Brain responses to facial attractiveness induced ... - Nature

Among these factors, averageness and symmetry were found to be important criteria. Attractiveness increases with an increasing level of ... Skiptomaincontent Thankyouforvisitingnature.com.YouareusingabrowserversionwithlimitedsupportforCSS.Toobtain thebestexperience,werecommendyouuseamoreuptodatebrowser(orturnoffcompatibilitymodein InternetExplorer).Inthemeantime,toensurecontinuedsupport,wearedisplayingthesitewithoutstyles andJavaScript. Advertisement nature scientificreports articles article Brainresponsestofacialattractivenessinducedbyfacialproportions:evidencefromanfMRIstudy DownloadPDF Subjects HumanbehaviourPerception AbstractBrainresponsestofacialattractivenessinducedbyfacialproportionsareinvestigatedbyusingfunctionalmagneticresonanceimaging(fMRI),in41youngadults(22malesand19females).ThesubjectsunderwentfMRIwhiletheywerepresentedwithcomputer-generated,yetrealisticfaceimages,whichhadvaryingfacialproportions,butthesameneutralfacialexpression,baldheadandskintone,asstimuli.Statisticalparametricmappingwithparametricmodulationwasusedtoexplorethebrainregionswiththeresponsemodulatedbyfacialattractivenessratings(ARs).TheresultsshowedsignificantlineareffectsoftheARsinthecaudatenucleusandtheorbitofrontalcortexforallofthesubjects,andanon-linearresponseprofileintherightamygdalaforonlythemalesubjects.Furthermore,canonicalcorrelationanalysiswasusedtolearnthemostrelevantfacialratiosthatwerebestcorrelatedwithfacialattractiveness.AregressionmodelonthefMRI-derivedfacialratiocomponentsdemonstratedastronglinearrelationshipbetweenthevisuallyassessedmeanARsandthepredictiveARs.Overall,thisstudyprovided,forthefirsttime,directneurophysiologicevidenceoftheeffectsoffacialratiosonfacialattractivenessandsuggestedthattherearenotablegenderdifferencesinperceivingfacialattractivenessasinducedbyfacialproportions. DownloadPDF IntroductionFacialattractivenessisafacialattributethatconveyssignificantbiologicaladvantagesasexpressedinmatingsuccess1,earningpotential2andlongevity3,acrossdifferentculturesandagegroups4.Facialattractivenessjudgmentevenexistsininfancy5.Alargebodyofstudieshavefoundthatfacialattributesthatcontributetoattractivenessincludeaverageness6,7,8,9,symmetry9,10,sexualdimorphism11,12,expression13,andskintexture14.Amongthesefactors,averagenessandsymmetrywerefoundtobeimportantcriteria.Attractivenessincreaseswithanincreasinglevelofaveragenessandsymmetry,whichcanbeunderstoodasevolutionarypressuresthatoperateagainsttheextremesofthepopulation15.Apartfromaveragenessandsymmetry,thesizesofindividualfeaturessignificantlyinfluencetheperceptionoffacialattractiveness.Previousstudiesthataddressmeasuringcertainfacialfeaturesormanipulatingindividualfeaturesizeshavefoundthatfemalefacesaremoreattractivewhenthefaceshavecertainfeatures,suchaslargeeyes,prominentcheekbones,thicklips,thineyebrowsandasmallnoseandchin16,17.Itisbelievedthatanaveragefaceisattractive,butnotoptimallyattractive,becauseattractivecompositefacescanbemadetobemoreattractivebychangingsomeofthefeaturesizestobedifferentfromthesamplemean18.Moreover,enhancedmasculinefacialcharacteristicsincreasebothperceiveddominanceandnegativeattributes(e.g.,coldnessordishonesty),whichsuggeststhathumanshaveaselectionpressurethatlimitssexualdimorphismandencouragesneoteny11.Inaddition,someoftheotherbehavioralstudieshavealsoattemptedtomakeclearwhetherthefacialmeasurementsshouldfollowcertaindefinedratiosforattractivefaces19,20.Forexample,menwithhigherwidth-to-heightratios(fWHRs)offacesarepositivelyassociatedwithhavingself-perceivedpower21,perceiveddominance,andattractivenesstowomenforshort-termrelations22,23.Arecentstudyreportedthattherepossiblyexistgoldenratiosoftheface’sverticaldistancebetweentheeyesandthemouthandofthehorizontaldistancebetweentheeyes,althoughdifferentfaceshavevaryingattractiveness24.Inparticular,Fanetal.25investigatedtheeffectoffacialproportionsonfacialattractivenessusingcomputergeneratedfaceimageswithcontrolledskintoneandexpression,andtheyfoundadifferentidealfacialproportionforfacialattractiveness.Despitethesubstantialnumberofbehavioralstudiesontheinfluenceoffacialproportionsorfeaturessizesonfacialattractiveness,littleisknownofhowhumanbrainsrespondtofacialattractivenessasinducedbyvaryingthefacialproportions.Pastneuroimagingstudiesusingrealfaceshavefoundthatsomeemotion-andreward-relatedregions,whichinvolvetheorbitofrontalcortex(OFC),anteriorcingulatecortex,nucleusaccumbens,caudate,andamygdala,respondtofacialattractiveness26,27,28,29.Morerecentmeta-analyseshaveobservedconsistentactivationstobeassociatedwithfacialattractivenessacrossneuroimagingstudies30,31,especiallyinputativerewardcircuitry,suchaslinearlyincreasedresponsesinthemedialorbitofrontalcortexandnucleusaccumbensforattractivefaces27,andnon-linearresponsesintheamygdalaforbothattractiveandunattractivefaces26.However,therearemultipledimensionalaspectsoffacialcharacteristicsthatinfluencefacialattractivenessperception,suchasfacialexpressions,hairstylesandskintones.Hence,itisextremelydifficulttodisentangletheeffectofanindividualfactorfromthecompositeeffectsofthesefactorsinnaturalfaceimages,despitearecenttrendinexploringthehigh-dimensionalnatureofneuralrepresentationsthatcharacterizesocialperceptioninnaturalandrealisticenvironments32.Inthepresentstudy,wethereforeusedcomputer-generated,yetrealisticfaceimagesthathadvaryingfacialproportionsasstimulitoinvestigatethebrainresponsestofacialattractivenessasinducedbyfacialproportions.Thisapproachallowedustoexploreonlytheeffectoffacialratiosonfacialattractivenessusinganevent-relatedfMRIdesign,bymanipulatingtheindividualfacialratiosandcontrollingtheeffectsofotherconfoundingfactors,suchasthehairstyle,skintextureandexpression,totestwhethersomeofthefacialratiosortheircombinationscontributetofacialattractiveness.Anotherobjectiveofthisstudyistoinvestigatewhetherthereisanydifferenceinthebrainresponsestofacialattractivenessasinducedbyfacialproportionsbetweenmaleandfemaleviewers.Behavioralstudieshaveindicatedthatmenassigngreaterimportancetophysicalattractivenesscomparedwithwomenwhenevaluatingapotentialmate33.Morespecifically,sexualpreferencesforsomefacialfeatures,suchaslipsize,mouthwidth,cheekbonesheightandchinsize,couldresultingenderdifferencesintheperceptionoffacialattractiveness20,34.Furthermore,ourrecentbehavioralstudieshavedemonstratedgenderdifferencesinrankingfacialphysicalattractiveness25.Infact,therehasbeensomeevidenceofsubjectgender-by-attractivenessinjudgingfaceattractiveness,especiallyinevaluatingopposite-sexfaces26,33,35.Forexample,increasedresponsesinsomereward-encodingregionssuchasOFC35andanteriorcingulate26wereobservedonlyformaleparticipants.Whenviewingmalefaces,womenshowedstrongerlineareffects,whilemenshowedstrongernonlineareffects,etc.Thesefindingssuggestthattherewardvalueoffacialattractivenessismorepronounceinmenthanthatinwomen,andsupportsexdifferencesinmateselectioninthatmenidentifyattractivenessasastrongermotivation35.Inspiredbythesepreviousfindings,wethereforespeculatethatthereexistgenderdifferencesbetweentheneuralresponsesofmaleandfemalebrains,whichunderliethebehavioraldiscrepancybetweengendersinthejudgmentoffacialgeometricattractiveness.MethodsFaceimagesasstimuliAllofthefaceimagesthatwereusedasstimuliinthisstudywereselectedfromourpreviousstudy25.Allofthemwerecreatedbyalteringcertainkeyfacialdimensionsusingcomputersoftwarefromasocalled“OriginalFace”(SeeFig.1),whichwasobtainedbyaveragingthefeaturesofthefacialimagesofsomearbitrarilychosenfamousorientalladies(Japanese,KoreanandChinese)thatwereavailableontheinternet.Furthermore,thefaceswerehairlesstoeliminatetheeffectsofhairstyles,andtheywerecontrolledtohavethesameneutralfacialexpressionandskintone.Finally,wecreatedsufficientbutrealisticvariationsinthefacialdimensionsandratiosofthefacialimages,byapplyinggradualalterationtothe“OriginalFace”withninedifferentapproaches(SeeFig.2andalsorefertoref.25fordetailsofthemethod).Duringeachapproaches,themaximumpossiblealterationofthetargeteddimension(s)withoutbeingoverlyunrealisticwasdeterminedbyapreliminaryvisualassessment.Then,thechangeineachstepofthealterationswascalculatedbydividingthemaximumpossibledimensionalalternationby14.Thisstrategyresultedin14progressivefacialimagesineachapproach.Theabovenineapproachescreatedwidevariationsinthelengthandwidthoftheface,thelengthofthenose,andthepositionsoftheeyes,whilekeepingotherfeaturessuchasthesizeoftheeyes,thewidthofthenose,andthesizeandshapeofthemouthunchanged.Furthermore,basedonthe“originalface”andtheselected18facialimagescreatedbytheabove9approaches,themouthwidth,eyefissurewidth,nosewidthandlipheightwerealteredinapre-definedproportionbytheorthogonalexperimentaldesign25.Usingthismethod,atotalof432faceimageswerecreatedforfurtherstudy.Figure1The“OriginalFace”andasetof29faciallandmarksusedinthegeometricmorphometricsanalysis.FullsizeimageFigure2Thealterationsofthe“OriginalFace”togeneratetheotherfacialimages.(A)demonstratesninedifferentapproaches(a–i)ofchangingthefacialratios(refertoFanetal.,25fordetail).(B)Examplesoffacialimagesgeneratedforthisstudy.FullsizeimageTheaboveapproachofmanipulatingindividualfeaturesizesprovidesasimplebutefficientwayinwhichwecansearchtheentirespaceoffaceshapesstartingfromthemeanfaceofacohort,withminimizationofthesamplesizeandpreservationofthesymmetry.Unlikepreviousstudiesthatintentionallyexaggeratemasculineorfemininefacialcharacteristics11,variationsinthefacialproportionsorfeaturesizeswerecreatedbyastepwiseprocedure(e.g.,−10%,−5%, + 5%and + 