920
FIGURE2
WEIGHTOFFOODTAKENBYCONDITION(STUDY1)
patternshowedthatparticipantsperceivedtheconfederatetobeheavierwhenshewaswearingthesuit(Mp0.46)than whenshewas not (Mp1.93).
DependentMeasures.Thequantitiesoffoodtakenandeaten
werestandardizedwithinfood(granolaorM&Ms)priortoanalysis.Tofacilitateinterpretation,however,un-standardizedmeansarereportedbelow.
AnANCOVAwithquantityoffoodtakenfromthebowlasthedependentmeasurerevealedonlyamaineffectforconfederatebodytype(F(1,60)p3.96,pp.05).Partici-pantstookmorefood(measuredingrams)whentheconfed-eratewasthin(Mgranolap41.33,MM&Msp74.27)thanwhenshewasobese(Mgranolap33.47,MM&Msp58.20).Thein-teractionbetweenfoodtypeandconfederatebodysizewasnonsignificant(F!1),indicatingthatregardlessofwhetherthefoodwasperceivedtobehealthyorunhealthy,participantsshowedrestraintafterobservinganobesepersontakingalotoffoodascomparedtoathinpersontakingthesameamount.Comparedtothecontrolgroup,participantstookmoreonaveragefromthefoodbowliftherewasaconfederatepresent(Mgranolap38.06,MM&Msp66.23)thanwhenthere
wasnot(Mgranolap22.33, MM&Ms p22.71; F(1,85)p 19.53,p!.001).Controlgroupparticipantstooksignifi- cantly less than those alongside an obese confederate (F(1,85)p9.21,p!.01) anda much smallerquantity thaniftheconfederatewasthin(F(1,72)p20.09,p! .001).Fullresultscanbeseeninfigure2.
AnANCOVAonparticipants‘actualconsumptionalsore-vealedthesamepatternastheirchoicebehavior(F(1,60)p5.67,pp.02).ParticipantsatealmosttwiceasmuchofboththegranolaandtheM&MswhentheconfederatewasthinMgranolap21.47,MM&Msp33.00)thanwhenshewasobese(Mgranolap13.13,MM&Msp20.47),regardlessofwhetherthefoodwasperceivedtobehealthyorunhealthy(F!1).
Comparedtothecontrolgroup,participantsatemoreiftherewasaconfederate(Mgranola p18.00,MM&Msp26.73)thanwhentherewasnot(Mgranola p12.44,MM&Msp11.86;F(1,85)p4.40,p!.04).Controlgroupparticipantsateless
JOURNAL OFCONSUMER RESEARCH
thanthosealongsideathinconfederate(F(1,72)p8.39,p!.01)anddirectionallylessthaniftheconfederatewasobese(seefig.3
).
Discussion
Study1providesevidencethatpeopleanchoronthefoodchoicesofothersintheirenvironment.Participantstooksignificantlymorewhentheyfirstobservedalargeanchorsetbyanotherconsumerversuswhentheymadetheirchoiceinisolation.However,weshowedthattheextentofad-justmentfromtheestablishedanchorismoderatedbythebodytypeoftheotherconsumer.Participantsadjusteddown-wardtoagreaterdegreewhentheotherconsumerwasobesethanwhenshewasthin.Interestingly,weobservenearlyidenticaleffectswhetherthefoodwasperceivedtobehealthyorunhealthy.Itseemsthatsocialinfluenceeffectsinvolvingobesityaregeneralizedtobothhealthyandun-healthyfoods.Ourresultssuggestthatitisportionsizechoicealonethatdrivestheeffectratherthanpairingobesitywithstereotype-consistentfoodchoices.
Wealsofindthatfoodchoicedecisionscarryovertoactualeatingbehavior,evenwhenparticipantswereisolatedwatchingamoviebythemselves.Ourresultsareinconsis-tent,therefore,withanimpressionmanagementaccount.Theparticipantwasescortedintoaseparateroom,whereshewatchedthevideounaccompaniedbyanyone.First,whileitseemspossiblethatparticipantsmayhavechosena portion to conveya desiredimpressiontotheconfederate,itwouldnotexplainwhysheshouldalsoeatmorewhileisolated.Second,socialinfluenceeffectsineating
behaviorhavebeenshowntopersistevenwhentheconfederateisfictionalornotphysicallypresent(Rothetal.2001).
