JudgmentandDecisionMaking,Vol.6,No.4,June2011,pp.333 - PDF document

JudgmentandDecisionMaking,Vol.6,No.4,June2011,pp.333–342NudgetonobesityII:MenupositionsinuencefoodordersEranDayanMayaBar-HillelAbstract“Verysmallbutcumulateddecreasesinfoodintakemaybesufcienttohavesignicanteffects,evenerasingobesityoveraperiodofyears”(Rozinetal.,2011).Intwostudies,onealabstudyandtheotherareal-worldstudy,weexaminetheeffectofmanipulatingthepositionofdifferentfoodsonarestaurantmenu.Itemsplacedatthebeginningortheendofthelistoftheircategoryoptionswereuptotwiceaspopularaswhentheywereplacedinthecenterofthelist.Giventhiseffect,placinghealthiermenuitemsatthetoporbottomofitemlistsandlesshealthyonesintheircenter(e.g.,sugareddrinksvs.calorie-freedrinks)shouldresultinsomeincreaseinfavorofhealthierfoodchoices.Keywords:choicearchitecture,menu,middlebias,edgebias,nudge,obesity,positioneffects.1IntroductionObesityisagrowingproblemthroughouttheworld.Fightingitviadietingisapparentlyineffective(e.g.,Mannetal.,2007;Garner&Wooley,1991).Inacompan-ionpaper,Rozinetal.(2011)presentargumentsandfactstosubstantiatethesetwoclaims,whichweshallnotre-peathere.Theythensuggestthatthewaronobesitycouldbenetfromnudges(Thaler&Sunstein,2008),notonlyfromheavyeffortsandinvestmentsinresources.Nudgesaresmall,cheap,easilyimplementableandoftenhardlynoticedchangesinthechoicearchitecture(i.e.,theman-nerorsettinginwhichthechoicesetispresented)thatdonotaffectthechoicesetitself,yetaffecttheappealofdifferentoptionsinit.Rozinetal.'snudgetonobesityisverysimple:ifyouwanttoincreaseordecreasethepop-ularityofafooditem,makeiteasierorhardertoaccess,respectively.Inthesamespirit,thepresentpaperexploresanotherpossiblenudgetonobesity.Weshowthatplacingafooditemonamenuatthebeginningortheendofitscategoryincreasesitspopularitycomparedtoplacingitinthemiddle.Restaurantspresentcustomerswithlistsoftheiroffer-ings.Whenthemenuisdisplayedinwriting,itemsarepresentedsimultaneously.Whenawaiterrecitestheday'sspecials,itemsarepresentedsequentially.Ourstudyin-volvedonlyprintedmenus.Menuitemsmaybeorga-nizedinvariousways,suchasbytype(e.g.,Soups;Sal-ads;etc.),oraccordingtomainingredients(Fishdishes; ThisresearchwasanMAthesisconductedbytherstauthorun-derthesupervisionofthesecondauthor.WethankCaféB.fortheirkindcooperation.WethankPaulRozinandthejournaleditorforsograciouslylettingusattachNobesityIItoNobesityI,andfortheirhelpdoingso.TheHebrewUniversity,JerusalemyCorrespondingauthor:Address:MayaBar-Hillel,CenterfortheStudyofRationality,TheHebrewUniversity,Jersualem91904,Israel.Email:maya@huji.ac.il.Vegetariandishes;etc.).Withineachcategorytheyaretypicallylistedinverticalordering.Whendesigningmenus,doesthisordermatter?Onemayseekanswersfromtwokindsofsources—the“howto”literatureonmenudesign,andthepsycho-logicalliteratureonpositioneffects.Familiarpositioneffectssuchasprimacyandrecencyrefertostimulipre-sentedsequentially,andtheirdependentvariableisnotusuallychoice.Buttheeffectcalled“edgeavoidance”(Rubinstein,Tversky&Heller,1986),“centralityprefer-ences”(Shawetal.,2000),“middlebias”(e.g.,Attali&Bar-Hillel,2003),or“center-stageeffect”(Valenzuela&Raghubir,2009)referstochoicefromamongsimultane-ouslypresentedoptions—andthevariousnamesindicatethetypicalndings:“Peoplechoosingfromanarrayofidenticaloptionsreliablypreferthemiddleones”(Chris-tenfeld,1995).Whenitemsarenotidentical,theeffect'smanifestationisthatwhenoptionsarepresentedinthemiddleofanarraytheyarechosenmoreoftenthanwhentheyarepresentedonitsedges.Thesestudiesdonot,ofcourse,applytooptionsforwhichpositionmaybeinherentlyimportant,suchasthe-aterorairplaneseats,skyscraperoors,restauranttables,orplaceinqueues.Rathertheyuseoptionsforwhichitishardtoimaginewhypositionwouldmatter,suchas:i.inwhichof4opaqueboxespeoplechoosetohide,orseek,a“treasure”(Rubinstein,Tversky&Heller,1986);ii.sim-ilarly,inwhatpositionpeopleplace,orguess,answersinmultiple-choicetests(Attali&Bar-Hillel,2003);iii.whichgoodtheychoosefromasetofidentical(Chris-tenfeld,1995;Shawetal.,2000)ornon-identical(Valen-zuela&Raghubir,2009)goodsoffered;iv.whatstalltheyheadforinapublicbathroom(Christenfeld,1995);etc.Allthesestudiesfoundthatplacinganiteminthemid-dle,ratherthantheedges,ofthechoicesetenhanceditspopularity.333 JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus334Table1:Itemorderinthefourmenus:A=Appetizers;E=Entrées;S=Softdrinks;D=Desserts. Baselinemenu:A1,A2,A3,A4E1,E2,E3,E4,E5,E6,E7,E8,E9,E10S1,S2,S3,S4,S5,S6D1,D2,D3,D4,D5,D6,D7,D8Mirrormenu:A4,A3,A2,A1E10,E9,E8,E7,E6,E5,E4,E3,E2,E1S6,S5,S4,S3,S2,S1D8,D7,D6,D5,D4,D3,D2,D1Inside-Outbase:A2,A1,A4,A3E5,E4,E3,E2,E1,E10,E9,E8,E7,E6S3,S2,S1,S6,S5,S4D4,D3,D2,D1,D8,D7,D6,D5Inside-Outmirror:A3,A4,A1,A2E6,E7,E8,E9,E10,E1,E2,E3,E4,E5S4,S5,S6,S1,S2,S3D5,D6,D7,D8,D1,D2,D3,D4 Weareawareofonlythreeexceptionsinwhichthereseemstobeanadvantagetobeingrstorlastinasimul-taneouslypresentedchoicesetratherthaninitsmiddle.NisbettandWilson(1977)askedtheirsubjectstocon-sideralineararrayof4identicalpairsofstockings(afactofwhichtheirsubjectswerenotaware),andserendip-itouslyfounda“pronouncedleft-to-rightpositionef-fect,suchthattheright-mostobjectinthearray[whichwasalsothelastperused]washeavilyover-chosen”(p.243)—namely,“last-is-best”.Incontrast,KoppellandSteen(2004)analyzedrealballot-votingdatathatwasal-mostlikeacontrolledstudy,inasmuchas“theorderofcandidates'nameswasrotatedbyprecinct”(p.267),andfoundthat“candidatesreceivedagreaterproportionofthevotewhenlistedrstthanwhenlistedinanyotherpo-sition”(p.267)—namely,“rst-is-best”.Finally,Chris-tenfeld(1995)askedrespondentstochoosearoutebe-tweentwopoints,eitheronhypotheticalmapsorforreal.Thedestinationpointcouldnotbereachedbywalkingastraightline,butthepathstobechosenfromhadthesametotallengthandnumberofturns.Respondentsshowedapreferenceforthepathreachedbymakingtherstturnaslateaspossible.Thepathscannotbeclassiedintorst,last,ormiddle,butthepossiblepointsoftakingtherstturncan,andinthatsense,respondentspreferredthelast.Incontrasttoalltheabove-mentionedndings,therestauranttradepublicationsonmenusadvocatebothedges(namely,therstandlast)asthepositionswhereoneshouldplacetheitemswhosepopularityonewantstoenhance(e.g.,“Amenuitem'spositionwithinalistcanalsoaffectsales.Peopletendtorememberthetoptwoitemsonalistandthebottomitem...”,Panitz,2000,p.82;“Peopledonotreadmenus,theyscanthem...Asaresult,themostfrequentlyselecteditemsarethoseintherstandlastpositioninthecategorylist...”Main,1998,p.80).Theserecommendations,however,wereneverbackedbyresearch,andnone,tothebestofourknowledge,exists(Panitz'sclaimiscertainlyvalid,butitisnotclearwhyoneneedstorelymuchonmemorywhenchoosingfromamenu).Moreover,whenKincaidandCorsun(2003)attemptedtoputotheracceptedtruthsregarding“theimpactofmenulayoutonitemsales”toanempiricaltest,theirtitlequestion,“Areconsultantsblow-ingsmoke?”