpretedwithcautionandcross-checkedwithothermethodstoconfirmthattheycorrespondtobehaviorintherealworld.Ourresultsarepreliminary,buttheysuggestthatpeoplerespondtofictionalgossipeventsmuchastheywouldrespondtogossipintheirownlives(e.g.,ourcattlerancherstudycompareswellwiththebehaviorofactualcattleFigure19.3Averagesubjectratingforcredibilityoftheinformationforfourversionsoftwoepisodes;oneinwhichitisimportanttoknowthefactsofthematter(Cheating)andoneinwhichitisnot(Pants).Thefourversionsareeye-witnessaccount(blackhistogram),twohearsayaccounts(moderatelyshadedhistogram),onehearsayaccountfromatrustedperson(darklyshadedhistogram),andonehearsayaccount(lightlyshadedhistogram).Maleandfemalescoreswerenotstatisticallysignificantandarecombinedforanalysis. GossipandLanguageasGroup-LevelAdaptations363ranchersasreportedbyEllickson,1991).Ifso,thenitwillbepossibletomakerapidempiricalprogresstowardunderstandingthecontoursofgossipinparticularandlanguageingeneralfromamultilevelperspective.SummaryWehavetriedtoprovidebothaconceptualframeworkandanempiricalmethodol•ogyforstudyinggossip.Ourstartingpointwasthelongstandingconflictbetweengroupandindividual-levelperspectivesthatexists,independently,inbiologyandthesocialsciences.Weshowedthateffectivesocialcontrolmechanismsallowgroupstoevolveintoadaptiveunitsandthatgossiphasbeenreportedtofunctionasasocialcontrolmechanisminculturesaroundtheworld.Methodologically,weexploredtheuseofsimplepaper-and-pencilteststhathavebeenwidelyusedtostudyothersubjectsinpsychologybutnotgossip.Weshowedthatpeopleareeasilyengagedbyfictionalgossipepisodeswhoseelementscanbesystematicallyvaried.Theyhighlydisapproveofself-servinggossipbutapproveofgossipinginresponsetonormvio•lations,aslongasthegossipisconductedinaresponsiblemanner.Indeed,insomecontexts,failingtogossipcanbemorehazardousthangossipingtoone"sreputation.Theresultsofourexperimentsagreewithdescriptiveaccountsofgossiparoundtheworldandshowhowgossipcanfunctionasasocialcontrolmechanismwithoutdamagingthereputationofthegossiper.Ourstudyofgossipledtothemoregeneralstudyoflanguagefromamultilevelperspective.Languageisamongthemostcommunalofhumanfaculties,yettheindividualisticperspectivedominantinthestudyofhumanevolutionhasretardedthestudyoflanguageassomethingthatevolvedtobenefitwholegroups.Weboldlyaskedwhatlanguagewouldlooklikeasthemediumofagroupmindandpredictedcontextdependentsensitivitytothequalityofinformation.Ourpredictionwassup•portedbyasingleexperiment.Obviously,wehaveonlyscratchedthesurfaceofaverylargesetofsubjects,butperhapssufficientlytoshowthatourconceptualframework,coupledwithoursimplemethods,offersafertileinterplaybetweenhypothesisformationandtesting.ReferencesAbrahamsRD(1970)Aperformance-centeredapproachtogossip.Man,5:290-301.BarkowJH(1992)Beneathnewcultureisoldpsychology.In:Theadaptedmind:Evolutionarypsychologyandtheevolutionofculture(BarkowJH,CosmidesL,ToobyJ,ed),pp627-638.Oxford:OxfordUni•versityPress. 364DavidSloanWilsonetal.BesnierN(1989)InformationwithholdingasamanipulativeandcollusivestrategyinNukulaelaegossip.LanguageinSociety,18:315-341.BesnierN(1990)Conflictmanagement,gossip,andaffectivemeaningonNukulaelae.In:Disentangling:ConflictdiscourseinPacificsocieties(Watson-GegeoKA,WhiteGA,ed),pp290-334.Stanford,CA:StanfordUniversityPress.BleekW(1976)Witchcraft,gossip,anddeath:Asocialdrama.Man,11:526-541.BoydR,RichersonPJ(1992)Punishmentallowstheevolutionofcooperation(oranythingelse)insizablegroups.EthologyandSociobiology,13:171-195.BussDM(1994)Theevolutionofdesire.NewYork:BasicBooks.CosmidesL,ToobyJ(1992)Cognitiveadaptationsforsocialexchange.In:Theadaptedmind(BarkowJ,CosmidesL,ToobyJ,ed),pp163-225.NewYork:AcademicPress.CoxBA(1970)WhatisHopigossipabout?InformationmanagementandHopifactions.Man,5:88-98.deRaadB,CaljeH(1990)Personalityinthecontextofconversation:Person-talkscenariosreplicated.EuropeanJournalofPersonality,4:19-36.DunbarRIM(1996)Grooming,gossipandtheevolutionoflanguage.Cambridge,MA:HarvardUniver•sityPress.DunbarRIM,DuncanNDC,MarriottA(1997)Humanconversationalbehavior.HumanNature,8:231-246.EderD,EnkeJL(1991)Thestructureofgossip:Opportunitiesandconstraintsoncollectiveexpressionamongadolescents.AmericanSociologicalReview,56:494-508.EllicksonRC(1991)Orderwithoutlaw.Cambridge,MA:HarvardUniversityPress.GluckmanM(1963)Gossipandscandal.CurrentAnthropology,4:307-316.GluckmanM(1968)Psychological,sociologicalandanthropologicalexplanationsofwitchcraftandgos•sip:Aclarification.Man,3:20-34.GoodmanRF,Ben-Ze"evA(1994)Goodgossip.Lawrence:UniversityofKansasPress.HamiltonWD(1964)Thegeneticalevolutionofsocialbehavior:IandII.JournalofTheoreticalBiology,7:1-52.HandelmanD(1973)Gossipinencounters:Thetransmissionofinformationinaboundedsocialsetting.Man,8:210-227.HarringtonCL,BielbyDD(1995)Wheredidyouhearthat?Technologyandthesocialorganizationofgossip.SociologicalQuarterly,36:607-628.HavilandJB(1977)Gossip,reputationandknowledgeinZinacantan.Chicago:ChicagoUniversityPress.LevinJ,ArlukeA(1994)Gossip:Theinsidescoop.NewYork:PlenumPress.MaynardSmithJ,SzathmaryE(1995)Themajortransitionsinevolution.Oxford:W.H.Freeman.MerrySE(1984)Rethinkinggossipandscandal.In:Towardageneraltheoryofsocialcontrol:Funda•mentals(BlackD,ed),pp271-302.Orlando,FL.:AcademicPress.NevoO,NevoB,Derech-ZehaviA(1993)Thedevelopmentofthetendencytogossipquestionnaire:ConstructandconcurrentvalidationforasampleofIsraelicollegestudents.EducationalandPsychologi•calMeasurement,53:973-981.PaineR(1967)Whatisgossipabout?Analternativehypothesis.Man,2:278-285.PaineR(1968)Gossipandtransaction.Man,3:305-308.PaineR(1970a)Lappishdecisions,partnerships,informationmanagement,andsanctions—anomadicpastoralistadaptation.Ethnology,9:52-67.PaineR(1970b)Informalcommunicationandinformationmanagement.CanadianReviewofSociologyandAnthropology,7:172-188. GossipandLanguageasGroup-LevelAdaptations365PowdermakerH(1933)LifeinLesu:ThestudyofaMelanesiansocietyinNewIreland.ForewardbyDr.ClarkWissler.NewYork:Norton.RobertsJM(1964)Theself-managementofcultures.In:Explorationsofculturalanthropology:EssaysinhonorofGeorgePeterMurdock(GoodenoughWH,ed),pp