OurfindingsarealsotheoreticallyconsistentwithwhathasbeenshownrecentlybyBergerandRand(2008),whofoundthatconsumersatelessjunkfoodafterbeingtoldthatanoutgroup(vs.aningroup)werethelargestconsumersofjunkfoodoncampus.However,whilethelinkbetweenout-groupsandadjustmentprocesseshasbeendocumented,the
FIGURE3
WEIGHTOFFOODCONSUMEDBYCONDITION(STUDY1)
BODYTYPESANDFOODCHOICES
BergerandRandstudyleavesopenthepossibilitythattheeffecttheyidentifymightbemoderatednotonlybygroupstatusbutalsobythequantitytheoutgroupwaspurportedtoconsume.IntheBergerandRandstudy,theoutgroupwasalwayslinkedtoahighanchorofunhealthyconsump-tion.Inoursecondstudy,wesetupalowanchortoexaminethecasewhereoutgroupsarelinkedtobehaviorswithalessseverehealthrisk.Forinstance,whileBergerandRand(2008)foundthatpeopledivergedawayanddranklesswhentoldthatgraduatestudents(anoutgroup)drankalot,whatiftheyhadinsteadencodedthatgraduatestudentswerelightdrinkers?Wouldpeopledrinkevenlessthantheywouldifthey were told that outgroup memberswereheavydrinkers,orwouldthisbackfire,resultinginpeopledivergingbydrinkingmore?Accordingtoreferencegroupresearch,con-sumersshouldadjusttoagreaterdegreeawayfromtheanchorpointofdissociativegroupsthanaspirationalones(BergerandHeath2008;WhiteandDahl2007).Basedonthistheorizing,wepredictthatconsumerswilladjustupwardfollowingasmallanchor(Wansinketal.1998)butthatthesizeofthisadjustmentwillbemoderatedbythegroupstatusoftheother consumer.Asa result, consumerswillconsumemoreafterseeinganobese(vs.athin)consumerchooseasmallportion.Study2willtestthispredictionandalsoallowustoexaminewhetherconsumerssimplyeatmorefoodinthepresenceofothers(deCastro1994),irrespectiveoftheanchortheysetup,ortheyeatlesswhenthereareobeseothersintheenvironment.
STUDY2
MethodandProcedure
Study2employeda2(confederatebodytype:thinvs.heavy)#2(confederatequantitytaken:littlevs.lots)+1(noconfederatecontrol)between-subjectsdesign.Participantsincluded115undergraduatefemalesfromtheUniversityofBritishColumbia,whocompletedthestudyinexchangefor
$10remuneration.Twoparticipantsindicatedthattheydidnotnoticewhattheconfederatetook,threeategreaterthanthreestandarddeviationsabovethemean(allindifferentconditions),andtwoparticipantsexperiencedafailedmanip-ulationasaresultofexperimentererror,sodatafromtheseparticipantswereexcludedfromtheanalyses.
Theproceduresofstudy2weresimilartothoseofstudy1,withthefollowingexceptions.Instudy2,therestrainedeatingscalewasnowadministeredat the end ofthe survey.Second,insteadofonefoodchoice,participantswereofferedtheirchoiceofsnacksfromsevenbowlsofdifferingsmallcandies.Finally,ratherthanalwaysindulgingwithalargequantitychoice,wealsomanipulatedthequantitythattheconfederatechosesothatshetookasmallportionhalfofthetime.Thispermittedamoredirecttestofouranchoringmodel.Ifparticipantschooseagreaterquantitywhenfollow-ing
anotherconsumerwhotakesalarge(vs.asmall)portion,thiswouldbeevidencethatconsumersanchoronthefoodchoicesofothers.
Theconfederatewashandedabowlfirst,andshechoseher
921
snacksinviewoftheparticipant.Inthelittlefoodcondition,theconfederaterandomlyselectedtwosmallcandies(suchasHershey‘sKissesorsmallsoursoothers)fromthesevenbowls.Inthelotsoffoodcondition,shetookapproximately30smallcandiestotalfromallsevenbowls(rangeis27–
35;Mheavyp31.42,SDp1.71;Mthinp30.90,SDp2.10;F!1).
