ResearchReportFacialStructureIsaReliableCueofAggressiveBehaviorJustinM - PDF document

ResearchReportFacialStructureIsaReliableCueofAggressiveBehaviorJustinM.CarreCherylM.McCormick,andCatherineJ.MondlochDepartmentofPsychologyandCentreforNeuroscience,BrockUniversityABSTRACT„Facialwidth-to-heightratioisasexuallydi-morphicmetricthatisindependentofbodysizeandmayhavebeenshapedbysexualselection.Werecentlyshowedthatthismetriciscorrelatedwithbehavioralaggressioninmen.InStudy1,observersestimatedthepropensityforaggressionofmenphotographeddisplayingneutralfacialexpressionsandforwhomabehavioralmeasureofag-gressionwasobtained.Theestimateswerecorrelatedstronglywiththefacialwidth-to-heightratioofthestim-ulusfacesandwiththeactualaggressionofthemen.TheseresultswerereplicatedinStudy2,inwhichtheexposuretoeachstimulusfacewasshortenedto39ms.ParticipantsestimatesofaggressionforeachstimulusfacewerehighlycorrelatedbetweenStudy2(39-msexposure)andStudy1(2,000-msexposure).These“ndingssuggestthatthefacialwidth-to-heightratiomaybeacueusedtopredictpro-pensityforaggressioninothers.Thehumanfaceisabasisforjudgmentsaboutgender,ethni-city,attractiveness,emotion,andpersonalitytraits(Zebrowitz,2006).Infact,personalityattributionsbasedonthecharacter-isticsofthefaceshowhighratesofconsensusacrossobservers(Penton-Voak,Pound,Little,&Perrett,2006;Todorov,Man-disodza,Goren,&Hall,2005),aremadeveryquickly(Bar,Neta,&Linz,2006;Willis&Todorov,2006),and,forcertaintraits,aresomewhataccurate(Penton-Voaketal.,2006).Peoplearerel-ativelygoodatidentifyingcheatersinaPrisonersDilemmagamebasedonfacialphotographs(Verplaetse,Vanneste,&Braeckman,2007),andwomensjudgmentsofmensinterestininfantsbasedontheirfacespredictedtheiractualinterestininfants(Roney,Hanson,Durante,&Maestripieri,2006).Thereissomeevidencethatbabyfacedness,characterizedbyroundfacesandbigeyes,isassociatedwithsocial,intellectual,andphysicalweakness(Zebrowitz,Fellous,Mignault,&Andre-oletti,2003).Also,peopleareaccurateinestimatingthephys-icalstrengthand“ghtingabilityofothersbasedonfacialinformation(Selletal.,2009),althoughthefacialmetricsusedtomakesuchjudgmentsarenotwellunderstood.Recently,Weston,Friday,andLio(2007)identi“edasexuallydimorphiccharacteristicofthefacethatwasindependentofbodysizefromamorphometricanalysisofanontogeneticseriesofhumanskulls.Theyfoundthatthegrowthtrajectoriesofmales AddresscorrespondencetoCherylM.McCormick,DepartmentofPsychologyandCentreforNeuroscience,BrockUniversity,500GlenridgeAve.,St.Catharines,Ontario,CanadaL2S3A1,e-mail: PSYCHOLOGICALSCIENCEVolume20„Number10Copyright2009AssociationforPsychologicalScience ParticipantsTwosamplesofundergraduatestudents(Study1:16women,15men;meanage19.94years,2.05years;Study2:16women,0men;meanage19.38years,1.41years)re-ceivedcoursecreditforparticipation.TheprocedureswereapprovedbyBrockUniversitysResearchEthicsBoard.Photographswereobtainedfromasampleof37menforwhichaggressivebehaviorandfacialWHRwasquanti“edpreviously(seeCarre&McCormick,2008).Thesemenwerevolunteersfromanintroductorypsychologyparticipantpoolwhoreceiveda$5honorariumandcoursecreditfortheirparticipation.Ag-gressivebehaviorwasmeasuredusingamodi“edversionofthepointsubtractionaggressionparadigm,awell-validatedlabo-ratorytask(e.g.,Cherek,Schnapp,Moeller,&Dougherty,1996).ThefacialWHRofthemenwasmeasuredusingNIHImageJsoftwareandinvolvedthelandmarksoriginallyusedbyWestonetal.