433-454.NewYork:McGraw-Hill.SeeleyT(1995)Thewisdomofthehive.Cambridge,MA:HarvardUniversityPress.SilvermanI,PhillipsK(1998)Theevolutionarypsychologyofspatialsexdifferences.In:Handbookofevolutionarypsychology(CrawfordC,KrebsDL,ed),pp595-612.Mahwah,NJ:Erlbaum.SoberE,WilsonDS(1998)Untoothers:Theevolutionandpsychologyofunselfishbehavior.Cambridge,MA:HarvardUniversityPress.WalkerCJ,BlaineB(1991)Thevirulenceofdreadrumors:Afieldexperiment.LanguageandCommuni•cation,11:291-297.WilsonDS(1990)Speciesofthought:Acommentonevolutionaryepistemology.BiologyandPhilosophy,5:37-62.WilsonDS(1995)Languageasacommunityofinteractingbeliefsystems:Acasestudyinvolvingconducttowardsselfandothers.BiologyandPhilosophy,10:77-97.WilsonDS(1997)Incorporatinggroupselectionintotheadaptationistprogram:Acasestudyinvolvinghumandecisionmaking.In:Evolutionarysocialpsychology(SimpsonJ,KendrickD,ed),pp345-386.Mahwah,NJ:Erlbaum.WilsonDS,NearD,MillerRR(1996)Machiavellianism:Asynthesisoftheevolutionaryandpsychologi•calliteratures.PsychologicalBulletin,199:285-299.WilsonDS,NearDC,MillerRR(1998)IndividualdifferencesinMachiavellianismasamixofcoopera•tiveandexploitativestrategies.EvolutionandHumanBehavior,19:203-212.WilsonPJ(1974)Filcherofgoodnames:Anenquiryintoanthropologyandgossip.Man,9:93-102. ContributorsBernardW.BalleineD.P.GriffithsDepartmentofPsychologySectionofNeurobiology,PhysiologyUniversityofCalifornia,LosAngelesandBehaviorLosAngeles,CaliforniaUniversityofCalifornia,DavisPatrickBatesonDavis,CaliforniaSub-DepartmentofAnimalBehaviourBerndHeinrichUniversityofCambridgeDepartmentofBiologyCambridge,EnglandUniversityofVermontMichaelJ.BeranBurlington,VermontLanguageResearchCenterCeciliaHeyesGeorgiaStateUniversityDepartmentofPsychologyAtlanta,GeorgiaUniversityCollegeLondonM.E.BittermanLondon,EnglandBekesyLaboratoryofNeurobiologyWilliamA.HillixUniversityofHawaiiLanguageResearchCenterHonolulu,HawaiiGeorgiaStateUniversityRobertBoydAtlanta,GeorgiaDepartmentofAnthropologyLudwigHuberUniversityofCalifornia,LosAngelesInstituteofZoologyLosAngeles,CaliforniaUniversityofViennaNicolaS.ClaytonVienna,AustriaSectionofNeurobiology,PhysiologyNicholasHumphreyandBehaviorCentreforPhilosophyofNaturalandUniversityofCalifornia,DavisSocialSciencesDavis,CaliforniaLondonSchoolofEconomicsJuanD.DeliusLondon,EnglandDepartmentofPsychologyMasakoJitsumoriUniversityofKonstanzDepartmentofCognitiveandKonstanz,GermanyInformationSciencesAnthonyDickinsonChibaUniversityDepartmentofExperimentalChiba,JapanPsychologyLouisLefebvreUniversityofCambridgeDepartmentofBiologyCambridge,EnglandMcGillUniversityRobinI.M.DunbarMontreal,Quebec,CanadaESRCResearchCentreinEconomicNicholasJ.MackintoshLearningandSocialEvolutionDepartmentofExperimentalSchoolofBiologicalSciencesPsychologyUniversityofLiverpoolUniversityofCambridgeLiverpool,EnglandCambridge,England 368ContributorsEuanM.MacphailDavidSloanWilsonDepartmentofPsychologyDepartmentofBiologicalSciencesUniversityofYorkBinghamtonUniversityYork,EnglandBinghamton,NewYorkPeterJ.RichersonDepartmentofEnvironmentalScienceandPolicyUniversityofCalifornia,DavisDavis,CaliforniaDuaneM.RumbaughLanguageResearchCenterGeorgiaStateUniversityAtlanta,GeorgiaSaraJ.ShettleworthDepartmentofPsychologyUniversityofTorontoToronto,Ontario,CanadaMartinaSiemannDepartmentofPsychologyUniversityofKonstanzKonstanz,GermanyKimSterelnyDepartmentofPhilosophyVictoriaUniversityofWellingtonWellington,NewZealandMichaelTomaselloMaxPlanckInstituteforEvolutionaryAnthropologyLeipzig,GermanyAlexandraWellsDepartmentofBiologicalSciencesBinghamtonUniversityBinghamton,NewYorkLauraWeiserDepartmentofBiologicalSciencesBinghamtonUniversityBinghamton,NewYorkCarolynWilczynskiDepartmentofBiologicalSciencesBinghamtonUniversityBinghamton,NewYork SubjectIndexAbstraction,33,104,123-124,128,134-136Brain,28-29,38-39,43,46-47,53,57-58,71,85,Adaptation87-88,117,201,205-206,208,218,234,256,altruism,348-351260-261,302,331,333,336-338,340,343grouplevel,355anatomy,333modificationofbehavior,31-32size,117,206-207,234,291,329,331,339-340,specialization,19,46-47,65,174,277,319,341344value,48,248,255-256Analogy,34,38,88,246,321,338Cacheretrieval(recovery),17,277,279,281-283,Animatebeings,168,170-171285Anthropocentrism,3,44Categorization,3,7,33,89,103-104,105,118,Apelanguageproject,229123-128,134-135,144,148,154,171,284Artificialgrammar,135-136,259conceptformation,124Association,11,18,28,32,53,64-65,68-69,89,natural,125,211104-105,108,118-119,128,149,222,224,relational,138,167-168,171-172,181258-259,261,266,316,319-320,340Causality,3,10-11,89,98,154,168,170-171,formation,258-259,261,266188,211,214,218,221-223,225-226,229,learning,4-6,8-10,12,14,18,32,43-44,46-48,232,235-23650,52-57,76,85,128,134-137,169,260-261,cognition,10,171,185,200,236264,266,269,337detector,222-223,229memory,32effectiveness,191-192,196,199process,9,97,258,300inference,211,221system,137,259-261judgment,186,191,202Attention,53,56,61,69-70,89,103,145,148,reasoning,199,205,209,211151-155,158,224,226,233,242,320relationship,137,143,154-155,168,186-188,selective,103190,192,199,201-202,221-223,225-226,visual,145,148,151-153,155,158228,232,235-236Aversion,32,47,54,65,195-196,200understanding,144,173,199,212,221,223,229,Awareness,4,20,232,259,262,265,269,289,355231-232,236-237Cenozoic,329,331-332,341BehaviorCentralnervoussystem(CNS),4,20adaptation,9,12-16,18,20,171Cheating,349-350,356,362competence,117,159-160,198Classification,6,31,35,89-91,96,105,201complexity,257Coalition,165ecology,26,34Cognitionflexibility,291,314,331,333adaptation,16,18,171,174,179-180,329,331homology,35complex,151,315,329,331-332,,333335-336,instrumental,72,76,175,186,190,195341,343,344learned,44,298ethology,26,34pattern,28,32,36,103,150evolution,38,44-45,47,57-58,61,76,143,predisposition,32180product,10-11human,24-25,154,168,179-180,191,333,344program,28,32,35map,44reading,143-145,151module,10,18,38,45-48,52-53,254social,86,98,151,157,169,320natural,185Behaviorism,4,27,138niche,335BehindtheMirror,29,31,61-62,64organization,45-46,57Belief,98,144-146,151,155,169,174,192-195,primate,165,180,205201,208-209,223,228,242,262,301,355process,4-9,11-13,15-17,19-20,39,43,75,Biology,3,23,25-26,30,58,167,198,245,300-195,201,259301,329,338,339,348-349,354,363science,4,29,39,47comparative,300self,264,266-268continuity,236skill,56,165-167,171,181need,194,242social,177,205,208smartness,234spatial,48,50Blocking,33,69,71,260system,18,29,193,329,337,344 