Measures
DependentMeasures.
Thevariablesofinterestwerehowmany
candiestheparticipanttookandateasafunctionof
theconfederate‘sfoodchoiceandbodytype.Tomeasurehowma
nycandiestheparticipanttook,thenumberofcan-diesremainingafterthesessionineachofthebowlswassubtracte
dfromthenumberthebowlstartedwith
andwhattheconfederatetook.Recallthat,forthepurposesoftestingourhypotheses,weareinterestedinhowtheeffectoftheotherc
onsumer‘s choice ismoderated
byherbodytype.Assuch,ourkeycontrastsexaminewhetherthereisadifferenceacrossbodytypewithinagivenamountchosenbytheother.Thereliabilityoftherestrainedeatingscaleinthissamplewasap.77,andthatoftheperceivedbodysizeoftheconfederateindex(sameitemsasstudy1)wasap.84.Asecondmanipulationcheckassessingthequantitytakenbytheconfederatewasmeas
uredwiththesingle7-pointitem:―Theotherstudenttookhowmuchfromthesnackbar?‖ (nofoodatall/alotoffood).
Results
ManipulationChecks.Themanipulationchecksweresuccessful.Atwo-factoranalysisofcovariance(ANCOVA)usingtheperceivedweightindexasthedependentvariable,amounttakenandconfederatebodytypeasindependentvar-iables,andparticipants‘restrainedeatingorientationandtimeof dayascovariates revealed onlyasignificantmaineffectforconfederatesize(F(1,81)p115.36,p!.001).Themeanpatternshowedthatparticipantsperceivedtheconfederatetobeheavierwhenshewaswearingthesuit(Mp0.10)thanwhenshewasnot(Mp1.93).
AnANCOVAontheperceivedamountofcandytakenbytheconfederatealsorevealedonlyamaineffect(F(1,84)p221.71,p!.001),suchthatparticipantsbelievedtheconfed-eratetookmorecandywhenshetook30candies(Mp5.74)than
whenshetookonlytwo(Mp2.40).
DependentMeasures.AnANCOVAwithquantityofcandiestakenfromthefoodbowlsasthedependentmea-surerevealedamaineffectofquantitytaken(F(1,84)p71.90,p!.001),suchthatparticipantstookmorewhentheconfederatetook30candies(Mp12.62)thanwhenshetookonlytwo(Mp4.72).Moreimportantly,themaineffectwasqualifiedbythepredictedbodytype#quantitytakeninteraction(F(1,84)p8.87,p!.01).Plannedcon-trasts(withinthehighandthelowanchorconditions)indi-catedthat,whentheconfederatetook30candies,participantstookfewerwhenshewasobese(Mp10.60)thanwhenshewasthin(Mp14.45;F(1,39)p5.07,pp.03).However,
922
whenshetooktwocandies,theoppositepatternemerged:participantstookagreaterquantitywhentheconfederatewas obese (Mp5.43) than when she was thin (Mp4.04; F(1,44)p4.22,p!.05).Thecontrolgroup(Mp8.50)differedfromthethin/little(F(1,40)p19.14,p!001),obese/little(F(1,39)p6.78,p!.01),andthin/lotscondi-tions(F(1,38)p10.40,p!.01).Seefigure4.
AnANCOVAonparticipants‘actualconsumptionrevealedanidenticalpatternastheirchoicebehavior.Themaineffectforamounttakenbytheconfederatewasagainsignificant(F(1,84)p22.18,p!.001),suchthatparticipantsatemorecandieswhentheconfederatetookalot(Mp8.12)thanwhenshetookonlyafew(Mp3.72).Therewasalsoamaineffectontherestrainedeatingscale(F(1,84)p7.71,p!.01).However,theselowerordereffectswereagainqual-ifiedbythepredictedbodytype#quantitytakeninteraction(F(1,84)p7.90,p!.01).Plannedcontrastsindicatedthat,whentheconfederatetookalargequantityofcandy,partic-ipantsatefewerwhenshewasobese(Mp6.25)thanwhenshewasthin(Mp9.82;F(1,39)p5.11,pp.03).Incon-trast, when shetook very fewcandies, participantsateagreaterquantitywhentheconfederatewasobese(Mp4.26)thanwhenshewasthin(Mp3.20;F(1,44)p2.76,pp .05,one-tailedtest).Thecontrolgroup(Mp7.88)differedsignificantlyfromthethin/little(F(1,40)p19.72,p!.001)andobese/litt
leconditionsonly(F(1,39)p9.94,p!.01).Seefigure5.