(2007).Speci“cally,thedistancebetweentheleftandrightzygion(bizygomaticwidth)wasdividedbythedistancebetweentheupperlipandmid-brow(upperfacialheight)toyieldthefacialWHR(seeFig.1).Forthestudiesdescribedhere,thesampleofstimulusfaceswasreducedtoincludeonlyCaucasianmenwithoutfacialhair(toavoidjudgmentsbasedonstereo-types)anddisplayingneutralexpressions(24,meanage19.08years,1.41years).Faceswereconvertedto8-bitgrayscale,standardizedusingahairline-chindistanceof400pixels,andplacedwithinablackbackground.ProcedureInthe“rststudy,stimulusfaceswerepresentedonablackbackgroundusingE-Primesoftwareona14-in.LCDmonitor.Imageswereapproximately17cmwideby20cmhigh(or15.212.9visualdegreeswhenviewedfrom75cm).Faceswerepresentedinrandomorderfor1,000mstofamiliarizethepar-ticipantswiththerangeoffaces.Participantsweretoldhowtheaggressivebehaviorofthemenhadbeenassessed.Next,eachfacewaspresentedfor2,000ms(fullyrandomized),afterwhichthequestionHowaggressivewouldthispersonbeifpro-voked?appearedonablackbackgroundalongwitha7-pointLikertscale(1notatallaggressiveveryaggressiveParticipantsweregivenunlimitedtimetomaketheirresponseonanumericalkeypad,whichthencausedthenextstimulusfacetoappear.Aftercompletingtheestimatesofaggression,participantsratedeachfacefordominance,masculinity,trust-worthiness,andattractivenessusing7-pointLikertscales(orderofthesefourratingswasfullycounterbalanced).Insum,eachfacewaspresentedatotalofsixtimes,onceforfamiliarizationandonceforeachofthe“vetraits,andtheentiresetoffaceswasratedononeattributeatatime.Estimatesofaggressionwerehighlycorrelatedwithratingsofdominance(.92),mascu-linity(.86),trustworthiness(.90),andattractiveness.57).Onlyestimatesofaggressionandofdominancewerecorrelatedsigni“cantlywiththefacialWHR(.59and.54,respectively).However,giventhespeci“cevaluativecon-text(i.e.,participantsweretoldoftheobjectivemeasureofag-gression;Oosterhoff&Todorov,2008)andthehighdegreeofassociationofthe“veratings,statisticalanalysesusedestimatesofaggressiononly.Inthesecondstudy,presentationofastimulusfacewaspre-cededbyacentral“xationcrossthatappearedfor500ms.Afacewasthenpresentedfor39ms(orderoffaceswasfullyrandom-izedandwithoutanypriorfamiliarizationtothefaces),afterwhichthequestionHowaggressivewouldthispersonbeifprovoked?appearedonablackbackgroundalongwitha7-pointLikertscale(1notatallaggressiveveryaggressiveAsinStudy1,participantsweretoldhowtheaggressivebe-haviorofthemenhadbeenassessedbeforethepresentationofstimulusfaces.StatisticsCronbachsalphawascalculatedtoexaminetheconsistencyoftheratingsofestimatedaggressionacrossindividualpartici-pants.Foreachparticipantwithineachstudy,wecalculatedthecorrelationbetweentheestimateofaggressionforthe24facesandboththefacialWHRandactualaggressionofthestimulusfaces.One-sampletestswerecomputedtotesttheprimaryhypothesisthatthesecorrelationswouldbesigni“cantlydifferentfromthenullhypothesis(i.e.,noassociation).Foreachstimulusface,wecalculatedthemeanestimatedaggressionacrossparticipantsandcorrelatedthatwithboththefacialWHRandtheactualaggressionofthestimulusface.ThecorrelationbetweenestimatesofaggressioninStudy1andStudy2wasalsocalculatedforeachface.Signi“cancelevelwassetattwo-tailed,forallanalyses. Low-Ratio