JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2005, 84, 313–325 NUMBER 3(NOVEMBER) THE

2005,84,313–325

NUMBER

3(NOVEMBER)

THECHALLENGEOFCHARACTERIZINGOPERATIONSINTHEMECHANISMS

UNDERLYINGBEHAVIOR

WILLIAMBECHTEL

UNIVERSITYOFCALIFORNIA,SANDIEGO

Neuroscienceandcognitivescienceseektoexplainbehavioralregularitiesintermsofunderlyingmechanisms.Animportantelementofamechanisticexplanationisacharacterizationoftheoperationsofthepartsofthemechanism.Thechallengeincharacterizingsuchoperationsisillustratedbyanexamplefromthehistoryofphysiologicalchemistryinwhichsomeinvestigatorstriedtocharacterizetheinternaloperationsinthesametermsastheoverallphysiologicalsystemwhileothersappealedtoelementalchemistry.Inorderforbiochemistrytobecomesuccessful,researchershadtoidentifyanewlevelofoperationsinvolvingoperationsovermoleculargroups.Existingattemptsatmechanisticexplanationofbehaviorareinasituationcomparabletoearlierapproachestophysiologicalchemistry,drawingtheirinspirationeitherfromoverallpsychologyactivitiesorfromlow-levelneuralprocesses.Successfulmechanisticexplanationsofbehaviorrequirethediscoveryoftheappropriatecomponentoperations.Suchdiscoveryisadauntingchallengebutoneonwhichsuccesswillbebeneficialtobothbehavioralscientistsandcognitiveandneuroscientists.

Keywords:mechanisticexplanation,operations,laws,levelsoforganization,connectionism,symbolictheories

_______________________________________________________________________________Peopleandotheranimalsbehave,andamajorobjectiveofthebehavioralsciencesistocharacterizethatbehaviorandidentifythecircumstancesthatbringitaboutandtheconsequencesthatchangeit.Thegoaloftenistospecifylawsthatrelatesuchvariablestobehavior.Someinvestigators,however,arenotinterestedinjustdiscoveringthoselawsbutinexplainingthemintermsofongoingprocessesoccurringinsidetheorganism.Thatis,theyseektounderstandthemechanismsoperativewithintheorganismthatexplainwhy,givenparticularenvironmentalcircumstances,spe-cificbehaviorsresult.Thiswasamajorobjec-tiveofthosepsychologistswhointhe1950sand1960screatedcognitivepsychology(Mill-er,1956;Miller,Galanter,&Pribram,1960;Neisser,1967).Italsowasthegoalofpracti-tionersofotherdisciplineswhointhe1970scametogetherwithcognitivepsychologyun-derthedesignationcognitivescience(Bechtel,Abrahamsen,&Graham,1998)orinthe1980sand1990sunderthelabelcognitiveneuroscience(Bechtel,2001a).

IthankWilliamTimberlake,theSpecialEditorforthisissue,aswellasAdeleAbrahamsen,ColinAllen,JoelMyerson,andLeonardGreen,forencouragementandhelpfulcommentsonearlierdraftsofthispaper.

Contactaddress:DepartmentofPhilosophy-0119,Uni-versityofCalifornia,SanDiego,9500GilmanDrive,LaJolla,CA92093-0119(e-mail:bill@mechanism.ucsd.edu).

doi:10.1901/jeab.2005.103-04

UsingterminologythatIwilldevelopbelow,theseinvestigatorsweretryingtodiscoverthemechanismsresponsibleforproducingthebehavior.Amajorchallengefacingsuchinvestigatorsistodevelopappropriatecon-ceptsforcharacterizingtheoperationswithinthesemechanisms(operationisatechnicaltermthatwillbeexplicatedbelow).Mycontentioninthispaperisthat,despiteprogressinsomespecificdomains,anadequatecharacterizationoftheoperationsinpsychologicalmechanismsstilleludesinvestigators.

Lackinganadequateaccountoftheopera-tionswithinpsychologicalmechanisms,cogni-tivepsychologistsandcognitivescientistsoftenhavecharacterizedtheseoperationsasofthesametypeasoperationsperformedbythewholeorganism.Thisisparticularlytrueoftheappealstorepresentationsandoperationsonthemthatappearinmanycognitiveaccounts.Suchanapproachtocharacterizingoperationswithinamechanismisnotuniquetothosetryingtocharacterizepsychologicalmechan-isms,butitisproblematic.Operationswithinamechanismoccuratalowerleveloforganizationthanthebehavingmechanismitself,andtheseoperationsaretypicallydiffer-entthanthoseperformedbythewholemechanism.Thetypesofoperationsoccurringatagivenleveloforganizationmustbediscovered,andsuchdiscoveryisfrequentlydifficult.AfterIdeveloptheconceptual

313

314WILLIAMBECHTEL

frameworkofmechanismandmechanisticexplanationandtherelatednotionofleveloforganizationinthefollowingsection,Iwilldevelopanexamplefrom19thcenturyphysi-ologicalchemistrythatisillustrativeofthechallengeandwhatisrequiredtosurmountit.Ithenwillfocusontheparticularchallengeindevelopingaccountsoftheoperationsinpsychologicalmechanisms.

Investigatorswhoseprimaryinterestisindiscoveringlawscharacterizingthebehaviorallevelitselfmayquestiontherelevancetothemofeffortstodiscoverpsychologicalmechan-isms.Inmanycases,however,anunderstand-ingoftheinternaloperationofthemecha-nismsuggestsenvironmentalfactorstowhichthewholemechanismissensitive,suggestionsthatthenrequirebehavioralinvestigationtoevaluate.Inotherdisciplinesthereoftenisreciprocalfeedbackbetweenaccountsofthewholemechanism’sinteractionswithitsenvi-ronmentandmodelsofitsinternaloperation.Understandingtheenvironmentalconditionsunderwhichorganismsacquireadiseaseguidesresearchersinvestigatingdiseaseme-chanismswithintheorganism,butinforma-tionaboutthosemechanismsalsoaresugges-tiveofenvironmentalcontingenciesthatcanmitigateorexacerbatethedisease.Inquiriesintopublic,environmental,andoccupationalhealthandinternalpathologymutuallysup-porteachother.Iwillreturntotheissueofhowbehavioralinvestigationsandinvestiga-tionsofinternalpsychologicalmechanismscansupporteachotherinthefinalsectionofthepaper.

MECHANISMS,MECHANISTICEXPLANATIONS,ANDLEVELS

OFORGANIZATIONMostaccountsofthescientificmethodcharac-terizescienceasthequesttodiscoverandtestlawsthatcharacterizethephenomenoninaparticulardomain.Lawshavesimilarlyplayedacentralroleinwhathasbeenthereceivedphilosophicalaccountofscience(Suppe,1977).Accordingtothedeductive-nomological(D-N)model,scientificexplanationinvolvesthederivationofastatementofthephenomenontobeexplainedfromstatementsoflawsandinitialconditions(Hempel,1965).Suchaconceptionofscientificexplanationprovidesagoodcharacterizationofthe

attemptsofbehavioristsinpsychologywhosought,forexample,toexplainbehaviorbyidentifyinglawsrelatingcontingenciesofre-inforcementtoresultingbehavior.Lawsoftenprovideagoodwayofrelatingbehavingsystemstotheentitiesandeventsintheirenvironmentthataffecttheirbehavior.Amajorquestinthelifesciences,however,isnotjusttorelateentitiestoothersintheirenvironment,butalsotounderstandwhyagivenentityrespondstoenvironmentaloccurrencesasitdoes.Scientistsengagedinsuchquestscommonlycharacterizethemselvesasseekingtounderstandtheresponsiblemechanisms.Thus,inbiologyonefindsdiscussionofthemechanismsofbloodcircu-lation,ofthermalregulation,ofcelldivision,andofproteinsynthesis.

