Cognitive and Neurobiological Mechanisms of the La

CognitiveandNeurobiologicalMechanismsoftheLawofGeneralIntelligenceChristopherF.ChabrisDepartmentofPsychologyHarvardUniversityCambridge,[e-mail]~cfc5November2006ToappearasChapter19of:Roberts,M.J.(Ed.)(2006).Integratingthemind:Domaingeneralversusdomainspecificprocessesinhighercognition.Hove,UK:PsychologyPress.IntroductionWhydoscoresonmostcognitiveabilitytestscorrelatepositively?Thefactthatpeoplewhoscorehighlyononetesttendtoscorehighlyonothers—thatsomepeoplearemoreintelligentthanothers—isnowsointuitivelyobviousthatmostpsychologists,atleastsinceitwasfirstdiscoveredbySpearman(1904),takeitforgranted.Theyarenotinterestedintalkingabout,muchlessthinkingabout,letaloneinvestigating,individualdifferencesincognitiveability.Andthereisnothingwrongwiththis;researchtopicsandapproachesfadeinandoutofpopularity.EventheharshestcriticsofIQprofessnosurpriseatthepositivecorrelationsamongtests;StephenJ.Gould(1981,p.315)wrote,“Thefactofpervasiveintercorrelationbetweenmentaltestsmustbeamongthemostunsurprisingmajordiscoveriesinthehistoryofscience.”Butifwecouldstepoutofournormalmodesofthinking,ifwecouldfreeourmindsofthe“debaucheryoflearning”thatWilliamJamesfamouslynoted(1890,asquotedbyCosmides&Tooby,1994),wemightwonderwhythereshouldbeanycorrelationamongcognitivetests,letalonethestrong,consistent,positivecorrelationsfoundinhundredsofstudiesoverthepastcentury(Carroll,1993).Evolutionarypsychology,withitsemphasisonspecialisedadaptationsandinstinctsforgedinalong-agostoneage,haslittletosayaboutcurrentvariationinintelligence.Fewtheoriesincognitivepsychologyspecifyhowtheircomponentprocessesorrepresentationsmightvaryfrompersontoperson.Neurosciencelikewisegivesshortshrifttoindividualdifferences,thoughitwouldappeartocontainwithinitsranksfeweroutrightopponentsoftheveryconcept.NotwithstandingthepleaofCronbach(1957)forunificationofthe“twodisciplines”ofpsychology,instancesofcollaborationbetweenstudentsofhumancommonalitiesandstudentsofhumandifferences,orinstancesofindividualresearcherswhoeffectivelyoperatewithinbothofthesecamps,arestilltheexception,nottherule.ThosewhoemploytheAnalysisofVariancerarelyactuallyanalysevariance(Plomin&Kosslyn,2001).Theubiquitouspositivecorrelationsamongcognitivetestsareachallengefortheoriesofhowthemindworksthatfocusonindependent,domain-specific,encapsulatedprocesses,or“modules”Chabris–TheLawofGeneralIntelligence–p.2of40(Fodor,1983)astheelementaryunitsofstudy.Thisisespeciallysoifonebelievesthatthesemodulesmusthaveevolvedtosolvecriticalproblemsrapidlyandefficiently(Cosmides&Tooby,1994):Whyshouldtherebeanycorrelatedvariationintheefficacyofallofthesecarefullydesignedprocesses?Paradoxically,thenotionofintelligenceasanindividualdifference,afactthatmosthumanstakeforgranted,isoutsidethepredictivepowerofoneoftheleadingtheoreticalframeworksinmoderncognitivepsychology.Twogeneralapproacheshavebeentakentowardreconcilingmodularitywithgeneralintelligence.Oneistoarguethatthegeneralfactorrevealedbythecorrelationsamongcognitivetests—thegfactor—isapsychometricallyunitaryconstruct,butthatitiscausedbymultiplebiologicalfactors(Jensen,1998a).InalettertoCommentary(includedinChabrisetal.,1998,pp.20–21),ArthurJensenelaboratedonthishypothesis:…thedesignfeaturesofthebrain—itsneuralstructuresandfunctions—thatarenecessaryforthemanydistinctprocessesthatenterintoinformation-processing,orintelligence(suchasattention,perception,discrimination,generalization,learning,memory,language,thinking,problem-solving,andthelike)areessentiallythesameforallbiologicallynormalHomosapiens,i.e.,thosefreeofchromosomalandmajorgeneanomaliesorbraindamage.Correlatedindividualdifferencesinthefunctioningofthesevariousinformationprocessesarearesultofotherquantitativebiochemicalandphysiologicalconditionsinthebrain,mostofthemhighlyheritable,thatareseparatefromthebrain’sessentialdesignfeatures,or“hard-wiring,”butare,asitwere,superimposedonallofthemincommon,andaffecttheoverallspeedandefficiencyoftheirfunctioning.Thedomain-general“speedandefficiency”ofneuralfunctioningcanbeassessedinseveralways:bymeasuringspeedofresponseinsimplecognitivetasks,bymeasuringspeedofperceptualintakeviaasimplepsychophysicaltask,bymeasuringnerveconductionvelocity,andthelike.Globalneuralresourcescanbequantifiedbymeasuressuchastotalbrainvolume,totalgreymatter,andtotalwhitematter.Indeed,numerousstudiessupportthecorrelationbetweenthesesortsofmeasuresandperformanceonintelligencetests,especiallythegeneralfactor.Thealternativeapproachisthatcertainneuralsystems,perhapsonesthatareonlysemi-modular,areespeciallyrelatedtogbecausetheyareresponsibleforcontrollingtheoperationofotherprocesses,orformanaginglimitedcentralresourcesthatotherprocessesrequireinordertofunctionoptimally.Variabilityintheefficacyofthesespecialsystemscouldaccountforg(e.g.,Kyllonen&Christal,1990).Asothershavenoted,thisviewencompassestwodistinctideas:thatofalimitedpoolofcentralmemorycapacity,perhapsalsoconstrainingtheactionofdomain-specificprocesses,versusthatofalimitedabilityofacentralprocesstocontroltheoperationo