FaceHigh-Ratio FaceFig.1.ExamplesofstimuliusedinStudies1and2.Thefacesdifferinwidth-to-heightratio(i.e.,highandlowratios).Thelinesdrawnonthefaceswerenotshowntoobservers,andareincludedheretoillustratethelandmarksusedtomeasurethewidth-to-heightratio.Volume20„Number10 JustinM.Carre,CherylM.McCormick,andCatherineJ.Mondloch RESULTSStudy1Theestimatesofaggressionwerehighlyconsistentacrossindi-vidualobservers(Cronbachs.95).Single-samplecomparingindividualcorrelationstoanullvalueofzeroshowedthatestimatedaggressionwaspositivelyassociatedwiththefacialWHRofthestimulusfaces:maleobservers,(14)16.94,.001,rep.99;femaleobservers,(15)9.23,.001,rep.99;combined,(30)16.41,.001,rep.99(Fig.2).Thesetestsalsoindicatedthatestimatedaggressionwaspositivelyas-sociatedwithactualaggressionofthestimulusfaces:maleob-servers,(14)6.95,.001,rep.99;femaleobservers,(15)8.81,.001,rep.99;combined,(30)11.21,.001,rep.99(Fig.2).ThemeanestimatedaggressionforeachfaceacrossparticipantswasassociatedwithboththefacialWHR.59,.002,rep.98)andactualaggression(.42,.04,rep.89)ofthestimulusfaces(Fig.2).Study2Theestimatesofaggressionwerehighlyconsistentacrossin-dividualobservers(Cronbachs.89).Also,estimatesofaggressionamongtheseparticipantswerehighlycorrelatedwithestimatesofaggressionfromparticipantsinStudy1,whoweregiven2,000-msexposuretothestimulusfaces,.99(Fig.3).Single-sampletestscomparingindi-vidualcorrelationstoanullvalueofzeroshowedthatestimatedaggressionwaspositivelyassociatedwithboththefacialWHR,.99,andactualaggressionofthestimulusfaces,.99.Themeanestimatedaggressionforeachfaceacrossparticipantswasas-sociatedwiththefacialWHRbutnotwithactualaggression,.99,and.79(Fig.4).Theresultsindicatethatobserverscanmakeaccuratejudg-mentsofpropensityforaggressionfromfacesdisplayingneutral 3456 1234567Fig.2.ResultsfromStudy1:relationshipsbetweenobserversestimatesofthepicturedindividualsaggressionand(a)width-to-heightratio(WHR)ofthefacialstimuliand(b)thepicturedindividualsactualaggression.Thebargraphsshowtherelationshipsforeachindividualobserver(stimuliviewedfor2,000mseach).Blackbarsindicatefemaleobservers16);whitebarsindicatemaleobservers(15).Shadedareasrepresentthe95%con“denceintervals(CIs).ThevaluefortherelationshipbetweenestimatedaggressionandWHRwas.38(95%CI.33….43),andthemeanvaluefortherelationshipbetweenestimatedaggressionandactualaggressionwas.27(95%CI.21….31).ThegraphsontherightshowfacialWHRandactualaggressionasafunctionofestimatedaggression,acrossparticipants.Volume20„Number10 FacialStructureandAggressiveBehavior