380SubjectIndexCommunication,48,144,146,176-178,197,222,Convergence,34,58,66,191226,232-234,236,321,353Cooperation,151,185,236Comparativestudy,26,44-45,47Correspondence,23,27,29,37,105experiment,66,72,317Cortex,29,199,208Competition,70,97,144,151,173,291,295,320,frontal,50321,348insular,201Comprehension,229,266prefrontal,96,118,199,201,208Concept,34-35,97,103-104,118,124,127-128,visual,117131,138,146,156,168,185,212,236,264,VI,207-208295,297,319,324,338,354V2,208abstract,128,138Counting,43complexsocial,168Cross-modalmatching,38conceptualization,14,105,118,126-128Culture,3,18-19,48,175,178-179,205,245,303,formation,43,103,118,124311,329,333-334,336,339,342-343,347ofself,147,264-265,267,269chimpanzee,44,174-175,311thought,6,10,33,138complex,329,336,342-343Conditioning,8,31-33,64,67-70,76,105,107,evolution,178,181,335195-196,200,223,259-260,263,295,298human,178,234,331,335-336,341,352,363avoidance,64inheritance,174,180classical,33,53-54,68,89,221,225,230,274innovation,179,311conditionedassociation,149learning,174,179,322conditionedreaction,350tradition,178-179,333,342,344conditionedresponse(CR),64,198,221,225,transmission,18,175,177,311,314,321,323-256,260,275325conditionedstimulus(CS),55,63,198,222,225,296Darwinian,12,15,25,97,241-242,244-245,250,instrumental,99,118,263301,331observational,8paradigm,25operant,32,222,228,230,232Deadreckoning,6,46,50-52,55-56,137Pavlovian,8-9,63,68Deception,150-151,154,169second-order,33,68Decisionmaking,6,57,135,336unconditionedstimulus(US),55,63,221Desire,192-198,200-201,223Connectionistnetwork,124Devaluationofafoodreward,195Consciousness,3-4,6,10,26,30,53,201-202,Discrimination241,250,253-263,265,267,269,274-276,conditionalodor,275289,300-304novelty-familiarity,130animal,3,286,289spontaneous,70autonoetic,224,266,275,285-286trailing,107memory,264visual,88,90,94,125,127noetic,274-275Dishabituation,96,212-214recall,263-266,273-274sensation,241,250Ecology,14,34,47,58,290,316Constraint,26,36-39,45,86,215,339,344ecologicalcognitivist,25developmental,36,332ecologicallearningtheorist,26onlearning,12ecologicalniche,34,38Constructivism,27Ektostriatum,118Contiguity,63,68-69,89,98,108,188,190EmergenceContingencyabilities,223,229,233backward,89behavior,185,223betweenevents,98,158,199emergents,223effectof,33property,24,300instrumental,188-191,198-199,201transitions,23Continuity,5,14,30,35,137-138,154,172,236,Emotion,145,257,301265Emulationlearning,175,177-179 SubjectIndex381Encephalization,331,340-341,343information,23,27,32,35Environmentrelatedness,350,352change,26-27,338transmission,321,323deterioration,329Gestaltperception,33heterogeneity,144Gesturalsignaling,176-177informationallytranslucent,149Goal-directedactions,185-187,192,194-195,leaking,161197-202,233,236predictor,319-320Goal-directedbehavior,4,301shaping,175Gossip,347-350,352-361,363social,14,149,290,354EpistemologyHabit,31,64,186-187,198,200,245evolutionary,26Habituation,31,33,43,,64,95-96,198,211-214,naturalized,26318Equivalenceclass,106-107,113-119,127Hippocampus,11,17,46,260-261,313-314,319Ethology,12,25-26,34-35Holocene,330,333Evaluation,72,187,192,196,200,202,357Homeostasis,31,39EvolutionHomology,35,337analysis,5,10,16,35,43Howrule,6-14,17-20,39.SeealsoWhat/whenbiology,338-339,348-349,354rulecoevolution,200Hyperstriatumventrale,11,88,314history,43,62,66,301,303,311,329,352medicalpartof(IMHV),88-89,96-97,314human,178,349-350,354,363Hypothalamus,50organic,331-332perspective,119,300-302Imitation,6,10,17-19,33,38,44,53,143,145,process,3,14,25,30,325,334,343156-157,179,263,311-312,336,340,342-reconstruction,35344syntheticviewof,36programlevel,156-157theory,23,39song,314,319,325,341Exaptation,56,311Imprinting,6,10-11,14,31,46,53,64-65,85-91,Exploration,98,118,31293-99,312,314,338ofself,33InferenceExtinction,74,110,222causal,211,221practical,193-194,199Featuredetector,87,96reasoning,205,209Feedinginnovations,311,314-315,319,321strongcomparative,36Feeling,26,200-201,246,250,254-257,269,276,transitive,64301weak,36functionof,200,255InformationFieldstudy,26,62,143encapsulation,13,48Fitness,15,116,143,145,185,311,321,334-335,explicit,implicit,153339,342,348,351integrationof,180,276,285Foodprocessing,5,44,46-48,50,55,57,118,344caching,224,273,278-280,282,284-285,312processingparadigm,27storing,11,14,17-18,46,54,65,273,277,285-retrievalof,274,276286,314,319spatial,17,319Forebrain,118,314-317,332storage,76,87,273Foresight,43-44unreliable,236,355Fuzzylogic,30,115Inheritance,28,174,180,335Inhibition,8,128,256Generalization,33,61,94,104-105,109-110,116,Innateness123-126,128,134,137,147,226,274behavior,28,289Geneinformation,32,333,336-338,341,344genome,23,28-29,31,61,185,338programming,289,299,332,339genotype,37,335releasingmechanism,31 382SubjectIndexInnovation,14,23,25-26,30,34,39,178-180,language-competentchimpanzee,107,224,232311,314-316,319,321-325language-richenvironment,221,119,232cultural,311skill,228,232evolutionas,34Lawofeffect,andofexercise,63-64,185feeding,311,314-315,319,321Learningvocal,321ability,11,53,314,318-319,321,339Insight,26,32-33,199,226,289-290,292,294-adaptive,335295,298-303algorithm,186,337Instinct,28,34-35,97animal,93,260,266Intellect,14,57,117associative,4-6,8-10,12,14,18,43-44,46-48,Intelligence,5,25,29,48,53,56,76,143,205,50,53-57,76,85,128,137,168,260-261,264,233-235,258,289,300-301,332,343266,269,337creative,179aversion,54,195general,257avoidance,7,39,317human,124,331conditional,130individual,117,143discriminative,70,95-96,123,126,130,167,234,scalesof,25312social,6,10,14,53,143,149-150,160-161explicit,264supramodulargeneral,118heuristics,333,344tests,117implicit,135,259-261,264Intentional),146,168-169,172,175-176,193,incentive,194-198,201246individual,31,174-175,314,317-319,321,324,action,168,194329,334-343being,53,168,179-180instrumental,44,68,72,75cognition,198invertebrate,258intentional/causalcognition,10,171,236language,236intentionality,56,168-169,171,185,190,209,latent,69,98236lawoflearning,62-64movement,176observational,226,230,297,318relation,6,10-11,168,181perceptual,53,89,94,96,99representation,186,190,195-197,199-202recognition,6,10-11,87-88,96state,169-170,176,194reinforcement,143understanding,144,173relational,129,131-133Intermodality,33rewarded,89,96-97social,6,10-11,44,145,157,173-180,226,311,Knowledge,11,29,43,135,145,153,155,159,314,316-325,329,333-343165,169,185-188,209,228,231,248,259,song,6,7,10-11,38,53,86273,275,285,303,337,355spatial,65,314,318abouttheworld,205,209,274-275strategy,332,338acquiringprocessesinevolution,332system,261,269,324causal,198-199theory,26,226explicit,153,264trialanderror,44,64,155,171,226,289,294,implicit,135,151,153297,303-304,334lifeasaknowledge-gainingprocess,29,31vertebrate,71-72,76,258Lexigram,126,224,226-230,233-234Landmarks,6,10,51-53,55-57LinguisticLanguage,6,10,34,43,53,131,138,194,211,behavior,224222,224-226,228-230,232-236,245,250,competency,107,229,303258-269,303,336,342,347-348,351-352,interaction,221,229,232354-355,363Localenhancement,179acquisitionof,229,258animal,228-229,232Machiavellianintelligence,205animallanguageproject,229Macroevolution,23,35-36human,178,228,230Matching-to-sample,105,119,130,276 SubjectIndex383Memorycircuit,23,39,119autobiographical,263-265connectionism,198declarative,273-275,286groupselection,332,338episodic,11,14,224,265-266,273-277,281,network,23,39285-286substrate,11,311-312,314,316,319episodic-like,17,273,275,278,282,285Niche,34,38,43,46,48,117,254,331-333,335,long-term,52339,341,343semantic,273-275,285convergent,25short-term,71,268Norns,185-187,190,194,198,202spatial,6,10-11,17-19,46,66,312,314,318-Novelty,130,144,311,314,324320,325system,10,17,47,56,267,273-275,319Objectconstancy,125Mentality,3,5Occasionsetting,53capacity,291,340Omissiontraining,68concept,295Ontogeny,7,9,14,17-18,23,27mental-stateattributions,26ritualization,176-179rehearsal,205,214-215,217-218Openprogram,28,31,35representation,248,301,332Opportunism,314-317,325Rubicon,253,265Organization,23-24,30,37-38,45-47,57,151,scenario,289,296,303156,178,253,257,264,314,332,349-350,state,6,10,26,145-146,148-149,151,153,155,353-354160-161,170visualization,289Pain,186,200,246-247,251,253,256,262,268-Metarepresentationalcapacity,144269,300-301Mimicking,156,174Partialreinforcementeffect(PRE),74-75MindPatternrecognition,208human,33-34,44,48,103,241Peakshift,128,134-135,138theoryof(ToM),6,10,48,53,56,62,145-146,Perception151,155-157,160,205,209,218,265,343-344constancymechanism,146Mindreading,143-145,149,151,155,158,160mapsoftheworld,332Miocene,330-331,343motorskill,261Modularorganization,47social,170Modularity,10,13,17,43,45-48,50,53-54,76,system,137,267254,319,324,337Perspectivetaking,158Module,10,13,17-19,38,45-48,50,52-57,91,PETscan,20893-95,97,117,137,254,324,332,335,337,Phenomenology,250340,342Phenotype,13,23,36-37Monitoring,158,248,302adaptation,334Motivational),23,61,97,145,148,150-153,flexibility,330,332,338,344194-197,200,258Phylogeny,14,24,27,35-37,47,58,316state,23,148,150,153,194-195descent,35system,97,196-197distribution,324Mutation,311,314,323-324,332,342inertia,311Physiology,5,48,245Naturalvariation,29Planning,43,199Nature-nurture,28Plasticity,49,198-199Neobehaviorism,186,193,198adaptivebehavioral,49Neocortex,118,206-208,218,312neural,198-199Neophobia,312,314,317-319,325,340Play,155,158,206,290Neostriatum,118Pleistocene,329-333,338,343Nervoussystem,4-5,20,35,55,64,98,103,119,Pliocene,329,331198,201,253,255,257,262,300,336-337Policing,350,352,355,359Neuron,30,39,117-118,185,198,300,302-303,Preadaptation,344355Precipitation,330 384SubjectIndexPredationskills,311Recognition,12,75,85,88,91,94,96-97,116,Predisposition,27,32,87,97,279,282147,149,160,166,208,236,263,265,291Prelimbicarea,199rules,149,154,159-161Primatesystem,91,93-95brain,206,234Recollectiveexperience,276brainevolution,205Reductionism,34,98cognition,165,180,205Reflexivity,106-107,115,209cognitiveadaptation,171,180Rehearsal,mental,205,214-215,217-218,236cognitiveskills,165,181Reinforcementsociety,143allocation,104,107,110vision,45contingenciesofreinforcement,105,222Priming,274effect,74,223,320Privatization,201,241,250learning,143Problemsolving,25,168,225,291mechanismof,340Producer-scroungergame,322,324partialreinforcementeffect,PRE74Progression,30,251,289process,64,185,202Psychologyreinforcer,49,53,69,188,198,200,223,227-cognitive,4,134-135,138228,275,333comparative,26,43-47,57,75-76,125,128,switch113,118131-132,138,193schedulesof,117,187contemporary,5S-R/reinforcementmechanism,186,190,198,200developmental,54S-R/reinforcementtheory,193evolutionary,3,6,13,15-20,23,48,54,251,theoryof,185254,332,355-356Relationexperimental,253,342aboutness,266-267,269folk,145-146,156,159-160,192abstract,25,105,129human,47,103intentional,6,10-11,168,181neobehaviorist,186matching,167,181psychologicalcontinuity,236temporal-spatial,285Psychophylogenesis,23,25RelationshipPurposeaction-outcome,190,202general,44,324,331-332,335-338,341betweenevents,134,136,138generalandspecialpurposemodule319,333,336betweenrelationships,136,138generalandspecialpurposeheuristics,324,337betweenstimuli,115,123,128-130,133-135,generalandspecialpurposemechanism,331,355137,225generalpurposelearning,335,337-338,341contiguous,188,190purposiveactions,198probabilistic,223purposiveness,198social,138,165-166,236special,44,319,324,331,333,335-338,355spatial,258specialpurposealgorithm,333,337-338third-partysocial,138,165-167,170-171specialpurposerule,336,338Remembering,6,10,264,274-275,282,286,318-Puzzleboxes,214-215319RepresentationQualia,241,244,251,253explicit,43-44,153implicit,44,153Ratcheteffect,178intentional,186,190,195-197,199-202Reasoning,6,10,39,53,131,133,137-138,167,internal,44,266-267199,205,209-211,221,246,274-275,299levelof,123,134,137analogical,131,137motivational,193causal,199,205,209-211multimodal,33cause-effect,221neural,88inferential,205,209misrepresentation,191Recall,224,263-266,268,273-274,276-277,282,representationalcapacities,144,147285-286stimulus,138 