AlthoughwhatRobertBoyletermedthemechanicalphilosophyplayedanimportantroleinthescientificrevolutionofthe17thand18thcenturies,itreceivedlittleattentionin20thcenturyphilosophyofscience.WhenadvocatesoftheD-Nmodeltriedtoaccountforhowscientistsexplainwhyagivenentitybehavesasitdoes,theyextendedthesamedeductiveframework.Theyproposedthatthelawsofagivendisciplinemightbederivedfromthelawsofamorebasicscience(togetherwithbridgeprinciplesthatwouldrelatethedifferentvocabulariesofthetwosciencesandboundaryconditions).Theyre-ferredtosuchdeductiverelationsbetweensciencesasreductions(Nagel,1961),butIwillusethetermtheoryreductiontodifferentiatethisconceptionofreductionfromtheverydifferentoneprovidedbymechanisticexpla-nations.

Adifferentstrategyistotrytoexplicatewhatscientists,especiallythoseinthelifesciences,haveinmindwhentheyrefertomechanisms.Onmyanalysis,amechanismisanorganizedsystemofcomponentpartsandcomponentopera-tions.Themechanism’scomponentsandtheirorganizationproduceitsbehavior,therebyinstanti-atingaphenomenon(Bechtel&Abrahamsen,2005;forrelatedaccounts,seeBechtel&Richardson,1993;Glennan,1996,2002;Ma-chamer,Darden,&Craver,2000).Accordingtothisview,amechanismisasystemoperatinginnature,andamechanisticexplanationisanepistemicproduct.Toarriveatamechanisticexplanation,scientistsmustrepresent(some-timesverbally,butoftenvisuallyindiagrams)

THECHALLENGEOFCHARACTERIZINGOPERATIONS

315

thecomponentpartsandtheiroperationsandthewaysinwhichtheyareorganized.

Thecentralfeatureofsuchmechanisticexplanationsisthattheydecomposeasystemthatproducesabehaviorintocomponentpartsandcomponentoperations.Thepartsandoperationsintowhichamechanismisdecomposedarecloselyrelated:therelevantpartsarethosethatperformoperationsandhenceareworkingparts.However,itisimpor-tanttodistinguishpartsunderstoodstructur-allyfromoperationsunderstoodfunctionallysinceunderstandingamechanismrequiresboththestructuralandfunctionalperspec-tives.Moreover,differentinvestigatorytech-niquesarerequiredtoestablishstructuralandfunctionalpropertiesofcomponents,andagivengroupofresearchersmaybeabletosecureevidenceonlyaboutoneortheother.Asaresult,researchersoftenconfrontthechallengeofrelatingpartswithoperations(anactivityIrefertoelsewhereaslocalization).Manyofthecomponentsofamechanismarethemselvesmechanisms—theyperformoperationsinvirtueoftheirparts(nowsubpartsoftheoriginalmechanism)perform-ingoperationsoftheirown.Thismereologicalrelationgivesrisetoaclearsenseoflevels—partsareatalowerlevelthanthemechanismtheycomprise.Althoughthismakesmechanis-ticexplanationsinherentlyreductionistic,thefocusinmechanisticexplanationisnotexclu-sivelydownwards.Mechanismsalwaysoperateincontextsandthesecanaffectthebehaviorofthemechanismitself.Moreover,insofarasthereisappropriatesystemicorganizationatthathigherlevel,agivenmechanismmaybepartofanothermechanismthatregulatesitsbehavior.Further,withevolvedmechanisms,interactionswiththeenvironmentarecrucialinselectingbetweenalternativemechanisms.Toaddresssuchissuesinexplainingthephenomenonofinterest,investigatorsmayneedtotakeintoaccountseveralhigherlevelsoforganization.

Sincethemechanismisnotjustanaggre-gateofitsparts(Wimsatt,1986),butrequiresthecomponentpartsandoperationstobeorganizedsoastoproducethebehaviorofthesystem,IfollowWimsatt(1976)inreferringtolevelsaslevelsoforganization.Therehavebeennumerousattemptstocharacterizelevelsasstrataacrossnaturebyfocusingonsuchfeaturesasthesizeofthecomponentparts

(P.S.Churchland&Sejnowski,1992)orthefrequencywithwhichcomponentsinteract(Wimsatt,1976),butthesefailtocapturetheclusteringofentitiesintolevelsthatfigureinscientificinquiry.Myrecommendationistoforegotheattempttoidentifylevelsasstrataacrossnatureandfocusonlyonwhatresultsfromtheattempttodifferentiatethecompo-nentpartsandoperationsthatfigureinaccountingforagivenphenomenon(Craver,forthcoming;Craver&Bechtel,submitted).Whenweidentifylevelsintermsofcausalinteractionswithinamechanism,entitiesthatarestructurallyalikemayappearatdifferentlevels.ProtonsforexampleinteractwithmembranesinthechemiosmoticmechanismresponsibleforconvertingenergyliberatedinoxidativereactionsincellsintoaprotongradientthatdrivesATPsynthesis.Protonsalsooccurinthemoleculesthatcomprisethemembrane,buttheseareatalowerlevelthantheprotonsthataretransportedacrossthemembraneandthusinteractwithit.Thereisnotalevelofprotons,butlevelscorrespondingtotheentitiesthatcausallyinteractinagivenmechanism.Theresultisahierarchyoflevels,butonethatischaracterizedrelativetothephenomenonaninvestigatorinitiallysetouttoexplain.

Animportantfeatureofthecomponentsatdifferentlevelsoforganizationisthattheytypicallycarryoutdifferenttypesofoperationsthanthoseatlowerorhigherlevels.Somaticcells,forexample,dosuchthingsassecreteenzymesandexchangematerialswiththeblood,organellesofthecelldosuchthingsassynthesizeproteinsorextractenergyfromoxidationreactions,enzymeswithintheorga-nellescatalyzeparticularreactions.Thiswasimplicitlyrecognizedintheneedforbridgeprinciplestorelatethevocabulariesoftwosetsoflawsintheoryreductionaccounts,butthereasondifferentdisciplinesinvokedifferentvocabularieswasnotidentified.Itisthatthedifferentdisciplinesfocusondifferentkindsofoperations.

Itisworthemphasizingwhyitisthatwholemechanismscandothingsthattheirpartscannot.Thesecret,asengineershavelongknown,istoorganizecomponentsappropri-atelysothattheiroperationsareorchestratedtoproducesomethingbeyondwhatthecomponentscando.Itisfordiscoveringsuchorganizationthatengineerswinacclaimand

316WILLIAMBECHTEL

securepatents.Organizationalsoiscrucialinnaturallyoccurringmechanisms—itisonlyastheyareproperlyorganizedthattheopera-tionsofthepartsofamechanismcombinetogenerateaphenomenonthatisbeyondthecapacityofanygivenpart.

Withthissketchofexplanationintermsofmechanisms,Icancommentbrieflyontherelationofmechanismstolaws.Therearetwodistinctwaysinwhichlawsfigureinaccountsofmechanisms(Glennan,1996).First,asnotedabove,lawsoreffectsarecharacteriza-tionsofphenomenatobeexplainedbymechanisms.Alawidentifiesaregularitybetweenvaluesofdifferentvariables.Iftherelationismediatedbyamechanism,themechanismcanexplainwhythelawholds.Thus,themathematicalrelationsidentifiedwithinmathematicalpsychologyoftenlendthemselvestomechanisticexplanation(Bech-tel&Abrahamsen,inpress).Second,someoftheoperationswithinamechanismcanthem-selvesbecharacterizedintermsoflawsrelatingvariables.Insomecases,lawsidentifycon-straintsonthewaysinwhichdifferentcompo-nentpartscanoperateonothers.Forexample,rateequationsinbiochemistryspecifytherateatwhichanenzyme-catalyzedreactioncanproduceitsproduct.Lawsdonot,however,specifytheparticularparts,operations,andorganizationinplaceinaparticularmecha-nismand,inthisrespect,mechanisticexplana-tionsgobeyondwhatlawsprovide.