Discussion
Themaineffectofconfederate‘squantityontheamountparticipantstookconceptuallyreplicatespastfindings(seeHermanetal.2003)showingthatthepresenceofotherscausesonetoeatmore(orless)dependingonthepatterntheothershaveset.Weshowthatconsumersanchoronthenorms setbyother consumers‘ consumptionchoices.Thesenormsarepowerful,occurringafterobservingonlyoneotherpersonmakingafoodselection.Peoplegenerallychoselessthantheywouldinisolationafterseeinganot
her
FIGURE4
NUMBEROFCANDIESTAKENBYCONDITION(STUDY2)
JOURNAL OFCONSUMER RESEARCH
FIGURE5
NUMBEROFCANDIESCONSUMEDBYCONDITION(STUDY2)
consumerchooseasmallportion,buttheychosealargerportionthantheywouldhavealoneafterseeingthisindi-vidualtakealargequantity.Moreimportantly,theresultssupportour
hypothesizedinteraction.Whentheconfederatewasobservedtaking a largequantityoffood(settingahighanchor),participantsagainchoseandatelesswhenthatconfederatewasheavythanwhenshewasthin.However,whenshewasseentakingasmallquantity(settingalowanchor), the oppositepatternwasobserved;participantschoseandatemorewhentheconfederatewasheavythanwhenshewasthin.Importantly,thisworkshowsthatdi-vergenceawayfromthebehaviorsofothersdoesnotalwaysmeanreducingthebehavior.Ifconsumersencodeanout-groupasdoingverylittleofsomething,theymaydivergebyincreasingthatbehavior,atleastrelativetoseeinganaspirationalgroupmemberengageinthesameactivity.Ratherthaneatinglessafterseeingaheavypersonchooseasmallamount,participants consumed more,whichiscon-sistentwithadjustingupwardfromalowanchor(Wansinketal.1998).
Study2contributestotheliteratureonidentityandhealth(e.g.,Gerrardetal.2005;GibbonsandGerrard1995;Gib-bonsetal.1998)bypointingtothefactthattheimagesofthosenotengaginginabehavior(i.e.,nonsmokers,non-drinkers)mayalsoaffectthelikelihoodofadoptingthosebehaviors.Whilepriorresearchhasexaminedhowtheimageofsmokers,teenparents,andrecklessdriversaffectspeo-ple‘slikelihoodofadoptingtheirriskylifestyles,theimpactofimagesofthosewhoabstainhasnotbeenexamined.Accordingtoourresearch,theimageofan―uncool‖non-drinkermayactuallyleadtoincreasedbingedrinkingaspeoplestrivetoavoidanidentityassociatedwiththatgroup.Similarly,whileOyserman,Fryberg,andYoder(2007)ex-aminedhowminoritygroups‘healthknowledgeandper-ceivedfatalismcanshiftasafunctionofidentityperceptions(―How Blackis it to eat fried food?‖), theydidnotexaminehowfoodchoiceschangeasafunctionofseeingreferencegroupmemberseating(un)healthyfood,nordidtheyex-amineactualbehavior.
BODYTYPESANDFOODCHOICES
Whiletheresultsofstudies1and2providesupportfortheeffectsofsocial
influenceonfoodchoiceandhowtheseeffectsaremoderatedbythebodytypeoftheothercon-sumer,they aresilent onwhentheseeffectsaremoreorlesslikelytooccur.Iseatinglessinthepresenceofanobesepersonathoughtful,deliberativeeffect,orisitonethatoccurslessconsciously?Aretherecertainindividualorsit-uationalfactorsthatenhancesusceptibilitytoreferencegroupeffects?Researchinsocialcomparisontheory(Fes-tinger1954;KruglanskiandMayseless1990;Wood1989)arguesthatone‘sevaluationoftheselfisrelative,meaningthatpeoplecomparethemselvestothebehaviorandrelevantcuesofothersinformingtheirself-perceptions.Whileabasictenetofsocialcomparisontheoryisthatconsumersmakecomparisonswiththosewhoaresimilartothemselves,otherresearchhasfoundlittlesupportforthisproposition(seeWood1989).Ifitissimilaritytotheoutgroupdrivingtheeffectsofadjustment,wewouldexpectthosewhoareheavythemselves(highBMI)nottoadjusttothesamedegreewiththedissociativeoutgroup,butwehavefoundnoevidence for BMIaffectingourresults.However,peoplealsodifferintheirsatisfactionwiththeirownappearance,andthismaynotcorrelatewiththeirownBMI.Forinstance,personssufferingfromanorexianervosaareoftenobjec-tivelyverythin,buttheyalsohaveaveryhighdegreeofbodydissatisfaction;similarly,thereareconsumerswithahighBMIwhoarequitesatisfiedwithandconfidentcon-cerningtheirappearance. Thereisreasontoexpectthatthemechanismmaybepsychologicalratherthanphysiological.Miller(1984)ar-guedthatpeoplearemorelikelytousedimensionsimportantto their self-definition whenengaginginsocialcomparison.Millerfoundthatwomenwhowereself-schematicongenderuseditasacomparisonpointevenwhenitwasnotrelevant.Forthosedissatisfiedwiththeirbody‘sappearance(lowinappearanceselfesteem[ASE]),thisdimensionshouldbeespeciallyrelevantandshouldthusbeadeterminantofso-cialcomparison.ThistenetissupportedinarecentpaperinwhichTrampe,Stapel,andSiero(2007)found
thatthosehighinbodydissatisfactionhavebeenshowntobemoresensitivetosocialcomparisonfollowingexposuretothebodytypesofothers.Giventhatdissatisfactionwithone‘sownbodyincreasespronenesstosocialcomparison,wehy-pothesizethatthosedissatisfiedwiththeirappearancewillengageinadjustmenttoagreaterdegree.However,amongthoseconfidentintheirappearance,lesssocialcomparisonshouldoccur,resultinginnochangeinbehaviorasafunctionofthebodytypeoftheotherconsumer.Supportforthishypothesiswouldprovideevidencethatouradjustmentpro-cessdoesindeedstemfromsocialcomparison,anditwouldalsocontributetothereferencegroupliteraturebyidenti-fyingamoderatorfordivergence,onebasedondesired,ratherthanactual,groupmembership. Wealso
anticipateaboundaryconditiontotheeffectswehaveidentified.Whilerecentworkhasshownthatcon-sumptiondecisionscanbedrivenbyadivergenceorad-justment away from dissociative outgroups (Berger and
923
Heath2007,2008;EscalasandBettman2005;WhiteandDahl2007),evidenceofthenatureoftheprocessisstilllargelyuntested.Doesdivergencehappenautomatically,ormightitbeaneffortfulprocess?Whatwouldhavehappenedifparticipantsinstudies1and2hadlackedthecognitiveabilitytoadjusttheiranchor?Wouldlessrestrainthavebeenshownafterseeinganobesepersonorderalargequantityoffood?Inotherwords,whilethereissomeevidencethatthedivergenceshowninpaststudiesisasocialprocess,isitalsocognitive?Whileothershaveexaminedhowcognitiveloadmayaffecttheadjustmentfromnumericalanchors(e.g.,Epley and