SubjectIndex385Responsefacilitation,340consummatory,74,200foraging,317,320,325instrumental,74insectcolonies,350organization,23intelligence,6,10,14,53,143,149-150,160-161responsiveness,86,150intelligencehypothesis,14,143,161Revaluationlearningtheory,226goal,187,201life,343retrospective,93science,349,363Roletaking,156sociogenesis,173Specialization,18-20,30,46-47,54,57,65-66,Scatterhoarding,17,277117,174,277,311-312,314,318-319,324-Selection325,341cultural,15Stimulusgroup,15,332,338,348-350,354-355auditory,87,91,93kin,350classification,105natural,15-18,27,29,35,191,241,243-244,complex,107,110,118-119,123-124,226248-249,254,321,323-324,332,334,338,compound,69,90,93340,348conditional,225pressure,13,17,35,350conditioned(CS),55,63-64,67-69,222,225,sexual,321296Selfenhancement,145,174consciousness,147,267equivalency,103recognition,147,265multifeatured,104mirror,147,160natural,116,126representation,250proximal,146,160Selfishness,349,351-352representation,137-138Sensation,47,57,201,241-248,250-251,256,stimulus-response(S-R),185-186,190,193,198-301-302202,253,294Senses,5,43-46,48stimulus-responsebehaviorism,253bodily,242unconditioned(US),63-64,67-79,221data,246,333,337visual,52,65,86-87,91,93,117,119,192experience,201,248-251Stonetools,230information,43,46,56Strataofexistence,29modalities,147Strategicmodeling,338,344organs,45-46,248,337Stylisticinertia,245qualia,244,253Subjective,4,50,241,243,250-251,315-316sensory-motorsynapses,198experience,241stimulation,242-243,246,248present,250system(s),45-46,54,58,61,103,129,256-267,quality,251336,337Symbolic),109,174,225,229,232-234,236,Sensitization,31,33,64,67-68,198239Sentenceproduction,266communication,226,233Sentientbeing,242,268information,233Sex(ual),85,148,160,347,356interchanges,236behavior,86,103presymbolic,267dimorphism,312processes,269imprinting,86Symmetry,106-107,115,119selection,321andreflexivity107,115Simplification,30testforsymmetry,106Skinnerbox,61,333andtransitivity,107Sociality,39,207Synapse,28,198,302brain,205Systematics,36cognition,177,205,208Systemicanalysis,29complexsociallife,343Systemstheory,23,27,29,36 386Tailoring(evolutionary),65-66Taxis,31Teaching,27,115,177,366,342Tectumopticum,118Telencephalon,317,319,325Theoryofbiology,211oflearning,64,128,192,194,226,266ofmind(ToM)(seeMind)Thought,6,10,15,33-34,138,243,246,266,303,348-349Timingcircadian,6,10,48-50,55,137interval,6,10,49-50ordinal,49-50,53Tooluse,175,177Tradition,16,26,33,61,64,66,105,178-180,199,223,243,315,333,342-344Transitivity,106-107,113,115,167,180inference,64Translucentenvironments,149-150Transmissioncultural,18,174,178,311,314,321,323-325ofinformation,35,178,355Transposition,128-129,134,138Traumaticavoidance,64-65Triangulation,88Understandingintentional,144,173,181relational,138Universals,123Unpredictability,43,290Urbanization,317Verballanguage,34Vestibularorgans,51Violationofasocialnorm,356-359Vision,45,48,147,207,242,300acuity,208isolation,299processing,46,207-208system,93,117-118,207,338Vitalistexplanation,30Vocallabelingcompetency,117Vocalization,148,166,174,211,229,268Voluntarymovement,33Weber"slaw,46,49What-whereconfiguration,276What/whenrule,7-13,17-20,39.SeealsoHowrule SeriesForewordBiologypromisestotheleadingscienceinthiscentury.Asinallothersciences,progressinbiologydependsoninteractionsamongempiricalresearch,theorybuild•ing,andmodeling.Butwhereasthetechniquesandmethodsofdescriptiveandex•perimentalbiologyhavedramaticallyevolvedinrecentyears,generatingafloodofhighlydetailedempiricaldata,theintegrationoftheseresultsintousefultheoreticalframeworkshaslaggedbehind.Biologicalresearchiscurrentlymuchdrivenbypragmaticandtechnicalconsiderationsandremainslessguidedbytheorythanotherfundamentalsciences.Asaconsequence,theory-drivenresearchintosomeofthemajoropenquestionsofbiology,suchastheoriginandorganizationoforganismalform,therelationshipbetweendevelopmentandevolution,orthebiologicalbasesofcognitionandmind,hasbeenimpeded.Thisseriesintendstohelpfillsuchconceptualgapsbypromotingthediscussionandformulationofnewtheoreticalconceptsinthebiosciences.TheoreticalbiologyhasimportantrootsintheexperimentalbiologymovementthatformedintheearlydecadesoftwentiethcenturyVienna.PaulWeissandLudwigvonBertalanffywereamongthefirsttousetheterm"theoreticalbiology"inamodernscientificcontext.Theirusageoftheexpressionwasnotlimitedtomathe•maticalformalization,asisoftenthecasetoday,butratherappliedtoageneralthe•oreticalfoundationofbiology.Thissyntheticendeavoraimedatconnectingthelawsunderlyingtheorganization,metabolism,development,andgrowthoforganisms.Itincludedunderstandablylittleofpopulation,ecological,molecular,andevenevolu•tionarytheory,whichtodayrepresentthemajorconnectiveconceptsinbiology.Inadditiontothese,asuccessfulintegrativetheoreticalbiologymustencompassrele•vantaspectsofcomputationalbiology,semiotics,andcognition,andshouldhavecontinuitieswithamodernphilosophyofthesciencesofnaturalsystems.Itisthistraditionofacomprehensive,cross-disciplinaryintegrationoftheoreticalconceptsthatthepresentseriesintendstoreturntothecenterofbiologicalresearch.Thename"ViennaSeries"isreflectiveofthelocationofinitiatingdiscussionmeetingsfortheseriesandcommemoratestheseminalworkoftheaforementionedfoundingscientists.Theseriesisspawnedbytheyearly"WorkshopsinTheoreticalBiology"heldattheKonradLorenzInstituteforEvolutionandCognitionResearchnearVienna,Austria.TheKonradLorenzInstituteisaprivateinstitution,closelyassociatedwiththeUniversityofVienna.Itfostersresearchprojects,seminars,workshops,andsym•posiaonallaspectsoftheoreticalbiology,withanemphasisonthedevelopmental,evolutionary,andcognitivesciences.Theworkshops,eachorganizedbyleadingexpertsintheirfields,concentrateonnewconceptualadvancesoriginatinginthesedisci•plines,andaremeanttofacilitatetheformulationofintegrative,cross-disciplinary