Inthispapermyfocusisonthedecompo-sitionofamechanismintocomponentpartsandcomponentoperations.Thechallengeinconstructingmechanisticexplanationsisthatnormallyoperatingmechanismsdonotrevealeithertheirpartsoroperations.Notjustanywayofcarvingupthemechanismrevealstheappropriateparts.Therelevantpartsarethosethatactuallyperformtheoperationsinthemechanism.Toconsideranexample,al-thoughneuroanatomistsoverseveralcenturiessoughttodelineatepartsofthebrainsofhumansandotherspeciesintermsofthegyriandsulciproducedbythefoldingofthecortex,andthesestillserveasusefulland-markswhenidentifyingwhereoperationsoccurinthebrain,theydonotrepresenttheworkingparts.Brodmann(1909/1994)differ-entiatedbrainregionsusingcriteriasuchastypesofneuronsandthicknessofcorticallayersthathethoughtwouldmaponto

operations(althoughhelackedtoolsforactuallylocalizingoperationsinbrainre-gions).Modernbrainmappers(Felleman&vanEssen,1991;vanEssen&Gallant,1994;seediscussioninMundale,1998)useadditionalcriteriasuchasconnectivityandfunctiontodemarcatebrainareas.

Aschallengingasitistoidentifycandidateworkingparts,itisevenhardertoidentifythecomponentoperations.Theprobleminpartisthatidentifyingactualoperationsrequiresappropriateexperimentalinterventionsthatcanrevealevidenceaboutthem.Butevenmorefundamentalistodevelopconceptstocharacterizetheoperationspartsperform.Iwillillustratethechallengeandonewayithasbeenresolvedbydevelopingacasehistoryfrombiochemistry,acasethatisparticularlysuggestiveofthechallengesfacingresearchersinvestigatingpsychologicalmechanisms.

ABIOCHEMICALEXAMPLEOFTHECHALLENGEOFIDENTIFYINGOPERATIONS

Interestinthechemicalprocessesoperativeinlivingorganismshasalonghistory,buttheinvestigationwasradicallyreshapedattheendofthe18thcenturywhenLavoisierreconcep-tualizedwhatcountedasanelementandhenceachemicalbuildingblockofanysubstance.Hedeterminedthatcarbon,hydro-gen,andoxygenareconstituentsoforganicsubstances(Lavoisier,1781).Berthollet(1780)identifiednitrogenasanotherfrequentcom-ponent.Withthisfoundation,investigatorsbegantryingtocharacterizephysiologicalpro-cessesintermsofchangesinelementalcomposition(seeHolmes,1963).Forexample,Lavoisier(17)himselfcharacterizedfermen-tationasinvolvingtheoxygenationofcarboninpartofasugarmolecule,producingcarbondioxide,attheexpenseofthedeoxygenationoftheremaindertoyieldalcohol.ShortlythereafterLouisJacquesThe´nard(1803)andsubsequentlyJosephLouisGay-Lussac(1810)workedoutthegeneralequationforfermen-tation,representedinmodernsymbolismas

C6H12O6?2CO2z2C2H5OH:

Arrowsinchemicalequationssuchastheoneaboveindicatethatthechemistswerenot

THECHALLENGEOFCHARACTERIZINGOPERATIONS

317

interestedjustindeterminingtheelementalcompositionoforganicsubstancesbutincharacterizingorganicprocessesintermsofoperationsinvolvingchangesinelementalcomposition.Sincesuchreactionsdonotoccurspontaneouslyinordinaryenvironments(i.e.,thosetypicallyprevailingonthesurfaceoftheearth),somethingadditionalwasre-quiredtomakethemhappen.ThechemistJacobBerzelius(1836)namedtheresponsibleagentacatalystandmanychemistshopedthatcatalyticchemicalchangescouldaccountforthereactionsinlivingorganisms.Inthisregard,FriedrichWo¨hler’ssynthesisofureawasregardedasparticularlysignificant.Wo¨h-lerexpressedhisenthusiasmforhisaccom-plishmentinaletterhewrotetoBerzelius:‘‘Icannolonger,asitwere,holdbackmychemicalurine;andIhavetoletoutthatIcanmakeureawithoutneedingakidney,whetherofmanordog’’(quotedinFried-mann,1997,p.68).

Inthefirsthalfofthe19thcentury,theprospectofexplainingthechemicalreactionsinlivingorganismsintermsofchangesinelementalcompositionseemedpromising.Intermsofelementalcomposition,WilliamProut(1827)classifiedthenutrientsrequiredbyanimalsintothreeclasses:saccharine(carbo-hydrates),oleaginous(fats),andalbuminous(proteins).Proutalsonotedthattherewereonlyminordifferencesbetweenthechemicalcompositionofnutrientsanimalstookinfromplantsandthecompoundsthatcomprisedthefluidsandsolidsofanimalbodies.Perhapsthemostcelebratedchemistofthefirsthalfofthe19thcentury,JustusLiebigdrewuponthisideatoformulateacentralpartofhissyntheticandhighlyspeculativeaccountofthechemicaloperationsoccurringinanimalsinhisAnimalChemistry(1842).Sinceanimaltissuewaslargelycomposedofproteins,heproposedthatanimalssimplyincorporatedproteinintotheirtissueswhereastheyoxidizedthecarbo-hydratesandfatsintheirdiettogenerateheat.Wheninsufficientoxygenwasavailableforoxidizingcarbohydrates,Liebigproposedthatanimalsconvertedthemtofatandstoredthem.Heconjecturedthatwhenworkoc-curred,theproteinsincorporatedintotheanimalbodywerebrokendownandwasteproductsexcreted.Newproteinsthuswerecontinuallyrequiredinanimaldietstorebuildanimaltissues.Inthismanner,Liebigarticu-

latedageneralschemeforthechemicaloperationsoccurringinanimals,whichhefilledinwithdetailedformulae.

Theambitiousprogramoftheorganicchemistsofthefirsthalfofthe19thcenturysoonencounteredseriouscomplications.Notsurprisingly,giventhelimitedempiricalevi-denceuponwhichhebuilthistheory,Liebig’s(1842)proposalsfaredpoorlyasempiricalresultsemerged.Someofthisevidencere-mainedatthelevelofthewholeorganismandinvolvedfeedingexperimentsinwhichre-searchersmeasuredfoodintakeofvariousfoodgroups,resultingwasteproducts,andenergyexpenditure,anddemonstratedthatthesefailedtoconformtoLiebig’shypothesis.FickandWislicenus(1866),forexample,usedthemselvesassubjectsandpreparedtoclimbMt.FaulhornintheSwissAlpsbyconsuminganon-proteindiet.Theyalsomeasuredtheirurinebefore,during,andaftertheascent.Theycalculatedtheenergyexpendedontheclimbanddeterminedthatitgreatlyexceededtheamountaccountedforbythenitrogenwasteintheirurine.ContrarytoLiebig,theenergytheyexpendedontheclimbmusthavecomefromcarbohydratesandfats(otherfeedingstudieswereconductedbyFrankland,1866;Smith,1862).EquallyseriousforLie-big’sprojectwastherecognitionthatthechemicalreactionsinlivingorganismsweremorecomplexthanheanticipated.ClaudeBernard(1848),forexample,soughttotracewhereglucosewasconsumedinanimalsanddiscoveredthatitwasactuallysynthesizedintheliver.Thisshowedthatanimalmetabo-lismcouldnotbeunderstoodasalinearchainofcatabolicreactions.Identifyingthemorecomplexpatternofchemicalprocessesin-volvedinlivingsystems,however,waschal-lenging.

