Inferring the Popularity of an Opinion From Its Famili - PDF document

KimberleeWeaver,R.B.PamplinCollegeofBusiness,VirginiaPoly-technicInstitute;StephenM.Garcia,GeraldR.FordSchoolofPublicPolicy,UniversityofMichigan;NorbertSchwarz,DepartmentofPsychol-ogy,InstituteforSocialResearch,andRossSchoolofBusiness,UniversityofMichigan;andDaleT.Miller,GraduateSchoolofBusiness,StanfordAnearlierversionofthisarticlewasawardedtheFrancoNicosiaAward detailsoftheexposureepisodebecomelessaccessible,makingitlesslikelythatperceiversrealizethatallstatementscamefromthesamecommunicator.Finally,(v)theimpactofrepetitionshouldbemostpronouncedwhenperceiverslackmorediagnosticinforma-tionaboutthedistributionoftheopinion,asisthecasewhentheymakejudgmentsaboutunfamiliargroups.ThePresentResearchInthefirstsectionofthisarticle(Studies1Aand1B),wetestthebasichypothesisthatperceiverswillestimatemorewidespreadsupportforanopinionafterhearingonegroupmemberadvanceitrepeatedlythanafterhearingthesamepersonexpressthesameopiniononlyonce.Inthesecondsection(Studies2and3),weseektounderstandthepsychologicalmechanismdrivingthisproposedeffectbydistinguishingbetweentwotypesofexplanations,onebasedonamodelofmetacognitiveprocessesdescribedaboveandtheotherbasedonconsciousinferencesperceiversmaymakeaboutgroupdynamicsandthemeaningofopinionrepetitioninagroup.Thefinalsection(Studies4and5)examinesboundaryconditionsoftheproposedeffectbyfocusingonthemoderatingroleofknownversusunknowngroupnorms.TheBasicEffect:Studies1Aand1BStudy1A:PreservingOpenSpaceinNewJerseyInStudy1A,participantsreadopinionstatementstakenfromasupposedfocusgroupdiscussion.Wepredictedthatobserverswouldestimatemorewidespreadsupportfortheissueafterreadingonegroupmemberreiterateanopinionthreetimesthantheywouldafterreadingthesamepersonexpressthesamepointonlyonce.Atotalof177UniversityofMichiganunder-graduatesparticipated.Sixty-sevenreceived$8forcompletingthestudywithinalargerquestionnairepacket.Theremaining110participantswererecruitedinpublicplacesoncampus.Opinionfamiliarityandopinionsourceweremanipulatedinathree-conditionbetween-subjectsdesign.Participantsinthethreepersoncontrolconditionreadthreeopinionstatementseachmadebyadifferentgroupmember(Jim,Mark,andJohn),thoseinthere-peatedopinionconditionreadthesamethreestatementsbuttheywereallattributedtoonegroupmember(Jim),andthoseinthesingleopinioncontrolconditionreadoneopinionstatementfromonegroupmember(Jim).ParticipantsweretoldthattheresearcherswereinterestedinanissuefromthelastNewJerseystateelection,namelywhetherlandshouldbesetasidetopreserveopenspaceinthestate(afictitiousissuecreatedforthepurposeofthisexperi-ment).ArepresentativefocusgroupoffiveNewJerseyhomeown-ershadmettodiscusstheiropinionsontheissue.Participantsinthethreepersoncontrolandrepeatedopinionconditionsread,“Threeofthecommentsmadeduringthefocusgrouphavebeenrandomlyselectedandareprintedbelow,alongwiththehome-owner’sname.Becausethiswasarandomprocess,selectedstate-mentsmaybefromonepersonor,atmost,threedifferentpeople.”Thesingleopinioncontrolconditionwaswordedsimilarlybutonlyreferencedonecomment.Participantsthensaweitherthefollowingthreeopinionstate-mentsoronerandomlyselectedopinionstatement,allintendedtoconveytheopinionthatopenspaceshouldbepreservedforrec-reationalpurposes:“Iaminfavorofopenspacebecauseitisimportantthatpeoplehaveopportunitiesforoutdoorrecreation,”“IthinkthatopenspaceisagoodideabecausemoreopenspaceinNewJerseymeansthatwewillallhavethechancetodooutdooractivitiesduringthesummer,”and/or“Openspacepoliciesshouldbesupportedbecausetheyguaranteethatpeoplecanspendtheirleisuretimeinanaturalenvironment.”Inthethreepersoncontrolcondition,eachstatementwasattributedtoadifferentfocusgroupmember(Jim,Mark,andJohn);intherepeatedandsingleopinionconditions,onegroupmember(Jim)madeall(orone)oftheAfterthesemanipulations,participantsestimatedtheopinionsofthefocusgroupandNewJerseyhomeownersingeneraltowardpreservingopenspace(forboth,1stronglyagainstand7stronglyinfavor).TheyalsoestimatedthepercentageofNewJerseyhomeownerswhosupportedpreservingopenspaceinthestate(___%).Finally,participantscompletedamanipulationcheckandrecalledfromhowmanyfocusgroupmemberstheyhadreadResultsandDiscussionAcrossthestudies,ourprimaryinterestwasincomparingthegroup-levelestimatesmadebyparticipantswhoareexposedtorepeatedopinionsfromonegroupmembertotheestimatesmadebyparticipantswhoreadthesamespeakeradvancethesameopiniononlyonce.Accordingly,ourdataanalyticstrategywastotestthiscentralhypothesisusingtwoorthogonalcontrasts.Inthestudiesthatincludedbothathreepersoncontrolconditionandasingleopinioncontrolcondition,wefirstcomparedthetwocontrolconditionswitheachother.Second,wetestedthemainhypothesisbycomparingtherepeatedconditionwiththesingleopinioncon-trol.Attheendofthearticle,wepresenttheresultsofameta-analytictestascertainingwhether,acrosstherelevantstudies,thethreepersoncontroldifferedsignificantlyfromtherepeatedFocusgroupopinion.Asexpected,participantsreadingthreeopinionsinfavorofopenspacepreservationeachfromadifferenthomeownerestimatedmorefocusgroupsupportforopenspacepreservation(0.89)thandidparticipantsreadingonlyoneopinionfromonehomeowner((1,174)1.60.Ofmoreimportance,resultsalsosupportedthemainhypothesis.Participantswhoreadonehomeownerreiteratesupportforopenspacethreetimesalsoestimatedmorefocusgroupsupportfortheissue(1.02)thandidparticipantswhoreadthesamehomeownerexpressthesamepointonlyonce,(1,174)OpinionofNewJerseyhomeownersingeneral.Wealsoex-aminedtheinferencesobserversdrewabouttheopinionsofNewJerseyhomeownersingeneral,thegroupfromwhichthefocusgroupwasdrawn.Thismeasureallowedustoseewhethertherewasgeneralizationandalsotoruleoutapossiblealternativeexplanation,namelythatobserversthoughtthatJim’srepeatedcommentspersuadedtheothergroupmembers.Thetwoitemsmeasuringgeneralhomeownersupport(favorabilityandpercent-ageestimate)wereconceptuallysimilarbutmeasuredondifferentINFERRINGTHEPOPULARITYOFANOPINION patternofresultsandsignificanceofallanalysesremainedun-changedwhenincludingonlythoserespondingcorrectlytothememoryquestionnaire(50outof60participants).Discussion:Studies1Aand1BResultsfromStudies1Aand1Bshowthatanindividualinagroupwhosimplyreiteratessupportforaparticularstancecanswayobservers’estimatesofwherethegroupstandsmorethanifheorsheexpressesthesameopiniononlyonce.Resultsalsoshowedthattheseinferencesgeneralizetothelargergroupfromwhichtheoriginalsamplewasdrawnandcanhaveseeminglyimportantpolicyconsequences.TheMechanism:Studies2and3Thenexttwostudiessoughttoestablishthepsychologicalmechanismdrivingthiseffectbydistinguishingbetweentwotypesofexplanations,onebasedonmetacognitiveprocessesofopinionfamiliarityandtheotherinvolvingconsciousinferencesaboutgroupdynamicsandthemeaningofopinionrepetitioninagroup.Earlier,weproposedamechanismbasedonopinionfluencyorfamiliarity.Ourlineofreasoningwasasfollows.Hearingonegroupmemberrepeatanopinionincreasestheactivationofthatopinioninmemory.Whenperceiverssubsequentlyreflectonthegroup’sopinion—eitherspontaneouslyorwhenmakingagroup-leveljudgmentonasurvey—thatmemorytraceiscuedandexperiencedsubjectivelyasafeelingoffluencyorfamiliarityfortheopinion.Thissubjectiveexperience,inturn,spillsoverintoobservers’group-leveljudgments.AlthoughtheresultsofStudies1Aand1Bareconsistentwiththefamiliaritymodel,otherexpla-nationsmaystillapplyandneedtoberuledout.Forinstance,ratherthanbasingtheirjudgmentsonexperientialinformation,observersinsteadmayhavemadeaconsciousinferenceaboutthemeaningofopinionrepetitioninagroup.ObserversmayhavereasonedthataspeakersuchasJimwouldbelessinclinedtoreiterateanopinioninagroupsettingunlessheknewthattheopinionwasbroadlyshared.Thus,repeatedstatementsmadebyonepersonmayhaveledtomoreextensivegroup-leveljudgmentsnotthroughopinionfamiliaritybutthroughadifferentpsycholog-icalmechanism,onesimilartothatemphasizedbyworkonthespiralofsilence(Noelle-Neumann,1984).Incontrast,itisworthnotingthatperceiversmayequallyplausiblyassumethatJimrepeatshisopinionsooftenbecauseherealizesthathisattemptstopersuadethegrouphavenotyetbeensuccessful.Hence,itisdifficulttopredictwhichinferencesperceiversmaydrawfromobservedrepetitions.Toavoidtheseambiguities,weremovedthegroupdiscussionaspectfromthecoverstoryofStudy2.Usingan“instantreplay”paradigm,participantseitherreadthreeidenticalcopiesofthesameopinionstatementthathadbeenmadeoncebyonegroupmemberorreadthesameopinionstatementonlyonce.Ifthesubjectiveexperienceofopinionfluencyorfamiliarityissufficienttoproducetheeffect,thensimplyreadingtheidenticalopinionstatementoverandovershouldleadperceiverstoestimatewidergroupsupportforthepositionthanpeoplewhohavereadthesameopiniononlyonce.Incontrast,ifthepreviousresultsreflectanassumptionbyperceiversthatspeakersonlyrepeattheiropinionswhentheyknowagroupagrees,thenparticipants’extensityjudg-mentsshouldbeunaffectedbythenumberofcopiestheyhavereadofthesameopinionstatement.Study2:InstantReplayUndergraduatesattheUniversityofMichigan—DearbornandUniversityofToledo(305)werecompensated$8forfillingoutaquestionnairepacketincludingboththisstudyandseveralunrelatedones.Study2usedatwo-conditionbetween-subjectsdesign(duplicatestatementsandsingleopinioncontrol).Participantsweretoldthatthe“NationalAlliancePartyofBelgium”(afictitiousgroupcreatedforthisexperiment)hadrecentlydiscussedtheparty’spositiononreproductiverights.Partyofficialshadrandomlye-mailedasampleofregisteredpartymemberstosolicittheiropinionsandoneofthecommentsre-ceivedwasprinted.Participantsthensawacommentfavoringamoremoderatestancethatwassubmittedviaane-mailprogrambyapartymembernamedJacquesAndersi.Intheduplicatestate-mentscondition,theidenticalopinionstatementwascopiedtwomoretimesonthebottomofthepage.Thematerialsweredesignedsothecopieslookedlikeanirrelevantpropertyofthee-mailsoftware.However,participantsinthisconditionhadtoactuallyreadtheextracopiesbeforerealizingthattheywereidenticalcopiesofthesameopinion.Incontrast,thoseinthesinglestate-mentcontrolconditionreadthesameopinionstatementonlyonce.Afterward,participantsestimatedthepercentageofpartymemberssupportingtheadoptionofamoremoderatestance,reportedhowmanysourceshadmadeopinions,andnotedwhetherthestate-mentstheyreadwereidentical.ResultsandDiscussionConsistentwithourmainhypothesis,participantsestimatedmorewidespreadsupportforamoderatestanceafterreadingoneopinionstatementadvocatingthatpositionthreetimes(21.24%)thandidparticipantswhoreadthesameopinionstatementonlyonce(0.28.Memorydistortionormisreadingthesurveydidnotexplainthefindings.Thepatternandstatisticalsignificanceoftheanalysesremainedthesamewhenincludingonlythoseparticipantsrespondingcorrectlytobothquestionsincludedinthememoryquestionnaire,thatis,individualswhobothcorrectlyreportedonespeakerhadexpressedanopinionandcorrectlyreportedthattheopinionstatementswereidenticalcopies(271outof305participants).Theseresults,showingthatsimplyreadingthesameopinionoverandoveragain,evenwhenpeopleareawarethatitisanidenticalcopy,suggestthatopinionrepeti-tionaloneissufficienttoproducemoreextensivegroup-leveljudgments.Thisisconsistentwithanexplanationinvolvingopin-ionfamiliarityandinconsistentwiththeideathataninferenceaboutthemeaningofrepetitioninagroupisnecessaryfortheStudy3:MeasuringOpinionActivationDirectlyInStudy3,wesoughttoobtainadditionalconvergentevidenceforthepsychologicalmechanismbymeasuringopinionactivationdirectlyusingalexicaldecisionresponsetime(RT)task.PastworkINFERRINGTHEPOPULARITYOFANOPINION (1,194)0.50;andpercentage(1,188)Mediationalanalysis.Ourmainhypothesiswasthatpartici-pants’experienceofopinionfamiliarity,operationalizedhereastheirresponselatencytoidentifyingopenspacerelevantwords,wouldmediatetheirgroup-leveljudgments.Wetestedthismedi-ationalpredictionusingthemethodrecommendedbyMacKin-non,Lockwood,Hoffman,West,andSheets(2002;seealsoHerbst,Gaertner,&Insko,2003).First,wedummycodedthenumberofopinionsindependentvariable(1singleopinioncontrol,2repeatedopinions).Becauseitisamoresensitivemeasureandlesslikelytosufferfromrestrictedrange,weusedparticipants’estimatesofthepercentageofhomeownerssupport-ingopenspaceasthedependentvariableandtheirRTdifferencescore(openspacewords–neutralwords)asthemediator.Asrequiredbythemethod,thenumberofopinionspartici-pantsreadsignificantlypredictedboththedependentvariable(percentageestimates;.05)andthemediator(re-sponselatencies;.05).Thatis,participantsreadingrepeatedopinionsbothestimatedmorewidespreadsupportforopenspaceandrespondedmorequicklytoopen-space-relevantwordsrelativetoneutralwordsthandidthosereadingthesamesourceexpressthesameopiniononlyonce.Themediator(re-sponselatency)alsosignificantlypredictedparticipants’percent-ageestimateswhencontrollingfortheeffectofnumberofopin-ionsread(.05).Thatis,themorequicklyparticipantsidentifiedtheopenspacewords,thegreatersupportforopenspacetheyestimatedwhencontrollingforexperimentalcondition.Aspredicted,whenparticipants’responselatencieswerecontrolled,therelationshipbetweenthenumberofopinionstheyreadandtheirpercentageestimateswassignificantlyreduced.05).AmodifiedSobeltestoftheratioofthestandarderrorof(seeMacKinnonetal.,2002)confirmedthatthisreductionwasstatisticallysignificant(Theseresultssuggestthatopinionactivationorfamiliarityisasignificantdriveroftherepetitioneffect.Opinionimportance.Apossibilityraisedearlierwasthatpar-ticipantsmayhaveinferredopinionimportancefromrepeatedopinions.Perceiversdidjudgetheopiniontobemoreimportanttothegroupafterreadingrepeatedopinions(thantheydidafterreadingoneopinion((1,194)0.36.However,perceivedimpor-tancedidnotfullyaccountfortherepetitioneffect.Amultivariateanalysisofcovarianceusingimportanceasacovariateandthethreeopinionprevalenceestimatesasdependentvariablesshowedthattherepetitioneffectremainedsignificantevenwhencovaryingoutimportance,(3,187).01.Furtheranalysessuggestedthattheopinionextensityjudgmentsmayhavedrivenparticipants’importancejudgmentsratherthanviceversa.Whentheopinionextensitymeasureswereincludedascovariates,theeffectofnumberofopinionsonimportancedroppedtoanonsig-nificantlevel:focusgroupestimateascovariate,(1,193).31;homeownersingeneralascovariate,(1,193).21;andpercentageestimateascovariate,(1,187).21.Thepatternandstatisticalsignificanceofallofthegroup-levelestimatesremainedthesamewhenincludingonlythoseparticipantsrespondingcorrectlytothemanipulationcheck(161outof196participants),withtheoneexceptionthatthejudgmentsoftheimportanceoftheopiniontothegroupdroppedtononsignificance(Discussion:Studies2and3Resultsthusfarshowthatanindividualinagroupwhosimplyrepeatssupportforaparticularstancecanswayobservers’esti-matesofwherethegroupstandsontheissuemorethanifthesamespeakerhadexpressedthesameopiniononlyonce.Usingtwodifferentparadigms,Studies2and3showconvergentevidencethatopinionfamiliarityisasignificantdriveroftheeffect.Study2showedthatsimplyreadingthesameopinionoverandoverissufficienttoproducetheeffect.Study3measuredopinionactiva-tiondirectlyandshowedthatopinionrepetitionisassociatedwithincreasedactivationofopinion-relatedconcepts.Thisactivation,inturn,isitselfsignificantlyrelatedtoperceivers’group-leveljudgments.Althoughdemonstratingtheroleofopinionfamiliarity,resultsfromStudies2and3alsosuggestthatcounterexplanationsinvolvingconsciousinferencesaboutthemeaningofopinionrep-etitioninagroupandopinionimportancearenotnecessarytoexplainthefindings.Moderators:Studies4and5AlthoughStudies1A–3showthatperceiversusefeelingsoffamiliaritythatwereactuallycreatedbyonepersontomakeinferencesabouttheprevalenceofsupportforanopinioninagroup,indirectevidencefromthepersuasionliteraturesuggeststhatinsomecasesperceiverscananddocorrecttheirsenseofopinionfamiliarityforsourceinformation.Thatis,thedesignofourfirstthreestudiesbearsresemblancetothatusedbyHarkinsandPetty(1981),whichactuallyshoweddifferentresults.HarkinsandPettywereinterestednotinwhetheronesource’srepeatedopinionsinfluencedobservers’group-leveljudgmentsbutratherwhethertheyinfluencedpersonalpersuasion.Theirparticipants Sixrespondentsfailedtoreportapercentageestimate,leaving190participantsinthisanalysis.MacKinnonetal.