viiiSeriesForewordtheoreticalmodels.Volumesonemergingtopicsofcrucialtheoreticalimportancenotdirectlyrelatedtoanyoftheworkshopswillalsobeincludedintheseries.Theserieseditorswelcomesuggestionsforworkshopsorbookprojectsonnewtheoreticaladvancesinthebiosciences.GerdB.Miiller,UniversityofVienna,KonradLorenzInstituteGiinterP.Wagner,YaleUniversity,KonradLorenzInstituteWernerCallebaut,LimburgUniversityCenter,KonradLorenzInstitute SpeciesIndexAccipitrida(hawks),315-316Dark-eyedjunco(Juncohyemails),65Africanmouthbreeders(Tilapiamacrocephala),62,Dinosaur,3774Dog(Canissp.),33,63-64,67-68,149,222,230,Anseriformes(ducksandgeese),312,314-315254-255,269,279,283,300-301Ant,50-51,55,146,148,175,258Dolphin(familyDelphinidae),37,116,165,167desertant(Cataglyphisfortis),50-51,55Dove,315,317,320Ape,18,130,143,147,150-151,155-157,169-Dragonfly(orderOdonata),302170,199,208-212,217-218,221-222,224,227,Duck(familyAnatidae),35,87,315229-230,232-234,236,254,265,267,275,290,domesticduckling(Anasplatyrhynchos311,344domesticus),87Apodiformes(swifts),315mallardduckling(Anasplatyrhynchos),87Arachnids(classArachnida),32Arthropods(phylumArthropoda),146,160Elephant(Loxodontaafricana),167Baboon(Papiosp.),144,165,211Falconida(falcons),315-316Bat(orderChiroptera),37,46Falcon(Falcosp.),315Bee{Apissp.),14,43-45,48,52,56,62,64,66-Finch(familyFringillidae),31871,75-76,146,258,267,350-351cutthroatfinch(Amadinafasciata),318honeybee(Apismellifera),14,48,56,62,64,66-greenfinch(Carduelischloris),31771,75-76,350zebrafinch(Taeniopygiaguttata),318Black-cappedchickadee(Parusatricapillus),65Fish,14,37,62-64,72-76,245,258,267,290,Blackbird(Turdusmerula),317293,300Bonobo(Panpaniscus),39,148,160,212,222,228Flatworm(phylumPlathelminthes),64Bonyfish(classOstheichthyes),63,258Fringillidae(finches),319Fundulus(killifish),63Caribgrackle(Quiscaluslugubris),320Carnivore(orderCarnivora),206-208,290Galliformes(phaesants,partridges),312,314-315Cat(Felissp.),62-63,123,146,254-255,269Gibbon(Hylobatessp.),151housecat,341Giraffe(Giraffacamelopardis),37Cephalopod(classCephalopoda),32,37Goldeneagle(Aquilarapax),291Cetacean(orderCetacea),206,344Goldfish(Carassiusauratus),14,62,64,72-76Charadriida(gulls),315Goose(familyAnatidae),51,315Chicken(Gallusgallus),63,128,161Gorilla(Gorillagorilla),153,156-157,230Chimpanzee{Pantroglodytes),14,43-44,56,75,Gull(familyLaridae),36,124,314-315126,131,133-134,158,160,168-170,174-180,herringgull(Larusargentatus),124199,209-210,212,214-215,217,224,226-227,Guppy(Poeciliareticulata),341233-234,290,342Ciconiida(herons),315Hamster(Cricetussp.),49-51,56Clark"snutcracker(Nucifragacolumbiana),17,goldenhamster(Mesocricetusauratus),49318-319Hawk(Accipiter,sp.),315Columbidae(pigeons),319Hen(Gallussp.),91,161Columbid,317s-318Heron(familyArdeidae),315Columbiformes(pigeonsanddoves),315Horse(Equussp),311Corvida(crows),315-316Hyena(familyHyaenidae),147,165,167Corvidae(crows),319Corvid,117,130-131,138,201,318-319,344Insect(classInsecta),32,51,175,277,290,300-Cow(ftusp.),292,311303,321-322,350-352Cowbird(Molothrussp.),277Insectivore(orderInsectivora),46,206Crab(infraorderBrachyura),137Invertebrate,5,9,23,33,43,62-63,71,258,300Crow(Corvussp.),130,296,315Americancrow(Corvusbrachyrhynchos),296Japanesequail(Coturnixcoturnixjaponica),87Crustacean(classCrustacea),32,76Jay,14,17,130,266,273,277-279,282,285-286,Cuckoo(familyCuculidae),277318-319Cuttlefish(Sepiaofficinalis),64Europeanjay(Garrulusglandarius),277 370SpeciesIndexJay(cont.)Australianparrots,316Mexicanjay(Aphelocomaultramarina),318-319Passerida(songbirds),315-316pinyonjay(GymnorhinusCyanocephalus),318-Passeridae(sparrows),319319Piciformes(woodpeckers),315scrubjay{Aphelocomacoerulescens),14,17,266,Pigeon(Columbalima),14,33,52,55,62,64,75,273,278,282,285-286103-111,113-119,123-135,138,222,312,315,Junglefowl,95317,322-323,333,335feralpigeon,333,335Kea(Nestornotabilis),38,39Primate(orderPrimates),14,38,45-46,53,56,Kingfisher(Alcedosp.),12562,72,76,107,126,131,133,138,143-151,153-156,158-160,165,167-171,174,Lion(Pantheraleo),165,167,233177,179-181,198,201,205-209,211-212,217-218,228,234,255,258,265,275,300,Macaque(Macacasp.),167,174343Japanesemacaque(Macacafuscata),11AProsimian,206-207,218longtailmacaque(Macacafascicularis),167prosimianprimate,206Magpie{Picapica),291Psittaciformes(parrots),315Mammalia,63Mammal,26,32,37,45,50-51,63,72,75-76,85,Racoon(Procyonsp.),130116-118,138,165,167,180,206,208,211,255,Rat(Rattussp.),14,32,48,53-55,57,62-63,65,257-258,277,290-291,311,329,331-332,336,72,74-76,124,127,129-130,143,154-155,340-341,343-344180,186-188,190-197,199-202,205,211,222,Marsupial,336224,275,333,335,341Mouse(familyMuridae),62-63,180blackrat(Rattusrattus),341Monkey,7-8,14,54,62-63,75,130-131,137,laboratoryrat,186,201151,154-155,166,170,208-212,214,217-218,Norwayrat(Rattusnorvegicus),341226,232,276-277,311,321,344Old-worldrats(e.g.Rattus),63capuchinmonkey(Cebussp.),170,210whiterat,124NewWorldmonkey,214Raven(Corvuscorax),14,123,199,289-294,296-redcolobusmonkey(Procolobusbadius),212297,299-304,312rhesusmonkey(Macacamulatto),7,14,62,75,Reptile(classReptilia),72,290,300232Reptilian,33,75spidermonkey(Atelessp.),212,214Rodent(orderRodentia),47,63,275squirrelmonkey(Saimirisp.),131Rook(Corvusfrugilegus),130,133vervetmonkey(Cercopithecusaethiops),150,166,321Sandpiper(familyScolopacidae),314Muscicapidae(flycatchers),319Sea-hare(Aplysiacalifornica),33,198,257Sealion(familyOtariidae),116Nonhumanprimate,56,62,107,126,143,154,Slug,267,269168-171,174,177,179,258Snake,7-8,33,54,230Nonhumanvertebrate,257-258,261Songbirds(Passerida),315-316,321Songthrush(Turdusphilomelos),316-317Octopus(Octopussp.),70,71Spider(classArachnida),233Opossum(familyDidelphidae),75Sponge(phylumPorifera),254Orangutan(Pongopygmaeus),150,214,217Stentor,43,56Oscinebirds(suborderOscines),312,314,319Termite(orderIsoptera),43-44,175,179Paramecium,30Tit(Parus),17,316-317,319Paridae(tits),319bluetit(Paruscaeruleus),317Parid,291,319greattit{Parusmajor),317,319Parrot(orderPsittaciformes),7-8,117,130-131,marshtit(Paruspalustris),17,317,319133,138,229,315-316,344Toad(Bufoarenarum),75Africangreyparrot(Psittacuserithacus),111,Tubifex,72130,229Turtle(Geoclemysreevesii),75 