Inthesecondhalfofthe19thcentury,fermentationassumedacentralplaceinpointingtothelimitationsofattemptstoexplainphysiologicalprocesseschemically.UntiltheinvestigationsofKu¨tzing(1837),Schwann(1837)andCagniard-Latour(1838)thatindicatedthatalcoholicfermentationfundamentallyinvolvedlivingorganisms,mostchemistshadassumedthatfermentationwasanordinarychemicalreactionsimplyrequir-ingacatalyst.Leadingchemistsreactedharsh-lytotheclaimthatlivingorganismswereinvolvedsinceitseemedastepbackwardsin

318WILLIAMBECHTEL

theattempttoexplainphysiologicalprocesses.Wo¨hlerpublishedexcerptsofapaperbyTurpin(1838)followinguponCagniard-Latour’sresearchinAnnalenderPharmacie(ajournalheandLiebigedited),andfolloweditwithaheavy-handedsatireentitled‘‘Thedemystifiedsecretofalcoholicfermentation.’’Itpurportedtopresentdetailedobservationsmadewithaspecialmicroscopeoflittleanimalsshapedlikeddistillingflasksthathadcompletedigestivesystemsandeliminatedalcoholfromtheirintestinaltractafterdigest-ingsugar.ButthelinkageoffermentationwithlivingorganismswasfurthersecuredthroughtheinvestigationsofPasteur,whoconcluded‘‘Fermentationiscorrelatedtothevitalpro-cessesofyeast’’(Pasteur,1860,p.323).Asthechemistsfeared,thisseemedtoputfermenta-tionbeyondthereachofchemicalexplana-tion,andnearlyfortyyearsintervenedbeforeEduardBuchner(17)discovered,serendip-itously,thatfermentationcouldoccurinpressjuiceinwhichnowholecellsremainedandattributedittoacatalyst(catalystsexistingwithinlivingsystemsnowbeingdesignatedenzymes)henamedzymase.

Advancesinorganicchemistryalsoposedachallengetotheprojectofprovidinganaccountofphysiologicalprocessesintermsofchangesinelementalcomposition.Organicchemistsinthelaterdecadesofthe19thcenturydeterminedthatchemicalcompoundswerenotjustcomposedofatomsbutwerestructured.Aconsequentwasthatnoteverychemicalformuladesignatingacombinationofelementscorrespondedtoactuallyoccur-ringsubstances.Thisindicatedtheneedtoconsiderchemicalstructureinexplainingphysiologicalprocesses.

Thechallengewashowtodoso.Onesortofinvestigationorganicchemistspursuedwastodecomposeglucosewithanumberofalkalisintheattempttoidentifycompoundstructures,notelements,outofwhichitmightbecomposed.Researchersidentifiedseveralthree-carbonsugars—methylglyoxal,glyceral-dehyde,anddihydroxyacetone.Weretheseintermediatesintheprocessesinyeastthattransformedsugartoalcohol?Toanswerthisquestion,investigatorssuppliedthemtoafer-mentingsystem(yeastor,afterBuchner,cell-freeextracts)toseewhethertheywouldgeneratealcohol.Whatisparticularlyinterest-ingishowresearcherscharacterizedthese

investigations.Theyaskedwhethermethyl-glyoxal,forexample,wouldfermentasrapidlyassugar.Abandoningtheattempttoexplaintheprocessesinelementalterms,theynowcouldonlyusethesamevocabularyasappliedtotheoverallprocesstothepossiblecompo-nentoperations.

Thechallengeconfrontingthoseseekingtoprovidechemicalexplanationsofbasicphysi-ologicalprocesseswastocharacterizethecomponentoperations(reactions)atanap-propriateleveloforganization.Elementalcompositionwastoolowalevelatwhichtocharacterizechanges,whiledecomposingfer-mentationintofermentationssimplyinvokedthevocabularydesignedtoexplaintheoverallbehaviortodescribetheoperationofitscomponents.Itdidnotexplaintheprocessintermsofsomethingmorebasic.Fortunatelyfortheseresearchers,ataboutthissametimeanewframeworkbecameavailable.Organicchemists’effortstodeterminethestructureoforganiccompoundsrevealedthattheywerecomposedofgroupsofmoleculessuchasamino2(NH),and3+),carboxyl(COO2),hydroxyl(OHphosphate(POboundtoacarbon432)groupsthatwereringbackbone(Holmes,1992).Reactionswouldinvolvewholegroupsbeingadded,deleted,ormovedonthebackbone—suchasdeamination(removalofanaminogroup),carboxylation(additionofacarboxylgroup),dehydroxylation(removalofanhydroxylgroup),phosphorylation(additionofaphosphategroup),etc.Thisprovidedthebasisforconceptualizingtypesofreactionsatalevelabovethatofelementalcompositionandprovidedtheresourcebiochemistryneededtobeginworkingouttheintermediatestepsinnumerousphysiologicalprocesses.

Theviewofphysiologicalprocessesasin-volvingpathwaysofsuccessiveoperationsin-volvingchemicalgroups,togetherwiththeproposalthatthesereactionswerecatalyzedbyenzymes,providedtheguidingassumptionsofthenewlyemergingdisciplineofbiochem-istry.Forexample,oneofthebest-knownbiochemicalpathways,thecitricacidorKrebscycle,consistsofsuccessivestepsinvolvingoxidations(removalof2Hgroups,pickedupbyNAD+orFAD),hydrationsanddehydra-tions(addingorremovingH2Ogroups),decarboxylations(removalofCOorremovalofsulfhydryl-CoA2groups),additiongroups,etc.(seeFigure1).

THECHALLENGEOFCHARACTERIZINGOPERATIONS319

Fig.1.CitricacidorKrebscycle.Inthisprototypicalbiochemicalpathway,eachreaction(exceptforthecondensationreactionbetweenoxaloaceticacidandacetyl-CoA)involvesoperationsofaddingorremovinggroupsofmolecules(shownbyarrowscominginoroutoftheoverallcycle)fromtheprevioussubstrate.

Ultimately,thebiochemicallevelwasnottheonlylevelatwhichresearchershadtodiscoveroperationstodevelopacompletemechanisticaccountofbioenergetics.Theinvestigatorytechniquesofbiochemistryinvolveddestroy-ingcellstructurestosolubilizeenzymesinahomogenate,fosteringthesacofenzymesviewofthecell.Somecellprocesses,suchastheconversionofenergyfromfoodstuffsintoATP,whichprovidestemporaryenergystorage,dependonthestructureofcellorganellesaswell.Understandingthisleveloforganizationrequiredanothernewsetofresearchtech-niquesthatfiguredprominentlyinthede-velopmentofthenewfieldofcellbiology,inthe1940sand1950s(seeBechtel,2006).Understandinghowthenewstructuresfiguredinprocessessuchasthoseofbioenergeticsrequiredconceptualizingyetfurthertypesofoperationssuchasvectoraltransportacrossmembranes.Theprocessofidentifyingun-suspectedlevelsoforganizationsituatedbe-tweenexistinglevelsandconceptualizingthe

typesofoperationsthatoccurthereisare-curringstepinthedevelopmentofmechanis-ticexplanationinthelifesciences.

CHALLENGE:DETERMININGTHE

NATUREOFPSYCHOLOGICALOPERATIONSAsInotedattheoutset,whencognitivepsychologistsandcognitivescientistssetouttodiscoverthemechanismsresponsibleforbe-havior,theyfrequentlycharacterizedtheop-erationsinthesemechanismsusingconceptsdevelopedtodescribethebehaviorsinwhichcognitiveagentsengage.Thisperspectiveismostclearinthesymbolicorsymbolmanip-ulationapproachtomodelingcognitiveactiv-ity.Init,psychologicaloperationsareviewedastransformationsonsymbolstructures,wherethesesymbolstructuresareconstruedasbeingmuchlikesentencesinanaturaloraformallanguage.Fodor(1975)quiteappropriatelycharacterizedthesetheoristsascommittedto‘‘alanguageofthought.’’Theoperationsin

320WILLIAMBECHTEL

turnaremuchlikethosehumansthemselvesperformwhendoingsuchtasksaswritingamanuscript—typingwordsandphrases,readingthemback,alteringsome,etc.Themaindifferenceisthatthesesymbolsarethoughttobeencodedinsomewayinsideaperson’sbrain,andtheoperationsofreadingandwritingareinternaloperations,notoperationsonpaper.