expressions,evenwhenexposuretothefacesislimitedto39ms.InbothStudy1andStudy2,individualparticipantsreliablyjudgedmenwithlargerfacialWHRsasmoreaggressive;acrossparticipants,faceswithlargerWHRswereratedasmoreag-gressivethanfaceswithsmallerWHRs.ThestrongcorrelationbetweenestimatedaggressionandfacialWHRsuggeststhatthisfacialmetricmaybeoneofthefacialcuesusedtomakeaccurateestimatesofaggression.ItisnotsurprisingthatparticipantsestimatesofaggressionwerecorrelatedmorestronglywithfacialWHRthanwithactualaggression.WhereasfacialWHRisastablefacialcharacteristicthatprovidesastaticestimateofthepropensityforaggression,anyactualbehavioralaggressionwillvaryovertime(e.g.,asafunctionofstateorsituation).Indeed,itisimpressivethatthecorrelationbetweenestimatesofaggressionandactualaggres-sionofthestimulusfaceswasashighinthecurrentstudy(.42and.31)aswasthecorrelationbetweenfacialWHRandactualaggressioninouroriginalstudy(.38;Carre&Mc-Cormick,2008).Furthermore,facialWHRisonlyoneofmanycuestopropensityforaggression. Fig.3.Relationshipbetweenobserversestimatesofthepicturedindi-vidualsaggressioninStudy1(2,000-msexposure)andinStudy2(39-msexposure).Theplottedpointsrepresentindividualstimulusfaces;eachpointshowstheaverageestimateofaggressionforastimulusacrossparticipantsinStudy1andacrossparticipantsinStudy2. 34561234567 Fig.4.ResultsforStudy2:relationshipsbetweenobserversestimatesofthepicturedindividualsaggressionand(a)facialwidth-to-heightratio(WHR)and(b)thepicturedindividualsactualaggression.Thebargraphsshowtherelationshipforeachindividualobserver(16;stimuliviewedfor39ms).Shadedareasrepresentthe95%con“denceintervals(CIs).ThemeanvaluefortherelationshipbetweenestimatedaggressionandWHRwas.44(95%CI.35….53),andthemeanvaluefortherelationshipbetweenestimatedaggressionandactualaggressionwas.19(95%CI.10….28).ThegraphsontherightshowfacialWHRandactualaggressionasafunctionofestimatedaggression,acrossparticipants.Volume20„Number10 JustinM.Carre,CherylM.McCormick,andCatherineJ.Mondloch FutureresearchshouldinvestigatewhetherthefacialWHRmeetsthecriteriaforanhonestsignalofaggressivepotential,similartohonestsignalsguidinginterindividualbehaviorinotherspecies(Setchell,Smith,Wickings,&Knapp,2008;Tib-betts&Dale,2004).Honestsignalsareusedinotherspeciesasameanstogaugeonesrelativestatuswithinthehierarchy(Setchelletal.,2008;Tibbetts&Dale,2004)andmayservetomodulateadaptivebehavior(Senar&Camerino,1998;Tibbetts&Lindsay,2008).Theabilitytogaugeaggressivebehaviorfromneutralfacesmayre”ectanovergeneralizationofemotionalexpressions(Montepare&Dobish,2003).Inotherwords,neu-tralfacesmaybeevaluatedaccordingtotheirsimilaritytocertainemotionalexpressions,suchasangerandhappiness,whichmayinturnbeusedbyperceiverstoguideadaptiveso-cialbehavior(e.g.,approach/avoidance;Oosterhof&Todorov,2008).Notably,angryfacialexpressionsconsistofloweringthebrowandraisingtheupperlip,afacialmovementthatinevitablyincreasesthefacialWHRand,byimplication,increasesthesaliencyofthesignaladvertisingpropensityforaggression.Thus,itisalsopossiblethattherelationshipbetweenfacialWHRandaggressionre”ectssocialconditioningwherebyapersonsaggressivebehaviorhasbeenshapedbyothersex-pectationsoftheiraggressivebehavior.Furthermore,itmaybesomeothercueinthefacecorrelatedwiththefacialWHRthatisin”uencingestimatesofaggression.Forexample,usingcom-puter-generatedfaces,Todorov,Baron,andOosterhof(2008)foundthatvariationinthebrowridge,cheekbones,chin,andnosesellion(i.e.,wherethenoseandbrowmeet)wererelatedtoob-serverratingsoftrustworthiness,asocialattributethatisnega-tivelycorrelatedwithaggression.Thus,itwillbeimportanttotestwhetherthefacialWHRpredictsaggressionindependentlyoftheseothercues.Nevertheless,thepresentresultsraisethepossibilitythatsubtledifferencesinfacialstructurein”uencetraitjudgments,whichmay,inturn,guidesocialbehavior.Acknowledgments„WethankMarkVidaandAlexandraHa-tryfortechnicalassistance.C.M.M.andC.J.M.holdNaturalSciencesandEngineeringResearchCouncil(NSERC)Discov-eryGrants.J.M.C.holdsanNSERCCanadaGraduateSchol-arship.REFERENCESBar,M.,Neta,M.,&Linz,H.