(2002)empiricallyassessedthestatisticalpropertiesofthe14mostcommonlyusedstatisticalmethodsforestablishingmedi-ation,includingBaronandKenny’s(1986)method.Theiranalysiscon-cludedthatthemethodwassuperiortotheother13methodsonthedimensionsofcontrollingTypeIerrorandmaximizingpower.Becausetheresponsesofindividualsrespondingincorrectlytothemanipulationcheckareambiguous(i.e.,itisunknownwhethertheirfailuretoreportcorrectlywasduetomisreadingtheformorexperiencingmemorydistortion),themediationalanalysiswasconductedwithonlythoseindi-vidualsrespondingcorrectlytothemanipulationcheckaskingforthenumberofopinionsources(161outof196participants).Table2Study3:MeanResponseLatenciesinMillisecondsasaFunctionofWordType(Issue-RelevantWordsvs.Neutral ConditionRelevantwordsNeutralwordsDifferenceOneopinion530.49552.73Repeatedopinions501.78565.34INFERRINGTHEPOPULARITYOFANOPINION within-subjectsfactorrevealedsignificantmaineffectsforbothnumberofopinions,(1,105).01,andprevious(1,105).01.However,thesemaineffectswerequalifiedbythepredictedNumberofOpinionsPreviousKnowledgeinteraction,(1,105)Twofollow-upcomparisonswereconductedtodeterminethenatureoftheinteraction.Aspredicted,whenparticipantsdidnothavepreviousknowledgeaboutwherethegroupstoodontheissue(e.g.,NationalAllianceParty),theirgroup-leveljudgmentswereaffectedbythenumberofopinionstheyread.Observersreadingonepartymemberrepeatedlyadvocateamoderatepositionesti-matedmoregroup-levelsupportforthatviewpoint(favorability:1.25;percentage:thandidthosereadingthesamepersonexpressthesamepositiononlyonce(favorability:1.07;percentage:(1,51)1.08.Incontrastandalsoaspredicted,whenparticipantshadpriorknowl-edgeoftheopinion(RepublicanParty),theywereabletocorrecttheirjudgmentsandtheirgroup-levelestimateswerenotaffectedbythenumberofopinionexpressionstheyread.Observersesti-matedasimilaramountofsupportforamoderatepositionregard-lessofwhethertheyreadrepeatedopinionsfromonesource1.14;percentage:14.72%)oronlyoneopinionfromthesameperson(favor-1.32;percentage:(1,54)–0.04(seeTable3forthepatternofmeans).Itisimportanttonotethattherewerenomaineffectsorinteractionsoftherepeated-measuresfactor(all.10),confirmingthatthefavorabilityandpercentagemeasuresactedsimilarly,andthepatternofdataandsignificanceoftheanalysesreportedremainedthesameevenwhenretainingonlythoseparticipantswhorespondedcorrectlytothememoryques-tionnaire(87outof110participants).Study5:KnownOpinionsandTimeDelayStudy4showedthatobserverswhowereintheknowaboutagroup’sopinioncorrectedtheirsenseofopinionfamiliaritywithsourceinformation.Thismaysuggestthattherepetitioneffectonlyholdswhenpeopleareformingestimatesofgroupopinionbutnotwhentheyarechanginggroup-levelimpressions.However,someworkonfamiliarityandjudgmentsuggeststhatpeople’sabilitytodiscounttheirfeelingsoffamiliaritymaybeshortlived.Ifpeople’sgroup-leveljudgmentsaredrivenbytheirfeelingsoffamiliarity,thenundersomeconditionstherepetitioneffectshouldemergeinknownnormsituationsaswell.Althoughresearchshowsthatpeopleareinitiallyreluctanttousefamiliarityintheirjudgmentswhenitsdiagnosticityiscalledintoquestion,itcanspilloverintoobservers’judgmentswhenthecauseofthefamiliaritybecomeslessobvious.Jacobyetal.(1989),forinstance,hadparticipantsreadalistofnonfamousnames(e.g.,SebastianWeisdorf).Participantsthenmadefamejudgmentsofnamesonasecondlist,someofwhichwerefromthefirstlistandsomeofwhichwerenew.Whenparticipantssawthesecondlistimmediatelyafterthefirst,familiarityinformationdidnotinflu-encetheirfamejudgments.Thisisbecauseparticipantsknewthenamesfeltfamiliaronlybecausetheyhadjustreadthemfortheexperiment.However,aftera24-hrtimedelayparticipantsshoweda“falsefameeffect,”attributingfameatTime2tonamestheyhadstudiedatTime1.Thetimedelaypreservedparticipants’feelingoffamiliarityforthenamesbutdiminishedtheirsourcememoryforwhytheyseemedfamiliar.Thispredictionisakintoasleepereffect(see,e.g.,Hovland&Weiss,1951;Pratkanis,Greenwald,Leippe,&Baumgardner,1988).Wehypothesizethatasimilarprocessmayoccurhere.Initially,perceiverswithprioropinionknowledgemaybeabletosubtracttheeffectofrepetitionfromtheirjudgments.However,thisabilitymaybeshortlived.Afteratimedelay,observerswillbeleftwithafeelingoffamiliarityfortheopinionbutimpairedconsciousmemoryforitssource.Asaconsequence,followingatimedelayevenperceiverswithpreviousopinionknowledgemayshowtherepetitioneffect.Atotalof196PrincetonUniversitystudentsparticipated.Onegroup(169)wascompensated$8forcom-pletingthisstudyalongwithseveralunrelatedonesaspartofa“QuestionnaireDay”packetsponsoredbythePsychologyDepart-ment.Theremaining27studentscompletedthestudyinexchangeforcreditintheirpsychologylaboratorycourse.Thestudyuseda3(numberofopinions:threepersoncontrol,repeatedopinion,andsingleopinioncontrol)2(timedelay:delayornodelay)between-subjectsfactorialdesign.Theuniversitywherethestudywasconducteddoesnothaveprofessionalschools,suchasalaworbusinessschool.Thenormoncampusisagainstestablishingprofessionalschoolsandinfavorofkeepingtheliberalartstradition.Studentsweretoldthattherehadbeenrecentdiscussionoftheprofessionalschoolissueoncampus,thatundergraduateopinionshadbeensolicitedandpostedonabulletinboarddiscussiongroup,andthatsomeofthecommentssubmittedhadbeenprinted.Participantstheneithersawthefollowingthreeopinionstate-mentsoronerandomlyselectedopinionstatement,allfavoringadoptingprofessionalschoolstocompetewith“peerinstitutions:”“Princetonneedstoestablishprofessionalschoolstocompete Thestudentscompletingthestudyinexchangeforcoursecreditcompletedthestudyunderdelayconditionsandwereeitherassignedtothesinglestatementcontrolorrepeatedcondition.Analysesrevealednosig-nificantmaineffectsorinteractionsofstudytype(paidvs.coursecredit)inanyoftheanalyseswithanyofthedependentvariables,sothisvariableisnotdiscussedfurther.Table3Study4:TheEffectsofPriorOpinionKnowledgeandNumberofOpinionsReadonEstimatesofPartyFavorabilityandPercentageofPartyMembersSupportingaMoreModerate KnownopinionUnknownopinionFavorability%Favorability%Singleopinioncontrol3.3742.973.9642.66Repeatedopinions3.5039.264.6862.80Highernumbersonthefavorabilitymeasuresindicatemoreagree-mentwiththeopinionstatements(greaterfavorabilitytowardtheissue).INFERRINGTHEPOPULARITYOFANOPINION familiaritybecomeslessobvious,people’sfeelingsdospilloverintotheirgroup-leveljudgments.GeneralDiscussionWhatwethinkothersthinkgreatlyinfluencesourownpersonalthoughts,feelings,andbehavior.Ouraccuracyinformingimpres-sionsofgroupopinionandgroupnormsisalsoanessentialcomponentinguidingoursocialinteractions.Nevertheless,littleisknownabouthowweestimatetheprevalenceofanopinioninagroup.Quiteobviously,anopinionislikelytobemorewidelysharedthemoredifferentgroupmembersexpressit.Ourpartici-pantsclearlyrecognizedthisandprovidedhigherprevalenceesti-mateswhenthesameopinionwasexpressedoncebyeachofthreedifferentgroupmembersthanwhenitwasexpressedoncebyonegroupmember.Moresurprising,andconsistentwithourhypoth-eses,however,ourstudiesshowedthathearingonepersonexpressanopinionrepeatedlyalsoleadsperceiverstoestimatethattheopinionismorewidespreadrelativetohearingthesamecommu-nicatorexpressthesameopiniononlyonce.Acrossourstudies,wefoundthatalthoughthreepeopleeachexpressingthesamesenti-mentismoreinfluentialthanonepersonexpressingthesamebeliefthreetimes,thelatterwas,onaverage,90%oftheformer.TheUnderlyingProcessWealsosoughttounderstandtheprocessesdrivingtherepeti-tioneffectbydistinguishingbetweentwotypesofmechanisms,onebasedonmetacognitiveexperiencesandtheotherbasedonconsciousinferencesthatperceiversmaymakeaboutgroupdy-namicsandthemeaningofrepetitioninagroup.Drawingonpastresearchofmetacognitiveexperiencesinhumanjudgment(forareview,seeSchwarz,2004),weproposedthatrepeatedexposuretoanopinionincreasestheaccessibilityoftheopinioninmemoryandresultsinafeelingoffamiliaritywhentheopinionisencoun-teredagain,suchasthroughspontaneousreflectionoraspartofaquestionaboutitsprevalence.Thissubjectiveexperienceleavesperceiverswiththe(correct)impressionthattheyhaveheardthisopinionmanytimes.Inourdailylives,wearelikelytohearanopinionmanytimeswhenitisofferedbydifferentpeopleindifferentsituations.Wepredictedthatpeopleoverapplythisusu-allycorrectassumptionandinferextensityfromfamiliarityevenunderconditionsinwhichtheopinionisrepeatedlyofferedbythesamesingleindividual.OurresultsconsistentlysupportthisAstheoreticallyexpected,theobservedeffectcanbetracedtotheincreasedaccessibilityoftheopinionandrelatedconcepts.Opinionrepetitionfacilitatesfastresponsestoissue-relevantwordsonalexicaldecisiontask(Study3).Thisincreasedopinionacces-sibility,inturn,mediatestheinfluenceofopinionrepetitiononperceivers’extensityjudgments.Incontrast,weobtainednosup- ItisinterestingtonotethattheworkofLinvilleandcolleagues(Linville&Fischer,1993;Linville,Fischer,&Salovey,1989)showsthatincreasingfamiliaritywiththeindividualmembersofagroupcanincreaseperceivers’senseoftheheterogeneityofopinionsinthegroup,whereasourworksuggeststhatincreasedfamiliaritywithoneparticulargroupmem-ber’spositionleadstoadecreaseinperceivers’estimatesofopinionAmeta-analysisoftheresultsforthethreestudiesusingthethreepersoncontrolcondition(Studies1A,1B,and5)showedthat,aswouldbelogicallyexpected,hearingthreedifferentpeopleeachadvanceanopinionleadsobserverstoattributegreatergroup-levelsupportfortheissuethandoeshearingonegroupmemberrepeatthesameopinionthreetimes, Agreement with opinion statements One opinioncontrolRepeatedOpinionsThree personcontrolNo time delayTime delay Figure1.Study5:Theeffectsoftimedelayandnumberofopinionsonestimatesofgeneralstudentsupportforprofessionalschools.Highernumbersindicatemoreagreementwiththeopinionstatements(greaterfavorabilitytowardprofessionalschools).INFERRINGTHEPOPULARITYOFANOPINION Baron,R.M.,&Kenny,D.A.(1986).Themoderator–mediatorvariabledistinctioninsocialpsychologicalresearch:Conceptual,strategic,andstatisticalconsiderations.JournalofPersonalityandSocialPsychology,Begg,I.M.,Anas,A.,&Farinacci,S.(1992).Dissociationofprocessesinbelief:Sourcerecollection,statementfamiliarity,andtheillusionofJournalofExperimentalPsychology:General,121,446–458.Cialdini,R.B.,Kallgren,C.A.,&Reno,R.R.(1991).Thefocustheoryofnormativeconduct:Atheoreticalrefinementandreevaluationoftheroleofnormsinhumanbehavior.InM.P.Zanna(Ed.),Advancesinexper-imentalsocialpsychology(Vol.24,pp.201–234).SanDiego,CA:Deutsch,M.,&Gerard,H.B.(1955).Astudyofnormativeandinforma-tionalsocialinfluencesuponindividualjudgment.JournalofAbnormalandSocialPsychology,51,629–636.Festinger,L.(1954).Atheoryofsocialcomparisonprocesses.Relations,7,Feustel,T.C.,Shiffrin,R.M.,&Salasoo,A.(1983).Episodicandlexicalcontributionstotherepetitioneffectinwordidentification.JournalofExperimentalPsychology:General,112,309–346.Fishbein,M.,&Ajzen,I.(1972).Attitudesandopinions.AnnualReviewofPsychology,23,Haber,R.N.,&Hershenson,M.(1965).Theeffectsofrepeatedbriefexposuresongrowthofapercept.JournalofExperimentalPsychology,40–46.Hamill,R.,Wilson,T.D.,&Nisbett,R.E.(1980).Insensitivitytosamplebias:Generalizingfromatypicalcases.JournalofPersonalityandSocialPsychology,39,578–589.Harkins,S.G.,&Petty,R.E.(1981).Effectsofsourcemagnificationofcognitiveeffortonattitudes:Aninformation-processingview.JournalofPersonalityandSocialPsychology,40,401–413.Hasher,L.,Goldstein,D.,&Toppino,T.(1977).Frequencyandtheconferenceofvalidity.JournalofVerbalLearningandVerbalBehavior,Hastie,R.,&Park,B.(1986).Therelationshipbetweenmemoryandjudgmentdependsonwhetherthejudgmenttaskismemory-basedorPsychologicalReview,93,258–268.Herbst,K.C.,Gaertner,L.,&Insko,C.A.(2003).Myheadsaysyesbutmyheartsaysno:Cognitiveandaffectiveattractionasafunctionofsimilaritytotheidealself.JournalofPersonalityandSocialPsychol-ogy,84,1206–1219.Hovland,C.I.,&Weiss,W.(1951).Theinfluenceofsourcecredibilityoncommunicationeffectiveness.PublicOpinionQuarterly,15,635–650.Jacoby,L.L.,Kelley,C.,Brown,J.,&Jasechko,J.