Inthisregard,itisinterestingtonotethatTuring(1936;seealsoPost,1936),inpro-posingtheTuringmachineasacomputationaldevice,wasexplicitlytryingtomodelhumancomputers—humanswhoseoccupationwastocarryoutcomplexmathematicalcomputa-tions.Subsequently,theTuringmachineoftenhasbeeninvokedbyadvocatesofthesymbolicaccountastheexemplarforthekindofdevicethemindistakentobe.Inthisinstance,anactivityperformedbyhumansprovidedthemodelforoperationsoccurringintheirminds.Itshouldbeapparentthatsuchinvocationofsymbolprocessingtoexplainhowmindsworkiscomparabletophysiologicalchemists’in-vocationoffermentationsasintermediateprocessesinalcoholicfermentation.Thecomponentoperationswithinthepositedmechanismareofthesamesortasthebehaviorsofthemechanismitself.

Cognitivepsychologyisnotjustatheoreticalenterprisehypothesizinginternaloperations;likephysiologicalchemistry,itspractitionersofferempiricalevidencefortheirhypotheses.Thisevidenceisoftensecuredthroughbehav-ioralmeasuressuchasreactiontimes(Don-ders,1868).Earlycognitiveresearchinpsy-cholinguisticsprovidesanillustrativeexample.PsychologistsextendedChomsky’s(1957)pro-posalsforgenerativegrammar,developedinitiallysimplytoprovideacompactaccountofthestructureoflanguageitself,tocharac-terizetheoperationsperformedwhenpeoplecomprehendorconstructsentences.Sen-tenceswhosegrammaticalanalysisinvolvedmoretransformationswerehypothesizedtorequireadditionalpsychologicaloperations,whichwouldrequireadditionaltime.Reactiontimestudiesrevealedthatsentencesrequiringmoretransformationsinthegrammardidtakelongertoprocessthansentencesrequiringfeweroperations,suggestingthatthegram-maticaltransformationswerealsopsychologi-callyreal(Miller,1962;forhistoryandperspective,seeAbrahamsen,1987;Reber,

1987).Earlyresearchonmemoryexhibitedasimilarpattern.Sternberg(1966)compareddifferentmodelsofmemorysearch,whichallassumedthatmemoryinvolvedthestorageofsymbolicstructuresandmentallyscanningthem.Thesepredicteddifferentpatternsofreactiontimesandhearguedthatthemodelthatfitbestcharacterizedactualhumanpsychologicaloperations.

Oneofthemostpowerfultoolsforcon-structingartificialintelligencemodels,NewellandSimon’smethodofprotocolanalysis,madeconceptualizinginternalpsychologicaloperationsonthebasisofagent-levelbeha-viorsalmostinevitable.TheyrequiredsubjectstotalkaloudastheysolvedproblemssuchastheTowerofHanoiproblemsoastoelicitthestepsthesubjectsemployedinsolvingtheproblem.Theseoperationsthenbecamethebuildingblocksoftheircomputationalmod-els,whichwerethenfurthertestedbydatasuchasthatprovidedbyreactiontimemeasures(Newell&Simon,1972).Thepro-duction-systemarchitecture,whichbecamethefoundationforsomeofthemostpowerfulcomputationalmodelsofhumanperformance(Anderson&Lebiere,1998;Rosenbloom,Laird,&Newell,1993)developedoutofthisperspective.Thefundamentalideaofthisarchitectureisthatjustashumanagentshaveavarietyofstrategiesthatcanbeelicitedbytheproblemstheyaretryingtosolve(andpartialsolutionsalreadyobtained),theirmindsareassumedtobeequippedwithproductionsthatareexecutedwhenappropriatesymbolstringsareactiveinworkingmemory.

Itispossiblethatoperationswithinpsycho-logicalmechanismsdohavethesamecharac-terasthoseperformedbyhumanagents,butifsothisisaveryunusualcaseinthehistoryofscience.Typically,theoperationswithinamechanismthatenableittoperformitsbehaviorsaredifferentinkindfromthosebehaviors.Theabilityofmechanismstoperformbehaviorsdifferentfromthosethattheircomponentpartsperformiswhatmakesmechanisticexplanationssopowerful.Asnotedabove,organizationisthekeytoachievingthis.Althoughevolutionaryargu-mentsaresubjecttomuchabuse,aminimalappealtoevolutionenablesustonotethatdistinctivehumanbehaviorslargelyoriginatethroughreorganizationofcomponentsfoundinthebrainsofourcloseprimaterelatives.Itis

THECHALLENGEOFCHARACTERIZINGOPERATIONS

321

alsooperationsperformedintheseotherspecies,organizedinnovelways,thatpermithumanperformance.Itseemspeculiartoproposethatsymbol-processingcomponentswouldhaveevolvedinspeciesthathadyettodevelopthecapacitytomanipulatesymbols.Ifnotfromcharacterizationsofthebehaviorofhumans,wherecaninvestigatorsdrawinsightsastothenatureofinternalpsycho-logicaloperations?Theprimealternativetowhichtheoristshaveappealedisneuroscience.Suchwastheoriginoftheprimecompetitortothesymbol-processingparadigmincognitivescience.Duringthesameperiodasthesymbol-processingparadigmwasdeveloping,otherscientistsappealedtobasicideasabouthowbrainsworktoconstructanalternativeper-spective.Inthisalternative,partsofthemechanism(commonlycalledunits)passactivationstoeachotherandindividualcomponentsbecomeactivewhentheyreceivetheappropriateactivationfromotherunitstowhichtheyareconnected(McCulloch&Pitts,1943;Pitts&McCulloch,1947;Rosenblatt,1962).Althoughitencounteredseverelimita-tionsinitsfirstincarnation(Minsky&Papert,1969),theapproachreappearedinthe1980sunderthebannerofparalleldistributedproces-sing(PDP)orconnectionism(McClelland&Rumelhart,1986;Rumelhart&McClelland,1986;foranintroductiondesignedtobeaccessibletonon-specialists,seeBechtel&Abrahamsen,2002).

Althoughconnectionistaccountsdonotfacetheobjectionofusingthebehaviorstobeexplainedasmodelsfortheoperationsappealedtointhemechanismexplainingthem,theyexhibittheoppositeshortcomingofappealingtowhatislikelytobetoolowaleveloforganizationtocharacterizetheoperations.Recallthatintheearly19thcenturymanychemistsattemptedtoexplainphysio-logicalprocessesdirectlyintermsofelementalcomposition.Althoughitiscertainlytruethatchangesinelementalcompositionofsub-stratesoccurinphysiologicalprocesses,therelevantoperationsinvolvedhigher-levelmo-lecularunits.Likewise,operationswithinpsychologicalmechanismsinvolveneurons,buttheoperationsthemselveslikelyinvolvepartsatahigherlevelthanindividualneurons.Inperceptualprocessing,neuroscienceit-selfhasmadesignificantprogressinidentify-inghigher-levelstructures.Forexample,the

componentpartsincontemporaryaccountsofvisualprocessingarenotindividualneurons,butbrainareasinvolvingpopulationsofneurons.InvestigatorscharacterizeareassuchasV1,V4,andMTasextractingdifferenttypesofinformationfromtheinputsignal(edgesofobjects,shapeandcolor,motion)andmakingitavailabletoareasdownstreamforfurtheranalysis(vanEssen&Gallant,1994;seeBechtel,2001b,foranalysisandanhistoricalaccountofthediscoveryofvisual-processingmechanisms).Discoveringmechanismsofvi-sionwasfacilitatedbybothafruitfultechnique(single-cellrecording)andthefactthatthevisualsystemprocessessensoryinput.Al-thoughsingle-cellrecordingactuallyrecordsfromindividualneurons,itrevealedthatneuronsinaparticularareaallprocessedsimilartypesofinformationfromdifferentpartsofthevisualfield.Aswell,withineachregiontherewasinternalstructure:neuronsorganizedintocolumnsinvolvinglayersofconnectedunitsthatprocessinformationfromthesamepartofthevisualfield.Todeterminewhatsortofinformationagivenareaextractedresearcherscouldvarythestimuliandcorre-lateinputswithresponses.Inmanyrespectsthekindsofinformationthatvisualareasextractarewhatonemightexpectfromcharacterizingperformanceatthebehaviorallevel—peopleseecolors,shape,motion,etc.Butthedetailsareoftensurprising.Theshapesdetected,forexample,arefrequentlynotsimpleCartesianshapesbutrathermorecomplexforms,andthemotionsareoftennonlinear(vanEssen&Gallant,1994).