SpeciesIndexVertebrate(subphylumVertebrata),5,9,32-33,43,49,56,62-63,66-72,75-76,210,217,245,257-258,261,267,300,311Wasp,64,233-234,258diggerwasp(Philanthustriangulum),52Waxworm,278-279,281Wolf(Canislupus),233,290-291,301,303,312Woodpecker(familyPicidae),315-316Worm,62,72-74,256,267,278-279,281-282Zenaidadove(Zenaidasp.),317,320 AuthorIndexAbercombie,B.,89Boyd,R.,6,14-15,43,319,324,349,355,358.Abrahams,R.D.,347Seealsochapter18(329-346)Aggleton,J.P.,276Braithwaite,V.A.,258Ahlquist,J.E.,316Brakke,K.E.,229Aiello,L.C,336Brenowitz,E.A.,312Aitken,M.R.F.,134-135Brewer,W.F.,259Akins,C.K.,18Broecker,W.S.,330Aldavert-Vera,L.,118Brooks,L.R.,134Altevogt,R.,290Brothers,L.,205Amsel,A.,72,75Brown,J.L.,312,320Andrew,R.J.,96Brown,M.W.,277Aristotle,34Brunswik,E.,125,221Arluke,A.,347Bugnyar,T.,311Aronson,L.R.,62Burgess,W.,199Asano,T.,107,116Bush,R.R.,7,63Astley,S.L.,104,116,126Buss,D.M.,3,16,356Atz,J.W.,34-35Byrne,R.W.,25,156,161,169,205,265,315,343Aydin,A.,129Calje,H.,347Baer,K.L.,311Call,J.,18,126,130,165-166,168-169,205,209Baker,A.E.M.,311Cambefort,J.P.,321Balda,R.P.,277,289,312,318-319Campbell,C.B.G.,25Balleine,B.W.,4,6,11,14,33,143,154,205,Campbell,D.T.,14,18,26,332209,236,275.Seealsochapter10(185-204)Cant,J.G.H.,150Bandura,A.,226Caraco,T.,322,324Barkow,J.H.,332,347Carey,S.,56Barlow,H.B.,46Carlier,P.,291,317,320Baron-Cohen,S.,53,56Caro,T.,341Barton,R.A.,46,58,205Carp,J.S.,198Basalla,G.,178,343Carr,J.A.R.,49Bateson,P.P.G.,6-7,11,14-15,46,48,53,154,Catchpole,C.K.,7312,314.Seealsochapter5(85-102)Cavalli-Sforza,L.L.,334Bednekoff,P.A.,318-319Chantrey,D.F.,90,94Bennett,P.M.,312,316Charnock,D.J.,192Ben-Ze"ev,A.,347Chater,N.,138Beran,M.J.Seechapter12(221-238)Cheney,D.L.,166,178,211,342Berridge,K.C,200-201Cheng,K.,52-53Besnier,N.,347Cherfas,J.J.,86Bever,J.,206Chou,L.-S.,335,341Bhatt,R.S.,104Cipolla-Neto,J.,88Biebach,H.,286Clayton,N.S.,6,11,14-15,17,53,86,201,224,Bielby,D.D.,347266,314,319,344.Seealsochapter15(273-BierensdeHaan,J.A.,290288)Bischof,H.J.,86Clutton-Brock,T.H.,312Bitterman,M.E.,3-9,14,25,44,47,53-58,98,Cobey,S.,69180,223,254,258,300,332,335,337,355.SeeCollett,M.,56alsochapter4(61-79)Collett,T.S.,44,51,258Blaine,B.,347Colwill,R.M.,190Bleek,W.,347Connor,R.C,167Blough,D.S.,128Constancia,M.,85Boesch,C,174,177-178Cook,M.,7-8,54Boire,D.,317Cook,R.G.,132Bolhuis,J.J.,86-87,89,91Cook,S.E.,86Bolles,R.C,193Cosmides,L.,3,45,48,53,333,337,356 374AuthorIndexCourage,M.L.,264-265,267Eichenbaum,H.,275,285Couvillon,P.A.,64,68-71,74Eisenberg,J.F.,336,340-341Cowles,J.T.,75Ellickson,R.C,353-354,357-358,363Cox,B.A.,347Elliott,M.H.,72,74Cramer,A.E.,52Emery,N.J.,53Crawford,B.H.,311Empson,W.,241distance,D.,342Enke,J.L.,347Etienne,A.,51,56Daly,M.,3,312D"Amato,M.R.,107,132Fagen,R.M.,321Darwin,C,12,15,25,33-34,97,138,236,241-Fanselow,M.S.,286242,244-245,250,255,301,331,348-349Feldman,M.W.,334Dasser,V.,166Felsenstein,J.,316Davenport,R.K.,228Fersen,L.v.,104,107,110,116,118,125,130Dawkins,R.,144Ferster,C.B.,222Dawson,A.G.,329Fetterman,J.G.,129Dawson,G.R.,340Fields,P.E.,124deBeer,35Fisher,213-214deRaad,B.,347Fitzgerald,M.,262deVos,G.J.,89Flaherty,C.F.,74deWaal,F.,148,165,169Fletcher,P.C,274Delacour,J.,276-277Fodor,J.A.,13,47-48,54,103,160Delius,J.D.,6,11,14,33,125,127,131.SeealsoForsell,Y.,274chapter6(103-122)Fragaszy,D.M.,311Dennett,D.C,144-145,159,185,209,244Freeman,W.J.,29DeRivera,J.,129Frege,G.,250DeVoogd,T.J.,312,317,319Freud,S.,262Devore,I.,335,338Frey,U.,276Dickinson,A.,4,6,9,11,14,33,44,89,143,154,Frith,C.D.,208205,209-211,236,260,301.SeealsochaptersFunayama,E.S.,7110(185-204)and15(273-288)Ditlevsen,P.D.,330Gaffan,D.,275-276,286Dobzhansky,T.,23,35Galef,B.G.,Jr.,174,320,335,340-342Dolman,C,317,320Gallistel,C.R.,45,48,50,52,55Domjam,M.,211Gallup,G.G.,265Donahoe,J.W.,198Garcia,J.,195,224,333Dorken,M.D.,136Gass,C.L.,27-28Dougher,M.J.,105Gelman,S.A.,45,47Diicker,G.,290Giannakoulopoulos,X.,262Dugatkin,L.A.,341Gibbon,J.,49-50Dukas,R.,289Gigerenzer,G.,45,47,53-55,57Dunbar,R.I.M.,6,14-15,145,161,199,236,Gillan,D.J.,131,275290,312,343,347.Seealsochapter11(205-Giraldeau,L.A.,317,319,322-324219)Gluck,M.A.,118Dunford,C,334Gluckman,M.,347-349,353Dunning,J.B.,317Goldberg,J.,320Dusek,J.A.,275Goldstein,D.G.,337Dusenbery,D.B.,47Gomez,J.-C,153Dwyer,D.M.,93Gonzalez,R.C,73-74,128Dyer,F.C,43Goodall,J.,176,179Goodman,R.F.,347Eddy,T.J.,53,152,158Gottlieb,G.,86-87Edelman,G.M.,332,338Gould,C.G.,290Eder,D.,347Gould,J.L.,290 AuthorIndex375Grand,S.,185Howe,M.L.,264-265,267Gray,L.,312Huber,L.,3,6,13,43,65,125,311.SeealsoGreen,S.,165chapter2(23-41)Greenberg,R.,312Hull,C.L.,61,63,66,128,185,194,198Greene,S.L.,110Hume,D.,199,221,234Greenspan,R.J.,24,27Humphrey,N.K.,6,14,53,161,167,201,253,Gregory,R.L.,242-243256,301,343.Seealsochapter13(241-252)Griffin,D.R.,289Hutchison,R.E.,86Griffiths,D.P.,93-94.Seealsochapter15(273-Huxley,J.,34288)Grill,H.J.,200-201Immelmann,K.,86,338GRIP(GreenlandIce-coreProjectMembers),330Ince,L.P.,256Guntiirkun,O.,118Ishida,M.,75Guthrie,E.R.,61James,H.,89Hall,G.,128Jaynes,J.,256Hamilton,W.D.,350Jerison,H.J.,117,331,332Hammond,L.J.,188,192Jitsumori,M.,Seechapter6(103-122)Hampton,R.R.,54Joffe,T.,207-208Handelman,D.,347Johnson,M.M.,87-88,94Hanson,H.M.,128Johnston,T.D.,12,87,314Harcourt,A.H.,165Jolly,A.,161Harlow,H.F.,62,226Harrington,C.L.,347Kalat,J.W.,312Harvey,P.H.,35-36,312,316Kamil,A.C,17,277Hassenstein,B.,68Kammer,A.,302Hatch,K.K.,322Kandel,E.R.,33Hauser,M.,56,161,341Kant,I.,29Haviland,J.B.,347Kastak,D.,116Hawkins,R.D.,33,198Keeble,S.,211Haxby,J.V.,96Keller,F.S.,104Healy,S.D.,277,319Kelso,J.A.S.,28Heinrich,B.,6,14,33,199,201,312,344.SeealsoKim,J.J.,286chapter16(289-305)Kinderman,P.,209Heinroth,O.,34,85Kipling,R.,250Helder,R.,322Kirkpatrick-Steger,K.,104Hennessey,T.M.,31Klopfer,P.H.,312,317Herlitz,A.,274Klosterhalfen,S.,65,70Hermer,L.,48,53,57Klosterhalfen,W.,65Herrnstein,R.J.,103-105,124-125Knowlton,B.J.,261Hess,E.H.,85-86Koelling,R.,224,333Heyes,C.M.,33,39,48,54,62,138,146-147,Kohler,W.,128,210,233,290155-156,158-159,169,174,193,209,265,320,Krebs,J.R.,11,17,144,312,319323,340,342.Seealsochapter1(3-22)Kroodsma,D.E.,312Hillix,W.A.,154,199,205,209.SeealsochapterKrubitzer,L.,58,33312(221-238)Kruger,A.,178Hinde,R.A.,332Kruijt,J.P.,86Hirschfeld,L.A.,45,47Kruschke,J.K.,126Hodos,W.,25-26Kummer,H.,53,57,179,343Holland,P.C,53Kuno,H.,107Holliday,R.,35Hollis,K.L.,89Laland,K.N.,335,338,340-341Honey,R.C,88,90-91,94-97,128Lamb,H.H.,329-330Horn,G.,87-88,91,93-95Lashley,K.S.,61,124 