10%forthemouthwidth),regardlessofthesexualdimorphismdimension.Moreover,thelevelsofproportionalchangesforallofthefacialratioswereoptimallydetermined,tohaveawiderrangeofvariationbutwithoutcreatingtoounrealisticimages.Thesestrategiesforfaceimagegenerationfacilitatetheinvestigationofonlytheeffectoffacialratiosonfacialattractivenessasindependentvariables.Allofthefaceimageshavevariablegeometricmorphometricparametersthatconsistofmorphologicaland/orfunctionalpointsaswellascontoursoftheeyes,noseandmouth.Basedonthepreviousliterature36,wefurtheridentified29landmarks(SeeFig.1)foreachsampleimageusingcomputersoftware,andwegeneratedthe21ratioslistedinTable1bymeasuringtheverticalorhorizontaldistancesbetweentheselandmarks.Themeasuresofthe21ratioscomposethelow-levelfeaturesthatresultinfacialattractivenessforanindividualfaceimage.Furthermore,foreachface,itsdistancefromthe‘originalface’wascalculatedbytheEuclideandistancebetweenitsratiofeaturesandtheratiofeaturesofthe‘originalface’.Table1Definitionoffacialratios.FullsizetableHumanSubjectsfMRIdatawerecollectedfrom41healthysubjects(22males)withnormalorcorrected-to-normalvisionwhogavewritteninformedconsenttoparticipateinthisstudy.Allofthesubjectswererevealedtohavenon-homosexualpreferencesbyusingadebriefingquestionnaire.Therewasnosignificantdifferencebetweentheagesofthemalesandfemales(age:23.57 ± 1.10,p = 0.745).Alloftheparticipantswereingoodhealthandhadnopasthistoryofpsychiatricorneurologicaldiseases.Only36subjects(19males)wereavailableforfurtheranalysisafterheadmotioninspection(seetheDataPreprocessingsection).ThisstudywasconductedaccordingtotheprinciplesintheDeclarationofHelsinkiandapprovedbytheEthicsCommitteeoftheThirdXiangyaHospitalofCentralSouthUniversity,Changsha,China.ExperimentalparadigmEachhumansubjectunderwentanexplicitjudgmenttaskoffacialattractivenessonthefacestimulus.Faceimages(800 × 600pixels;subtending23 × 18°ofvisualangle)werepresentedinthecenteroftheSVGAdisplay,usinganMR-compatibleliquidcrystaldisplaygoggleoperatingataresolutionof800 × 600pixels,at60 Hz.ThestimuliseriesweregeneratedwiththeE-Primesoftware(http://www.pstnet.com/eprime.cfm).Thesubjectswerescannedinfourruns,andeachruninvolved108facepresentationtrialsinwhichafaceimagewaspresentedfor1 secondfollowedbyafixationcrossforafurther2 seconds,andthesubjectswereinstructedtomaintainfixation.Ineachrun,54nulleventtrials,inwhichafixationcrosswaspresentedfor3 seconds,wererandomlyinterspersedwiththefacepresentationtrials.Thesefaceimages(432facesintotal)wereselectedonceforallofthesubjects,andwererandomlypresentedduringthescanning.Thescannerwasintheacquisitionmodefor15 secondsbeforeeachseriestoachievesteady-statetransversemagnetization.Followingthepresentationofeachface,thesubjectswereinstructedtojudgetheattractivenessbypressingoneoffourbuttonswiththeirrighthands(denoting“highlyattractive”and“attractive”)andtheirlefthands(denoting“unattractive”and“highlyunattractive”)onaresponsekeypad,respectively.Theirresponsetimefromthepresentationofafacetothepressingofabuttonwasalsorecorded.fMRIdataDatawerecollectedona3.0TMRIscanner(Philips,MRIngenia,theNetherland)attheThirdXiangyaHospitalofCentralSouthUniversity,HunanProvince,China.Toreducetheirheadmotion,theparticipants’headswerefixedusingfoampadswithastandardbirdcageheadcoil.Thefunctionalimageswerecollectedusingagradient-echoT2*-weightecho-planarimaging(EPI)sequencewiththefollowingparameters:repetitiontime(TR) = 3000 ms,echotime(TE) = 40 ms,fieldofview(FOV) = 240 × 240 mm,flipangle(FA) = 90°,slicethickness = 2 mm,gap = 1 mm,andmatrix = 88 × 86.Thevolumesconsistedof36slicesangledat−30°tothehorizontal,whichcouldimprovethesignalqualityintheventralprefrontalcortexandamygdala26.Eachsessionlasted8 