Thefactthatadvancesindiscoveringwhatinformationanareaprocessedresultedfromfortuneandnotfromhypothesesbeingtestedinexperiments,suggestsofhowharditistofigureoutthecomponentoperationsfrombehaviors.Hubel(1982)reportsthatheandWiesel,forexample,discoverededge-detect-ingcellsinV1whenaslidestuckintheprojectorfromwhichtheyhadbeenprojectingdarkspotsonlightbackgroundsandviceversa.Gross(1998)reportsthatheidentifiedshape-detectingcellsininferotemporalcortexwhen,infrustrationafterprojectingstimulitowhichthecellsdidnotrespond,hewavedhishandinfrontofthemonkey’sface.

Movingbeyondvisiontowhatarethoughtofashighercognitiveprocesses(reasoning,memory,languageprocessing)ismorechal-

322WILLIAMBECHTEL

lengingbecauseinvestigatorscannotsoreadilycontroltheoperationsapersonperformsatagiventimebycontrollingthestimulus.Whathasemergedasthedominantapproachforlinkingpsychologicalprocesseswithbrainactivityisfunctionalneuroimaging,inwhichinvestigatorsmeasurebloodflowchangesassubjectsperformtasks.But,asPetersenandFiez(1993)madeveryclear,theobjectinsuchresearchwasnottolocalizetasks,althoughinearlyimagingstudies,findingincreasedbloodflowinonlyoneorasmallnumberofbrainareasassubjectsperformedtasks,fueledsuchinterpretations.Asimagingtechniquesma-tured,neuroimaginghasbeguntoidentifymultiplebrainareascharacterizedasnetworksengagedinperformingthetask.Butwhatdoeseachareado?HereneuroimagingconfrontsthesameproblemIhavebeenfocusingoninthispaper—characterizingthecomponentoperations.

Biochemistrywasfortunateinthatstructuralinformationaboutorganicmoleculesprovideditwithinformationabouthigher-levelstruc-turesonwhichenzymesoperated.Cognitivescienceandcognitiveneuroscienceareun-likelytobeabletotakeadvantageinanydirectwayofinformationcomingfromthebrain,makingthechallengeofdiscoveringthenatureofthecomponentoperationsmuchgreater.Oneproblemisthatcorticalstructuresdonotvarymuch.(Thisisincontrasttosubcorticalareassuchasthehippocampus,whereeachsubregionhasadistinctivepatternofconnectivity,withtheparticularconnectivitypatternsuggestingthatoneareamaybeperformingatasklikepatterngeneralizationwhileanotherisperformingpatternsepara-tion.SeeRolls&Treves,1998,whohavedrawnuponsuchcluestodevelopananalysisofitsoperation.)

Icanforeseetwostrategiesthatmayhelpguidethediscoveryofappropriatelycharac-terizedcognitiveoperations.Oneisthedis-coverythroughtechniquessuchasneuroima-gingthatthesamebrainareasareinvolvedinmultipletasks,andthentryingtoassesswhatmightbecommonrequirementsofthediffer-enttasks.Theother,involvingcomparativepsychology—discoveryofthetasksinwhichrelatedspeciesuseareashomologoustothoseinourbrains—maylikewiseleadresearcherstoconsiderwhatoperationscontributetobothtasks(seeDeacon,19,1997,forprobing

suggestionswithrespecttotheoperationsinvolvedinhumanlanguageprocessing).Ultimately,however,thereisnosimplealgo-rithmfordiscoveringthetypeofoperationsintowhichthebehavioralsystemshouldbedecomposed.Theremaybenoalternativebutforcognitivescientiststoemployaccountsofoperationsdrawnfromwhatarelikelytoohighortoolowalevelwhileawaitinginspiredtheorizing.IfIamright,though,suchatheo-reticaladvanceisessentialifcognitivescienceistosucceedinthesearchformechanisms.WHYWORRYABOUTMECHANISMS?IfthechallengeindiscoveringcomponentoperationsisasgreatasIhaveproposed,onemightquestionwhetherthequesttodiscoverthemechanismsunderlyingbehaviorisworthpursuing.Perhapsitwouldbewiserforbehavioralscientiststolimittheirfocustoregularitiesdiscoverableinbehaviorandnotconcernthemselveswiththemechanismsthatunderliethem.InthisfinalsectionIwillfocusonthevalueofaccountsofmechanismsforunderstandingbehavioralregularities,aswellastheconversevalueofbehavioralaccountsforinvestigatorsseekingtounderstandin-ternalmechanisms.

Anunderstandingoftheresponsiblemech-anism,evenapartialandflawedunderstand-ing,canserveasavaluableguidetodevelopingandarticulatingfurthertheaccountoftheoverallbehavior.Ahypothesisaboutthemechanismcansuggestdifferentcontextualvariablesthataffectbehavior.Forexample,anunderstandingofthemechanismoftenwillpointtokindsoffactorsthatcanpromote,alter,ordisruptit.Thesefactorsmaybeexternaltothemechanism,butwithouthypothesesaboutthemechanism,theremaybelittlemotivationtoexaminehowtheyaffectbehavior.Nutritionresearchprovidesagoodexample.Althoughthefirstvitamin-deficiencydiseaseswereidentifiedpriortoresearchlinkingvitaminswithmetaboliccoenzymes,thatdiscoverybroughtamajorchangeinresearchonvitamins.Untilthen,vitamin-deficiencydiseaseswerequitemysterioussincedietarysubstanceshadbeenviewedasbeingburned(oxidized)toyieldenergy.Vitamins,however,wererequiredinsuchsmallquanti-tiesthattheycouldnotbeusedinthisway.Cofactorsinenzyme-catalyzedreactions,how-

THECHALLENGEOFCHARACTERIZINGOPERATIONS

323

ever,arenotconsumedbutreusedrepeatedlyandhencearerequiredinverysmallquanti-ties.Oncethevitamin-coenzymelinkwasestablished,anytimeanadditionalcofactorwasdiscovered,thatdiscoverymotivatedasearchforwhetherthatsubstancealsomightbeavitamin,andthusarequirementofanadequatediet(Bechtel,1984).

Similarly,understandingthepsychologicalmechanismsunderlyingbehaviorcanguideresearcherstonewdiscoveriesatthebehavior-allevel.Forexample,thediscoveryofdifferentwhatandwhereprocessingstreamsinthemechanismsunderlyingvisionledChen,Myerson,Hale,&Simon(2000)tolookforandfindbehaviordifferencesintasksthatemphasizedoneortheotherprocessingstream.Withoutthecluesfromthemecha-nism,therewaslittlereasontosuspectsuchbehavioraldifferences.Likewise,P.M.Church-land(inpress)usedinformationaboutoppo-nentprocessinginthevisualsystemtopredictanddemonstratethehumanabilitytoseewhathecallschimericalcolors—colorsbeyondthosefoundintheMunsellcolorspindleofcolorsusuallyexperienced.

Justasan(evenpartial)understandingofthemechanismcancontributetothefurtherdevelopmentofbehaviorallevelaccounts,so,too,behaviorallevelaccountsarecrucialtothedevelopmentofmechanisticaccounts.Tobeginwith,ifscientistsaretheorizingaboutamechanismtoexplainaparticularkindofbehavior,itisindispensabletobeginwithagoodcharacterizationofthebehavior.Otherwise,theymayproduceaproposalforapossiblemechanismthatdoesnotinfactexist,andwhosebehaviorwouldnotcorre-spondtoanythingthatactuallyhappens.Moreover,onceamechanismisproposed,theevidencefororagainstitcomesnotjustfrominvestigationsofinternaloperationsbutfromwhetheritactuallycanaccountforfactorsthatareknowntoaffectthebehavior.Liebig’s(1842)proposalforanimalchemistry(discussedearlier)failednotjustbecauseitfalselyruledoutsyntheticprocesseswithinorganisms,butalsobecauseitfalselypredicteddietaryrequirementsoforganisms.Thede-velopmentofaproposedmechanismdoesnotobviatetheneedforbehavioralinvestigationbutinfactisaspurtodiscoveryofthebehavioralregularitiespredictedfromtheproposedmechanism.