(1989).Becomingfamousovernight:Limitsontheabilitytoavoidconsciousinfluencesofthepast.JournalofPersonalityandSocialPsychology,56,326–338.Jacoby,L.L.,&Whitehouse,K.(1989).Anillusionofmemory:Falserecognitioninfluencedbyunconsciousperception.JournalofExperi-mentalPsychology:General,118,126–135.Johnson,W.A.,Dark,V.J.,&Jacoby,L.L.(1985).Perceptualfluencyandrecognitionjudgments.JournalofExperimentalPsychology:Learn-ing,Memory,andCognition,11,Kelley,C.M.,&Rhodes,M.G.(2002).Makingsenseandnonsenseofexperience:Attributionsinmemoryandjudgment.PsychologyofLearn-ingandMotivation,41,Korte,C.(1972).Pluralisticignoranceaboutstudentradicalism.etry,35,576–587.Linville,P.W.,&Fischer,G.W.(1993).Exemplarmemoryandpropertyabstractionprocessesinperceiveddistributionsofin-groupandout-groupmembers.SocialCognition,11,Linville,P.W.,Fischer,G.W.,&Salovey,P.(1989).Perceiveddistribu-tionsofthecharacteristicsofingroupandoutgroupmembers.JournalofPersonalityandSocialPsychology,57,MacKinnon,D.P.,Lockwood,C.M.,Hoffman,J.M.,West,S.G.,&Sheets,V.(2002).Acomparisonofmethodstotestmediationandotherinterveningvariableeffects.PsychologicalMethods,7,McFarland,C.,&Miller,D.T.(1990).Judgmentsofself-othersimilarity:Justlikeotherpeopleonlymoreso.PersonalityandSocialPsychologyBulletin,16,475–484.McGlone,M.S.,&Tofighbakhsh,J.(2000).Birdsofafeatherflockconjointly?Rhymeasreasoninaphorisms.PsychologicalScience,11,424–428.Miller,D.T.,&Prentice,D.A.(1996).Theconstructionofsocialnormsandstandards.InE.T.Higgins&A.W.Kruglanski(Eds.),psychology:Handbookofbasicprinciples(pp.799–829).NewYork:Moscovici,S.(1985).Socialinfluenceandconformity.InG.Lindzey&E.Aronson(Eds.),Handbookofsocialpsychology(Vol.2,pp.347–412).NewYork:RandomHouse.Newcomb,T.M.(1943).Personalityandsocialchange.NewYork:Holt,Rhinehart&Winston.Noelle-Neumann,E.(1984).Thespiralofsilence:Oursocialskin.cago:UniversityofChicagoPress.Pratkanis,A.R.,Greenwald,A.G.,Leippe,M.R.,&Baumgardner,M.H.(1988).InsearchofreliablepersuasioneffectsIII:Thesleepereffectisdead.Longlivethesleepereffect.JournalofPersonalityandSocialPsychology,54,Prentice,D.A.,&Miller,D.T.(1993).Pluralisticignoranceandalcoholuseoncampus:Someconsequencesofmisperceivingthesocialnorm.JournalofPersonalityandSocialPsychology,64,PsychologySoftwareTools.(2002).E-Prime(Version1.2)[Computersoftware].Pittsburgh,PA:Author.Quattrone,G.A.,&Jones,E.E.(1980).Theperceptionofvariabilitywithinin-groupsandout-groups:ImplicationsforthelawofsmallJournalofPersonalityandSocialPsychology,38,Reber,R.,&Schwarz,N.(1999).Effectsofperceptualfluencyonjudg-mentsoftruth.ConsciousnessandCognition:AnInternationalJournal,338–342.Rothbart,M.,Fulero,S.,Jensen,C.,Howard,J.,&Birrell,P.(1978).Fromindividualtogroupimpressions:AvailabilityheuristicsinstereotypingJournalofExperimentalSocialPsychology,14,Schwarz,N.(2004).Meta-cognitiveexperiencesinconsumerjudgmentanddecisionmaking.JournalofConsumerPsychology,14,Shelton,J.N.,&Richeson,J.A.(2005).IntergroupcontactandpluralisticJournalofPersonalityandSocialPsychology,31,1189–Sherif,M.(1936).Thepsychologyofsocialnorms.NewYork:McGraw-Skurnik,I.,Yoon,C.,Park,D.C.,&Schwarz,N.(2005).Howwarningsaboutfalseclaimsbecomerecommendations.JournalofConsumerResearch,31,Terry,D.J.,&Hogg,M.A.(1996).Groupnormsandtheattitude–behaviorrelationship:Aroleforgroupidentification.PersonalityandSocialPsychologyBulletin,22,776–793.Tversky,A.,&Kahneman,D.(1973).Availability:Aheuristicforjudgingfrequencyandprobability.CognitivePsychology,5,Weaver,K.,&Schwarz,N.(2005).Usingsocialinfluencetochangehealth.Unpublishedmanuscript,UniversityofMichigan.Whittlesea,B.W.A.,Jacoby,L.L.,&Girard,K.(1990).Illusionsofimmediatememory:Evidenceofanattributionalbasisforfeelingsoffamiliarityandperceptualquality.JournalofMemoryandLanguage,716–732.Zajonc,R.B.(1968).Attitudinaleffectsofmereexposure.JournalofPersonalityandSocialPsychologyMonographs,9,ReceivedJanuary17,2003RevisionreceivedJuly29,2006AcceptedAugust18,2006INFERRINGTHEPOPULARITYOFANOPINION