minand15 seconds,and165volumeswereobtained.Finally,T1-weightedanatomicalimageswereacquiredforeachsubjectfordetailedanatomicalinformation.DebriefingTheparticipantsundertookonedebriefingtaskoutsidethescannerimmediatelyaftertheexperiments.Theywereaskedtorateallofthefacesintermofattractiveness,usingacomputerizedvisualanaloguescale.Thescalewasmarkedwiththeextremesof“highlyunattractive”and“highlyattractive”,withthemid-pointmarked.Theratingswerescaledbetween1and9,with9representingratingsofthehighestattractiveness.Themeanattractivenessrating(AR)foreachimagewasobtainedbyaveragingtheARsoverallofthesubjects.DatapreprocessingImagingdatawerepreprocessedandanalyzedusingSPM8(WellcomeDepartmentofImagingNeuroscience,London;http://www.fil.ion.ucl.ac.uk/spm).Foreachparticipant,thefirstthreevolumesofscanneddatawerediscardedbecauseofmagneticsaturationeffects.Theremainingvolumeswererealignedtothefirstimageforcorrectingtheheadmotion,andameanfunctionalimagewascorrespondinglyobtained.Threemaleandtwofemaleparticipantswhoseheadmotionexceeded1.8 mmwerediscarded,andtheresidual36participantshadlessthan1 mmoftranslationinthex,y,orzaxisand2°ofrotationineachaxis.TheimageswerethennormalizedtoanMNIechoplanarimagingtemplatewithanaffineregistrationfollowedbyanonlineartransformation,usingavoxelsizeof3 × 3 × 3 mm.Finally,a6-mmFWHMGaussiankernelwasusedtosmooththedata.StatisticalparametricmappinganalysisThefunctionalimageswereanalyzedstatisticallyusingatwo-stageproceduretoexplorethebrainresponseswhilejudgingtheattractivenessoffacesthathadvaryingfacialproportions.Inthefirststage,weestimatedtheeffectsizesforeachregressorinthedesignmatrixwithintheframeworkofthegeneralizedlinearmodel,withparametermodulationforeachsubject.Toidentifythebrainregionsthatshowedalinearornonlineareffectofattractivenessratings,weusedaparametricapproach37thatallowsonetocharacterizebrainresponsesasalinearcombinationof(basis)functionsoftheexperimentalparameter(e.g.,f(x) = a + bx + cx2,wherexisthesubject-specificattractivenessratings).Weusedafirst-orderandasecond-orderpolynomialexpansion,respectively,asthebasisfunctionsinwhicheachpolynomialtermrepresentsaninteractionbetweentheattractivenessratingsandaboxcarstimulusfunctionbox.Thestandardboxcarfunction(i.e.,zerothorder)modesthedifferencebetweenthefacepresentationandthesilentbaselinecondition,irrespectiveofthefaceattractiveness.Thelinear(i.e.,lin = r·box)termandthequadratic(i.e.,sec = r2·box)termaccountforthelinearandsecond-ordernonlinearchangesinthebloodoxygenleveldependent(BOLD)signalrelativetotheparameteroffacialattractivenessratings,respectively.Priortomodelfitting,theseterms(asparametricmodulationsinSPM8)wereconvolvedwithahemodynamicresponsefunction.Inthedesignmatrixofthefirstlevelmodel,thecovariatesofinterestincludethelineartermandthequadraticterm,bothofwhichmodelallofthetrialsononeregressorbutincludesalinearorsecond-ordernonlinearparametricmodulationofthisregressorbyusingsubject-specificratingsforeachimage.Alinear(f(x) = a + b·AR)andaquadraticnonlinearpolynomialfunction(f(x) = a + b·AR + c·AR2)ofattractivenessratings(ARs)wereusedtoestimatetheregionsthathadlinearandnonlineareffectsoffacialattractiveness,respectively.Inthelinearexpansionmodel,theremainingcolumns(confoundsorcovariatesofnointerest)includethezeroth-orderterm(modelingthemaineffectoffacepresentationirrespectiveofattractiveness),aconstantterm,regressorsthatpertaintothereactiontimeforeachtrialandsixheadmovementparameters.However,inthequadraticexpansionmodel,thelineartermofattractivenessratingswasalsomovedintothecovariatesofnointerest.Theobtainedeffectsofthelineartermandthesecond-ordertermacrossthepopulationofsubjectsweresubmittedtothesecondstageofanalysis,inwhichthesignificanceofthelinearandnonlineareffectswastestedbyusingrandomeffectsanalyseswithANOVAmodelsinSPM8.Inthesecondlevelanalysis,menandwomenweretreatedasasinglegroup,andaone-samplet-testwasusedtoevaluatethelinearorsecond-ordernonlineareffectsofthefacialattractivenessirrespectiveofthesubjectgender.Additionally,forestimatingtheeffectofthesubjectgender-by-attractiveness,themalesandfemaleswerealsotreatedasseparategroupsandatwo-samplet-test(withcontrastsof1forthemalesand−1forthefemales)wasappliedtoinvestigatethegenderdifferenceinthebrainresponsestofacialattractiveness.Alloftheactivationwasreportedbyusingthemoreliberalthresholdofp 5voxels)inthecaudatenucleus(Fig.3A),inferiorfrontalgyrus(Fig.3B)andrightpostcentralgyrus,aswellasasignificantnegativelineareffectintheleftpostcentralgyrus.Figure3Lineareffectsoftheattractivenessratingsforallofthesubjects.(A)demonstratesthelineareffectoftheattractivenessratingsinthecaudatenucleus(leftpanel)andtherelativechangeintheBOLDsignalsacrossfourdifferentgroups(rightpanel),withgroup1representingthehighlyunattractivefacesand4denotingthehighlyattractivefaces.(B)Theorbitofrontalcortex(OFC)showssignificantpositivelineareffectsofattractivenessratings(leftpanel).Inotherwords,moreattractivefacesevokestrongeractivationintheOFC(rightpanel).Here,p 0.001).Theresidualsixcolumnsrepresentthemostrelevantcomponentsoffacialratiostofacialattractiveness,whicharelistedinTable5.Fromthecorrelationsbetweenthesixcomponentsandtheratios,itcanbelearnedthatsomeoftheratioshavehigherweightsindeterminingthefacialattractiveness.Forexample,thefirstandsixthcomponentsarelargelyrelatedtothenosewidthandtheinteroculardistance.Thesecondandfourthcomponentsarelargelyrelatedtotheratiosofmideyedistanceandtheinteroculardistancetothenosewidth,andsoon.Inparticular,multiplecomponentsthatinvolvethecomponents1,3,5and6indicatethattheratiooftheearlengthtotheinteroculardistanceandtheratioofthelips-chindistancetotheinteroculardistanceplayimportantrolesinfactorsthatinfluencethefacialattractiveness,whichislargelyinlinewiththeresultsfrombehavioralexperiments25.Overall,theearlength,nosesize,interoculardistanceandlip-chindistanceareofmoreimportanceindeterminingthefacialattractivenesscomparedwiththeotherfeaturesizes.Wehavealsovisualizedtheeffectsofthesecomponentsbyaveragingthe20facesthathavethehighestandlowestscoresonthem(seesupplementalFig.S1).Table5CCAtransformationmatrixforthefacialattractivenessratings.FullsizetableEachcomponentisplottedagainsttheattractivenessratingsinFig.5.Byapplyingnonlinearregression,weobtainedthecoefficientsofmodel(4)andlistedtheminTable6.Figure6plotsthemeanattractivenessratingsofthe432facialimagesagainstthepredictedattractivenessratingsusingaregressionequation(6).Itisevidentthatthepredictedattractivenessratingsaresignificantlylinearlyrelatedwiththevisuallyassessedmeanattractivenessratings(R2 = 0.58,p