Iearliercharacterizedthesearchforme-chanismsasreductionisticinthatresearchersdecomposedmechanismsintotheirparts,whichareentitiesatlowerlevelsoforganiza-tion.Sometimesreductionisticresearchisthoughttounderminetheimportanceofresearchathigherlevelsoforganization—onceweunderstandthemechanismatthelowerlevel,itisthoughtthatnothingremainsforthehigherleveltoprovide(Bickle,2003).Butanunderstandingofmechanismsmakesclearwhysuchaviewismistaken.Thebehaviorofamechanismde-pendsasmuchonengagementwiththingsinitsenvironment(includingothermechan-isms)asonthepartsthatcompriseit.Fromtheperspectiveofscientists,amechanismisultimatelyamulti-levelintegrator,pro-vidingaframeworkforrelatinginformationaboutthecontextinwhichamechanismbehavesandtheinternaloperationofthemechanism.

CONCLUSION

Ihavefocusedonthechallengeofcharac-terizingtheoperationswithinthemechanismsthatunderliepsychologicalbehavior.Thechallengeisnotunique,asthediscussionofthehistoryofphysiologicalchemistryshows.Operationswithinamechanismaretypicallyofadifferentsortthanthebehaviorsofthemechanism—thisiswhyscientistsinvokeme-chanismstoexplainbehaviors.Thiswasamajorshortcomingofappealstofermenta-tionstoexplainfermentationandofmostaccountsincognitivepsychology,cognitivescience,andcognitiveneurosciencethatmod-elcomponentoperationsonsymbol-proces-singactivitiesperformedbypeople.Buttryingtoproceedfromneuralprocessingconfrontsthesamelimitationsastryingtoexplainphysiologicalprocessesintermsofelementalchemistry.Toaccountforthemechanismsunderlyingbehavior,investigatorsneedtodiscoveradistinctivesetofoperationsthatrelatestopsychologicalmechanisms,justasbiochemistshadtodiscovertheappropriatesetofcomponentoperationsunderlyingphys-iologicalprocesses.Asdifficultasthischal-lengeis,itisanimportantonetosurmountbecauseofthevaluablepayofftointegratingbehavioralaccountswiththoseofunderlyingmechanisms.

324

WILLIAMBECHTEL

REFERENCES

Abrahamsen,A.A.(1987).Bridgingboundariesversus

breakingboundaries:Psycholinguisticsinperspective.Synthese,72,355–388.

Anderson,J.R.,&Lebiere,C.(1998).Theatomiccomponents

ofthought.Mahwah,NJ:Erlbaum.

Bechtel,W.(1984).Reconceptualizationandinterfield

connections:Thediscoveryofthelinkbetweenvitaminsandcoenzymes.PhilosophyofScience,51,265–292.

Bechtel,W.(2001a).Cognitiveneuroscience:Relating

neuralmechanismsandcognition.InP.Machamer,P.McLaughlin,&R.Grush(Eds.),Theoryandmethodintheneurosciences(pp.81–111).Pittsburgh,PA:UniversityofPittsburghPress.

Bechtel,W.(2001b).Decomposingandlocalizingvision:

Anexemplarforcognitiveneuroscience.InW.Bechtel,P.Mandik,J.Mundale,&R.S.Stufflebeam(Eds.),Philosophyandtheneurosciences:Areader(pp.225–249).Oxford:BasilBlackwell.

Bechtel,W.(2006).Discoveringcellmechanisms:Thecreation

ofmoderncellbiology.Cambridge:CambridgeUniversityPress.

Bechtel,W.,&Abrahamsen,A.(2002).Connectionismand

themind:Parallelprocessing,dynamics,andevolutioninnetworks.(2nded.).Oxford:Blackwell.

Bechtel,W.,&Abrahamsen,A.(2005).Explanation:A

mechanistalternative.StudiesinHistoryandPhilosophyofBiologicalandBiomedicalSciences,36,421–441.

Bechtel,W.,&Abrahamsen,A.(inpress).Phenomenaand

mechanisms:Puttingthesymbolic,connectionist,anddynamicalsystemsdebateinbroaderperspective.InR.Stainton(Ed.),Contemporarydebatesincognitivescience.Oxford:BasilBlackwell.

Bechtel,W.,Abrahamsen,A.,&Graham,G.(1998).The

lifeofcognitivescience.InW.Bechtel,&G.Graham(Eds.),Acompaniontocognitivescience(pp.1–104).Oxford:BasilBlackwell.

Bechtel,W.,&Richardson,R.C.(1993).Discovering

complexity:Decompositionandlocalizationasscien-tificresearchstrategies.Princeton,NJ:PrincetonUniversityPress.

Bernard,C.(1848).Del’originedusucredansl’e´conomic

animale.Archivesge´ne´ralesdeme´decine,18,303–319.Berthollet,C.L.(1780).Recherchessurlanaturedes

substancesanimalesetsurleursrapportsaveclessubstancesve´ge´tales.Me´moiresdel’Acade´mieroyaledessciences,120–125.

Berzelius,J.J.(1836).EinigeKideenu¨berbeiderBildung

organischerVerbindungeninderlebendenNatur-wirksame,aberbishernichtbemerkeKraft.Jahres-Berichtu¨berdieFortschrittederChemie,15,237–245,

Bickle,J.(2003).Philosophyandneuroscience:Aruthlessly

reductiveaccount.Dordrecht:Kluwer.

Brodmann,K.(1909).VergleichendeLokalisationslehreder

Grosshirnrinde.Leipzig:J.A.Barth.Buchner,E.(17).AlkoholischeGa¨rungohneHefezellen

(Vorla¨ufigeMittheilung).Berichtederdeutschenche-mischenGesellschaft,30,117–124.

Cagniard-Latour,C.(1838).Memoiresurlafermentation

vineuse.Annalesdechimieetdephysique,68,206–223.Chen,J.,Myerson,J.,Hale,S.,&Simon,A.(2000).

Behavioralevidenceforbrain-basedabilityfactorsinvisuospatialinformationprocessing.Neuropsychologia,38,380–387.

Chomsky,N.(1957).Syntacticstructures.TheHague:

Mouton.

Churchland,P.M.(inpress).Chimericalcolors:Some

novelpredictionsfromcognitiveneuroscience.Philo-sophicalPsychology.

Churchland,P.S.,&Sejnowski,T.J.(1992).The

computationalbrain.Cambridge,MA:MITPress.

Craver,C.(forthcoming).Explainingthebrain:Whatascience

ofthemind-braincouldbe.

Craver,C.,&Bechtel,W.(2005).Explainingtop-down

causation(away).Manuscriptsubmittedforpublica-tion.

Deacon,T.W.(19).Theneuralcircuitryunderlying

primatecallsandhumanlanguage.HumanEvolution,4,367–401.

Deacon,T.W.(1997).Thesymbolicspecies.NewYork:

Norton.

Donders,F.C.(1868).Overdesnelheidvanpsychische

processen.OnderzoekingengedaaninhetPhysiolo-gischLaboratoriumderUtrechtscheHoogeschool:1868–1869.TweedeReeks,2,92–120.

Felleman,D.J.,&vanEssen,D.C.(1991).Distributed

hierarchicalprocessingintheprimatecerebralcortex.CerebralCortex,1,1–47.

Fick,A.E.,&Wislicenus,J.(1866).Ontheoriginof

muscularpower.PhilosophicalMagazine&JournalofScienceLondon,4thser.,31,485–503.