376AuthorIndexLawrence,D.H.,129,241,250McFarland,D.,43,55Lea,S.E.G.,104McGrew,W.C,44,175Lee,D.W.,17McLaren,I.P.L.,89,94Lefebvre,L.,6,11,14-15,18,33,58,291,335,Meeuwissen,G.B.,86339,341,355.Seealsochapter17(311-328)Menzel,C,224Lehman,S.,330MenzelJr.,E.W.,26,29Lenz,R.,125Merry,S.E.,347,354Leslie,A.,211-214Mill,J.St.,221Levin,J.,347Mineka,S.,7-8,54Lewin,R.,228-232Mishkin,M.,276-277Lewis,K.P.,206Mithen,S.,47-48,57Lewontin,R.C,25Moore,B.R.,18,25,38,53,342,344Lickliter,R.,86-87Moore-Ede,M.C,49Limongelli,L.,170-171,210-211,214Morgan,L.,223Lipkens,R.,107Morris,R.G.M.,276Lipp,H.P.,47Moss,C.J.,167Locke,A.,176Mosteller,F.,7,63Locke,J.,123-124,243-244Mowrer,O.H.,61LoLordo,V.M.,55Muzio,R.N.,75Lombardi,C.M.,132Londer,R.,329Nadel,L.,263Longo,N.,74Nagell,K.,175Lorenz,K.,13,23,26-35,39,61-62,64-65,68,Nakagawa,E.,108,116,12785,312Neisser,U.,267Loveland,D.H.,103Nevo,O.,347Lowes,G.,72Nilsson,L.-G.,274Lubow,R.E.,103Nissen,H.W.,75Luciano-Soriano,M.C,107Nosofsky,R.,126Lythgoe,J.N.,46-47Nottebohm,F.,314Nyberg,L.,274Machiavelli,N.,205Mackintosh,N.J.,6,8-9,11,14,33,44,61,70,O"Connell,S.,208-209,211,214116,154,167.Seealsochapter7(123-141)Odling-Smee,F.J.,14,29Macphail,E.M.,6,9,14,33,66,72,75,200-201,Oetting,S.,86250,300,312.Seealsochapter14(253-271)Olson,D.J.,17Manabe,K.,117O"Reilly,R.C,94Mandler,J.M.,263,276Owens,M.E.,262Marcel,A.J.,242Marino,L.,206Packer,C,165,167Markham,M.R.,105Pacteau,C,135Markowitsch,H.J.,224,265,273-275,285-286Pagel,M.D.,35-36Marler,P.,165Paine,R.,347-349,353Marr,D.,47Palameta,B.,321,323,341Martin,R.D.,336Papini,M.R.,71,75Martins,E.P.,35Parsons,C.H.,86Marzluff,J.M.,291Partridge,T.C,329Masterton,R.B.,62Patterson,M.M.,256Maturana,H.,27Pavlov,I.P.,8-9,12,32,61,63,67-68,71,149,MaynardSmith,J.,35,349222-223Mayr,E.,27-28,35Pawlowski,B.,206Mazmanian,D.S.,125-126Pearce,J.M.,55,129McAdam,M.R.,214Pepper,J.,291McCabe,87,96Pepperberg,I.M.,7-8,117,130-131,229McDonough,L.,263Perner,J.,274 AuthorIndex377Perruchet,P.,135Schmajuk,N.,75Pert,A.,75Schneider,S.H.,329Phillips,K.,356Schoenfeld,W.N.,104Piaget,J.,199,221,236Schrodinger,E.,27Pillemer,D.B.,264,274Schull,J.,312Pinker,S.,332Schusterman,R.J.,116Platon,29Schutz,S.L.,74Plotkin,H.C,14,23,28-29Scott,A.M.,86Poli,M.D.,333Seeley,T.,350Polt,J.M.,86Seferta,A.,317-318Portmann,A.,312,317Seyfarth,R.M.,166,178Povinelli,D.J.,25,53,150,152-153,155,158-Shallice,T.,199159,169,208-209Shanks,D.R.,154,191-192,260Powdermaker,H.,352Shapiro,M.S.,64,70Premack,D.,117,130-131,156,159,168-169,Shepard,R.N.,44-45,47265Sherry,D.F.,7,17,45-47,56,277,312,319-320Pulliam,H.R.,334Shettleworth,S.J.,3-7,10,13,25,33,65-66,75,Purvis,A.,31697-98,117,137,180,222,254,277,289,312,319,324,332,335,337,355.Seealsochapter3Rambaut,A.,316(43-60)Ratcliffe,D.,290Shyi,G.,263Ray,E.D.,18Sibley,C.G.,316Reaux,J.,171Sidman,M.,105-107,115,119Reber,A.S.,135,259-260Siemann,M.,Seechapter6(103-122)Reboreda,J.C,277Silverman,I.,356Redhead,E.S.,52Simpson,G.G.,35,62Regehr,G.,134Sjolander,S.,29,33,40Reid,Th.,242,246Skinner,B.F.,61-62,221-223Rensch,B.,25,290Slater,P.J.B.,7Rescorla,R.A.,7,50,55,63,190,225Sloman,S.A.,103Revusky,S.,224Small,W.S.,62Richerson,P.J.,6,1415,43,319,324,349,355,Smith,B.H.,69358.Seealsochapter18(329-346)Smolker,R.,290Riedl,R.,27,34,36,38Sober,E.,9,14.15,36,151,349-351,355Riley,D.A.,128-129Soha,J.A.,53Rips,L.J.,103Spalding,D.A.,85Ristau,C.A.,26Spelke,E.S.,48,53,57,337Roberts,J.M.,347Spence,K.W.,61,128Roberts,W.A.,75,116,125-126Spencer,H.,23Rogers,A.R.,335,340Sperber,D.,45,47Rogers,L.J.,86Spetch,M.L.,52Roitblatt,H.L.,118,130Squire,L.R.,274Rovee-Collier,C,263,265St.John,M.F.,260Rozin,P.,56-57,312Staddon,J.E.R.,43Ruffman,T.,274Staffeldt,E.F.,312Rumbaugh,D.M.,6,154,199,205,209.SeealsoSteadman,P.,245chapter12(221-238)Stephan,H.,207Russon,A.E.,156,311Sterelny,K.6,14.Seealsochapter8(143-Ruttan,L.,165162)Stewart,D.J.,90Sacher,G.A.,312Suhonen,J.,277Sasvari,L.,316-317,319Suomi,S.J.,86Savage-Rumbaugh,E.S.,126,169,227-234Sutherland,N.S.,70Schacter,D.L.,7,45,47,56,274,312,319-320Szathmary,E.,349 378AuthorIndexTempleton,J.,323White,P.A.,221tenCate,C,87White,S.H.,264,274Terkel,J.,341Whiten,A.,145,150-152,155-156,159,161,169,Terrace,H.S.,43205,315,342-343Teuber,H.L.,61Whitman,C.O.,34-35Thomas,R.K.,25,43Whittle,P.J.,317-318Thompson,R.K.R.,131Whittlesea,B.W.A.,136Thorndike,E.L.,61-66,137-138,185,223Wilczynski,C.Seechapter19(347-365)Thorpe,W.H.,290Wilkie,D.M.,49Timberlake,W.,25Williams,D.A.,55Tinbergen,N.,36,52Williams,K.,341Tinklepaugh,O.L.,75Wills,S.J.,128-129,134Todt,E.H.,262Wilson,A.C,314,323-324Tolman,E.C,61,193Wilson,B.,130-132Tomasello,M.,6,10,14-15,18,126,130,138,Wilson,D.S.,14,47,57.Seealsochapter19143-144,151,154-155,161,199,205,209,236,(347-365)311,342.Seealsochapter9(165-183)Wilson,M.I.,3Tooby,J.,3,45,48,333,335,337-338,356Wilson,N.E.,211Toth,N.,230-231Wilson,P.J.,347Tschudin,A.,206Wittgenstein,L.,244Tulving,E.,224,265,273-276,285-286Wolfer,D.P.,47Wolpaw,J.R.,198Urciuoli,P.J.,107Woodruff,G.,156,168-169,265Wrangham,R.W.,175Vaccarino,A.L.,17Wright,A.A.,110,131-132Valero-Aguayo,L.,107Wyles,J.S.,314,323Vallortigara,G.,6vanKampen,H.S.,89Yager,R.R.,115VanderWall,S.B.,277Yamamoto,J.,107,116Varela,F.,27Yoshihara,M.,116Vaughan,W.,105,107,110,114-116,119,127Young,M.E.,130,133Veenman,C.L.,118Veissier,I.,311Zabel,C.J.,167Verner,J.,312Zadeh,L.A.,115Vince,M.A.,290Zeil,J.,44,51Visalberghi,E.,155,170-171,210-211,214,218,Zentall,T.R.,18,107,109,127,342311Zhuikov,A.Y.,64Zola-Morgan,S.,263,275Wagner,A.R.,7,50,55,63Zolman,J.F.,89Walker,C.J.,347Walker,E.P.,62Walker,J.J.,71Wallenschlager,D.,291Walters,E.T.,257Washburn,D.A.,226,232Wasserman,E.A.,104,116,124,126-127,132-133,191Watanabe,S.,104,118Watson,J.B.,61Weber,E.H.,46,49Weinberg,M.,192Weiser,L.Seechapter19(347-365)Wells,A.Seechapter19(347-365)Wheeler,P.,336