Gilovich2006),itremainsuntestedwhethercog-nitiveprocessesarenecessarytodriveanadjustmentbasedsolelyonthedesirabilityofgroupmembership.Ifweweretoshowthatconsumersfailedtoadjustfortheothercon-sumers‘bodytypeundercognitiveload,thiswouldbestrongevidenceinsupportofouranchoringandadjustmentmodel.Researchonanchoringandadjustmentmodelshasshownthattheadjustmentprocesscanbeattenuatedbyalackofcognitiveresources(GilbertandGill2000;WegenerandPetty1995).Accordingly,wehypothesizethatthedeci-sion todivergeaway from outgroupassociationsisacog-nitiveone(atleastinthecasewherethefocusisonquantitychosen)andthatintheabsenceofcognitiveresources,theadjustmentwedocumentedinstudies1and2shouldbeattenuated.Thus,wepredictthatonlywhenampleprocessingresourcesareavailableshouldweseeanadjustmenteffectbasedonbodytype.Asinstudy1,weagainfocussolelyontheoverconsumptionanchor, asthisposesthegreaterpublichealthrisk.Assuch,we pre-dictathree-wayinteractionbetweentheotherconsumer‘sweight,appearanceself-esteem(ASE),andcognitiveload.Whenprocessingresourcesarenotconstrained,weexpectthatthosewhoarelowinASEwillchooseasmallerportionwhentheotherpersontakesalargequantity.However,amongthosewhoarehighinASE,thiseffectshouldbeattenuated.Withoutavailableprocessingresources,weex-pectthatneithertheweightoftheotherpersonnorASEwillhaveaneffectonparticipants‘foodchoice.Supportforthishypothesiswouldidentifyaboundaryconditiontotheeffects ofidentitysignaling and socialcomparison,namely,thattheadjustmentprocessrequiresconsciousresources.Study3teststhesehypotheses.
STUDY3
MethodandProcedure
Thepredictionsweretestedusinga2(bodytypeofpersoninfrontofyou:thinvs.obese)#2(cognitiveload:lowvs.high)between-subjectsexperimentaldesign,plusameasuredbodysatisfactionvariable.Similartostudies1and2,atruebaselinecondition(lowcognitiveloadandintheabsence
ofasocialother)wasalsorun.Participantsincluded173un-dergraduatestudents(118males,55females)fromArizonaStateUniversity,whocompletedthestudyinexchangeforpartialcoursecredit.
Participantswereinvitedintothelabtoparticipateina
924
studythatpurportedlytestedtheeffectsofmemoryonde-cisionmaking.First,cognitiveloadwasmanipulatedbyhavingparticipantsmemorizea10(highload)or2(lowload)digitnumberthattheywouldbeaskedtorecalllaterintheexperiment(ShivandFedorikhin1999).Followingthe manipulation,participantswere
toldthattheresearcherswouldbeexaminingconsumers‘decision-makingprocesseswhentheymakeselectionsamongmenuitemsandthat―Inordertomakethestudymorerealistic,themenuitems pre-sentedtoyouaredishesactuallyofferedbyaretailer‖(WhiteandDahl2006).Themenucontainedfourflavorsoficecream(FrenchVanilla,DutchChocolate,Cookies‘NCream,andStrawberry),allofwhichwereavailableinfivesizes(x-small,6ounces;small,9ounces;medium,12ounces;large,15 ounces; and x-large, 18ounces).Participantswerethenaskedtoimaginethefollowing
scenariocontainingourmanipulationoftheother‘sbodytype:
Youareinalonglineatanicecreamstore.Ithasbeenalongdayandyouarefeelinglikeyou‘dliketoorderacoldtreat,butyouarenotsureexactlywhatyouwouldliketoorder.Asyouwaitinline,youglanceoveratthemenu.Asyougetclosertothefrontoftheline,youglanceatthepersoninfrontofyou.Althoughtheyarethesamegenderasyou,youcannothelpbutnoticethembecauseoftheirweight:theyareveryoverweight(thin).―Wow!Thatisoneoftheheaviest(thinnest)peopleI‘veeverseen,‖youthinktoyourself.Youstillhavenotmadeupyourmindwhenthepersoninfrontofyouisabouttoorder.Youoverhearthepersoninfrontofyouordertheirsnack:anX-LargeIceCreamCone.Asthepersonreceivestheirorder,theclerkasksyouwhatyouwouldliketohave.
Participantswerethenaskedtochooseasizeandflavor(thelatterwasincludedtomakethescenariomorerealisticanddisguisethestudy‘spurposes)oficecreamthattheywouldchoose.Followingthiswasthenumberrecall,dummyquestionsaboutthemenuandflavorsoffered,scalesmea-suringrestrainedeatingandappearanceself-esteem,andba-sicdemographicinformation.