Fodor,J.A.(1975).Thelanguageofthought.NewYork:

Crowell.

Frankland,E.(1866).Onthesourceofmuscularpower.

RoyalInstitutionofGreatBritain.Noticesoftheproceedingsatthemeetingsofthemembers,4,661–685.Friedmann,H.(1997).FromFriedrichWo¨hler’surineto

EduardBuchner’salcohol.InA.Cornish-Bowden(Ed.),Newbeerinanoldbottle:EduardBuchnerandthegrowthofbiochemicalknowledge(pp.67–122).Valencia:UniversitatdeVale`ncia.

Gay-Lussac,J.L.(1810).Extraitd’unme´moiresurla

Fermentation.Annalesdechimie,76,245–259.

Glennan,S.(1996).Mechanismsandthenatureof

causation.Erkenntnis,44,50–71.

Glennan,S.(2002).Rethinkingmechanisticexplanation.

PhilosophyofScience,69,S342–S353.

Gross,C.G.(1998).Brain,vision,andmemory.Cambridge,

MA:MITPress.

Hempel,C.G.(1965).Aspectsofscientificexplanation.In

C.G.Hempel(Ed.),Aspectsofscientificexplanationandotheressaysinthephilosophyofscience(pp.331–496).NewYork:Macmillan.

Holmes,F.L.(1963).Elementaryanalysisandtheorigins

ofphysiologicalchemistry.Isis,,50–81.

Holmes,F.L.(1992).Betweenbiologyandmedicine:The

formationofintermediarymetabolism.Berkeley,CA:OfficeforHistoryofScienceandTechnology,UniversityofCaliforniaatBerkeley.

Hubel,D.H.(1982).Evolutionofideasontheprimary

visualcortex,1955–1978:Abiasedhistoricalaccount.BioscienceReports,2,435–469.Ku¨tzing,F.T.(1837).MicroscopischeUntersuchungen

u¨berdieHefeundEssigmutter,nebstmehrerenanderndazugeho¨rigenvegetabilischenGebilden.Journalfu¨rpraktischeChemie,11,385–409.Lavoisier,A.L.(1781).Me´moiresurlaformationde

l’acidenomme´airfixeouacidecrayeux,quejede´signeraide´sormaissouslenomd’acideducharbon.Me´moiresdel’Acade´mieroyaledessciences,448–458.

THECHALLENGEOFCHARACTERIZINGOPERATIONS

Lavoisier,A.L.(17).Traite´e´le´mentairedechimie,pre´sente´

dansunordrenouveauetd’apre`slesde´couvertesmodernes.Paris:Cuchet.

Liebig,J.(1842).Animalchemistry:ororganicchemistryinits

applicationtophysiologyandpathology.Cambridge:JohnOwen.

Machamer,P.,Darden,L.,&Craver,C.(2000).Thinking

aboutmechanisms.PhilosophyofScience,67,1–25.McClelland,J.L.,&Rumelhart,D.E.(Eds.)(1986),

Paralleldistributedprocessing:Explorationsinthemicro-structureofcognition.Vol.2.Psychologicalandbiologicalmodels.Cambridge,MA:MITPress.

McCulloch,W.S.,&Pitts,W.H.(1943).Alogicalcalculus

oftheideasimmanentinnervousactivity.BulletinofMathematicalBiophysics,7,115–133.

Miller,G.A.(1956).Themagicalnumberseven,plusor

minustwo:Somelimitsonourcapacityforprocessinginformation.PsychologicalReview,63,81–97.

Miller,G.A.(1962).Somepsychologicalstudiesof

grammar.AmericanPsychologist,17,748–762.

Miller,G.A.,Galanter,E.,&Pribram,K.(1960).Plansand

thestructureofbehavior.NewYork:Holt.

Minsky,M.,&Papert,S.(1969).Perceptrons:Anintroduction

tocomputationalgeometry.Cambridge,MA:MITPress.Mundale,J.(1998).Brainmapping.InW.Bechtel,&G.

Graham(Eds.),Acompaniontocognitivescience(pp.121–128).Oxford:BasilBlackwell.

Nagel,E.(1961).Thestructureofscience.NewYork:

Harcourt,Brace.

Neisser,U.(1967).Cognitivepsychology.NewYork:Apple-ton-Century-Crofts.

Newell,A.,&Simon,H.A.(1972).Humanproblemsolving.

EnglewoodCliffs,NJ:Prentice-Hall.Pasteur,L.(1860).Me´moiresurlafermentationalcooli-que.AnnalesdeChimie,3eSer,58,323–426.

Petersen,S.E.,&Fiez,J.A.(1993).Theprocessingof

singlewordsstudiedwithpositronemissiontomogra-phy.AnnualReviewofNeuroscience,16,509–530.

Pitts,W.H.,&McCulloch,W.S.(1947).Howweknow

universals:Theperceptionofauditoryandvisualforms.BulletinofMathematicalBiophysics,9,127–147.Post,E.L.(1936).Finitecombinatorialprocesses-FormulationI.JournalofSymbolicLogic,1,103–105.Prout,W.(1827).Ontheultimatecompositionofsimple

alimentarysubstances;withsomepreliminaryremarksontheanalysisoforganisedbodiesingeneral.PhilosophicalTransactionsoftheRoyalSocietyofLondon,117,355–388.

Reber,A.S.(1987).Theriseand(surprisinglyrapid)fall

ofpsycholinguistics.Synthese,72,325–339.

325

Rolls,E.T.,&Treves,A.(1998).Neuralnetworksandbrain

function.Oxford:OxfordUniversityPress.

Rosenblatt,F.(1962).Principlesofneurodynamics;perceptrons

andthetheoryofbrainmechanisms.Washington:SpartanBooks.

Rosenbloom,P.S.,Laird,J.E.,&Newell,A.(Eds.)(1993),

TheSoarpapers:Researchonintegratedintelligence.Cam-bridge,MA:MITPress.

Rumelhart,D.L.,&McClelland,J.L.(1986).Explorations

inthemicrostructureofcognition.Vol1.Foundations.Cambridge,MA:BradfordBooks,MITPress.Schwann,T.(1837).Vorla¨ufigeMitteilung,betreffend

Versucheu¨berdieWeinga¨rungundFaulnis.Pog-gendorf’sAnnalenderPhysikundChemie.41184–193.

Smith,E.(1862).Ontheeliminationofureaandurinary

water.PhilosophicalTransactionsoftheRoyalSociety,London,151,747–834.

Sternberg,S.(1966,August5).High-speedscanningin

humanmemory.Science,153,652–6.

Suppe,F.(1977).Thesearchforphilosophicalunder-standingofscientifictheories.InF.Suppe(Ed.),Thestructureofscientifictheories(pp.3–241).Urbana:UniversityofIllinoisPress.The´nard,L.J.(1803).Me´moiresurlaFermentation

vineuse.AnnalesdeChimie,46,294–320.

Turing,A.(1936).Oncomputablenumbers,withan

applicationtotheEntscheidungsproblem.ProceedingsoftheLondonMathematicalSociety,secondseries,42,230–265.

Turpin,P.J.F.(1838).Me´moiresurlacauseetleseffetsde

lafermentationalcooliqueetace´teuse.Annalesdechimieetdephysique,7,369–402.

vanEssen,D.C.,&Gallant,J.L.(1994).Neural

mechanismsofformandmotionprocessingintheprimatevisualsystem.Neuron,13,1–10.

Wimsatt,W.C.(1976).Reductionism,levelsoforganiza-tion,andthemind-bodyproblem.InG.Globus,G.Maxwell,&I.Savodnik(Eds.),Consciousnessandthebrain:Ascientificandphilosophicalinquiry(pp.202–267).NewYork:PlenumPress.

Wimsatt,W.C.(1986).Formsofaggregativity.InA.

Donagan,N.Perovich,&M.Wedin(Eds.),Humannatureandnaturalknowledge(pp.259–293).Dordrecht:Reidel.

ReceivedSeptember14,2004FinalacceptanceJuly8,2005