(2006).Very“rstimpressions.,268…278.,J.M.,&McCormick,C.M.(2008).Inyourface:Facialmetricspredictaggressivebehaviourinthelaboratoryandinvarsityandprofessionalhockeyplayers.ProceedingsoftheRoyalSocietyB:BiologicalSciences,2651…2656.Cherek,D.,Schnapp,W.,Moeller,F.,&Dougherty,D.(1996).Labo-ratorymeasuresofaggressiverespondinginmaleparoleeswithviolentandnon-violenthistories.AggressiveBehavior,27…36.Montepare,J.M.,&Dobish,H.(2003).Thecontributionofemotionperceptionsandtheirovergeneralizationstotraitimpressions.JournalofNonverbalBehavior,237…254.Oosterhof,N.N.,&Todorov,A.(2008).ThefunctionalbasisoffaceProceedingsoftheNationalAcademyofSciences,,11087…11092.Penton-Voak,I.S.,Pound,N.,Little,A.C.,&Perrett,D.I.(2006).Personalityjudgmentsfromnaturalandcompositefacialimages:Moreevidenceforakerneloftruthinsocialperception.,490…524.Roney,J.R.,Hanson,K.N.,Durante,K.M.,&Maestripieri,D.(2006).Readingmensfaces:Womensmateattractivenessjudgmentstrackmenstestosteroneandinterestininfants.ProceedingsoftheRoyalSocietyB:BiologicalSciences,2169…2175.Sell,A.,Cosmides,L.,Tooby,J.,Sznycer,D.,vonRueden,C.,&Gurven,M.(2009).Humanadaptationsforthevisualassessmentofstrengthand“ghtingabilityfromthebodyandface.ingsoftheRoyalSocietyB:BiologicalSciences,575…584.Senar,J.C.,&Camerino,M.(1998).Statussignallingandtheabilitytorecognizedominants:Anexperimentwithsiskins(CarduelisProceedingsoftheRoyalSocietyB:BiologicalSciences,1515…1520.Setchell,J.M.,Smith,T.,Wickings,E.J.,&Knapp,L.A.(2008).Socialcorrelatesoftestosteroneandornamentationinmalemandrills.Hormones&Behavior,365…372.Tibbetts,E.A.,&Dale,J.(2004).Asociallyenforcedsignalofqualityinpaperwasps.,218…222.Tibbetts,E.A.,&Lindsay,R.(2008).VisualsignalsofstatusandrivalassessmentinPolistesdominuluspaperwasps.BiologyLettersTodorov,A.,Baron,S.G.,&Oosterhof,N.N.(2008).Evaluatingfacetrustworthiness:Amodelbasedapproach.SocialCognitiveandAffectiveNeuroscience,119…127.Todorov,A.,Mandisodza,A.N.,Goren,A.,&Hall,C.C.(2005).In-ferencesofcompetencefromfacespredictelectionoutcome.,1623…1626.Verplaetse,J.,Vanneste,S.,&Braeckman,J.(2007).Youcanjudgeabookbyitscover:Thesequel.Akerneloftruthinpredictingcheatingdetection.EvolutionandHumanBehavior,260…271.Weston,E.M.,Friday,A.E.,&Lio,P.(2007).Biometricevidencethatsexualselectionhasshapedthehomininface.PLoSONE,1…8.Willis,J.,&Todorov,A.(2006).Firstimpressions:Makingupyourmindafter100-msexposuretoaface.PsychologicalScienceZebrowitz,L.A.(2006).Finally,faces“ndfavour.SocialCognitionZebrowitz,L.A.,Fellous,J.M.,Mignault,A.,&Andreoletti,C.(2003).Traitimpressionsasovergeneralizedresponsestoadaptivelysigni“cantfacialqualities:Evidencefromconnectionistmodel-PersonalityandSocialPsychologyReview,194…215.1/19/09;REVISIONACCEPTEDVolume20„Number10 FacialStructureandAggressiveBehavior