Measures
Bodydissatisfactionwasmeasuredwiththe6-itemap-pearanceself-esteem(ASE)scaledevelopedbyHeathertonandPolivy(1991).Thiscontainsitemssuchas,―Ifeelsat-isfiedwiththewaymybodylooksrightnow,‖―Iamdis-satisfiedwithmyweight,‖and―Ifeelunattractive.‖Thereliabilityoftheindexinthissamplewasap.87.Thereliabilityoftherestrainedeatingscale(sameasthatusedinstudies1and2)inthissamplewasap.79.Thecor-relationbetweentherestrainedeatingscaleandtheASEscalewas-.45(p!.001),andexcludingrestrainedeatingfromtheanalysisdoesnotaffecttheresults.ThecorrelationbetweenASEandBMIwas-.13andwasmarginallysig-nificantinthissample(pp.09).Becausenumerousstudieshaveshowndifferencesbetweenmenandwomenineating
JOURNAL OFCONSUMER RESEARCH
habits and preferences, aswellassensitivitytobodyimage,wetreatgenderasacovariateinouranalysis.
Toassessthevalidityofourmanipulations,wemeasuredperceivedconfederateweightusingasingleitem:―Theper-soninline infrontof meinthe scenario was(-3)veryoverweight/(3)veryunderweight.Cognitiveloadwasas-sessedwithtwo7-pointscales(―Ifounditchallengingtoreadthescenariowhiletryingtorememberthenumber‖and―Rememberingthenumberwaseasy,‖anchoredby(-3)completelyagree/(3)completelydisagree,withtheseconditemreverse-scored.Thereliabilityofthecognitiveloadmeasurewasrp.63(p!.001).
DependentMeasure.Choicewasassessedbyhavingparticipa
ntsselectasizeoficecreamthattheywouldorder.Giventhatthesizeswerelabeledasrangingfrom
6ouncesto18ouncesin3ounceincrements,sizechoicewas treated asacontinuousvariablerangingfrom1(x-small)to5(x-large).
Results
Totestmoderationwhereoneofthevariablesiscontinuous,analyseswereconductedusinghierarchicalregression(AikenandWest1991;DawsonandRichter2006).Inthefirststep,maineffectsforcognitiveload,ASE,andother‘sbody
type,alongwithparticipantgenderandtherestrainedeatingscalewereincluded.Inthesecondstep,thethreetwo-wayinter-actionsbetweenthefactorswereentered.Finally,inthethirdstep,thethree-wayASE#CognitiveLoad#Other‘sWeightwasentered.Allvariablesweremeancenteredtoreducemulticollinearityandfacilitateinterpretationoflowerordereffects(AikenandWest1991;IrwinandMcClelland2001).
ManipulationChecks.Manipulationchecksshowedthatourmanipulationsweresuccessful.Participantsdeemedtheparticipantheavierwhensheorhewasspecifiedasheavyinthescenariothanwhensheorhewaspurportedlythin(Bp.81,tp15.62,p!.001).Importantly,noothermainorhigherorderinteractionswerepresent.
Ourmanipulationcheckforcognitiveloadshowedthatthoseunderloaddeemedthesituationmoredifficulttoprocessthanthosenotunderload(Bp.59,tp8.32,p!.001).Noothermainorhigherordereffectswerepresent.
DependentMeasure.Resultsonthesizechoicevari-ableshowedasimplemaineffectofgender(bp.39,tp4.79,p!.001)andamarginallysignificantsimplemaineffectforrestrainedeating(bp-.15,tp1.67,p!.10),bothintheexpecteddirections.However,theseeffectswerequalifiedbythepredictedthree-wayASE#CognitiveLoad
#Other‘sWeightinteraction(bp.18,tp2.33,pp
.02;seetable1).Includinggender as a factor does not mod-erateourresults.To
facilitateinterpretationandexpositionoftheinteraction,simpleslopesanalyseswereconducted.
RegressionlineswereplottedforonestandarddeviationaboveandbelowthemeanforASE(Aikenand
West1991;Preacher,Curran,andBauer2006).Examiningthecondi-tionswhere participants were under low