(p.226),wasansweredintheafrmative.However,sincetheydidnotstudy“edgebias”speci-cally,wehavenodirectmenuresultstoeithercontrastwithoraddtothe“edgeavoidance”wereportedabove.Inthepresentstudy,therefore,wedidnothypothe-sizeabiaseitherinfavoroforagainstmiddlepositioneditems,butrathercheckedwhetheroneexists,using2-tailedsignicancetesting.2Study12.1MethodParticipants.240HebrewUniversitystudents,ages19–35,52%female,wererecruitedindividuallyaroundthecampus.Participantswereassignedatrandomtothe4conditions,inequalnumbers.Design,stimuliandprocedure.Fourmenuversionswereprepared,differingonlyinorderofitempresenta-tionwithincategory.Themenuoffered4appetizers(A),10entrées(E),6softdrinks(S)and8desserts(D),inthatorder.ThenamesoftheitemsandtheirdescriptionswerecopiedfromthatofanIsraelipizzeriachain.Nopricesweredisplayed.Thefourmenus(inHebrew)appearedinfourdifferentorders,shownschematicallyinTable1.CallonetheBaseline(arbitrarilydesignated).Thentheotherthreewere:Mirror(thatreversedtheBaselineordercompletelywithineachcategory);Inside-OutBase(thatreversedtheBaselineorderwithinthetophalfandwithinthebottomhalfofeachcategory,butnotthetopandbottomhalvesthemselves,therebyturningmiddleitemsintoextremeitemsandviceversa);Inside-OutMirror.Eachparticipantreceivedasinglemenu,andwasaskedtochooseasingleitemfromeachcategory.Theywerepromisedthatoneparticipantwouldbechosenbylottery,thewinnertoberewardedwitharealmealatthepizze- JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus335ria,consistingofhisorherexactquestionnairechoices.Therewardwassufcientlymotivatingthatparticipantsvolunteeredthefewminutesneededtomaketheirmenuselections.2.2ResultsTable2showsthepopularityoftheitems(namely,thenumberoforderstheyreceived)asafunctionoftheirwithin-categoryposition.Sinceweusedonlyfourdif-ferentorders,notallitemsappearedinallpossibleposi-tions(exceptingtheAppetizers).Allitemsdid,however,appearbothinthetophalfandinthebottomhalfoftheircategoryoffers(thiswasassuredbytheMirrorreversals).Additionally,itemsthatwereeitherrstorlastintheircategory(A1,A4;E1,E10;S1,S6;D1,D8)alwaysex-changedpositionswithitemsinitsmiddle(e.g.,A2,A3;E5,E6;S3,S4;D4,D5,respectively).Becauseallcat-egorieshadanevennumberofofferings,this“middle”consistedof2items.Amongtheitemsthatwerenei-therattheextremesnorinthemiddle,somenonethelessmovedcloserto,orfurtherfrom,themiddle(aswhenE2,E9andD2,D7exchangedpositionswithE4,E7andD3,D6,respectively),butsomeretainedtheirpositionvis-à-visthemiddleortheextremesinall4menuorders(thiswasthecaseforE3,E8,S2,andS5).Thelattercouldnotprovidedataforourhypothesis,soalthoughtheirdataarereportedinTable2,theywereignoredinthepositionanalyses.InTable2,M+andM-designatethemiddlepositions,whereM+isatopM-;M++andM--designatepositionsjustaboveorjustbelowthemiddleones,respectively;PUisthepenultimateposition.SinceTable2givesthedatainfull,allquestionscanbeansweredfromitdirectly.Nonetheless,forreaderease,wepreparedanothertablederivedfromit.Table3omitsthehypothesis-irrelevantmenuitems(E3,E8,S2andS5).Italsoomitsmenuitemswhichwerechosenbyfewerthan24(i.e.,10%)oftheparticipants(E1,E2,E4,E5,E6,S4,D1andD4).Thisremoved107obser-vationsfromthetable,whicharenonethelessincludedinallanalysesatthecategorylevel.Table3alsoconvertsfrequenciestopercents,tofacilitatebetween-itemcom-parisons.Finally,itemorderisrearranged,fromtheitemshowingthegreatestbenetforanextremeposition,D7,tothatshowingtheleastbenetforanextremeposition,D5(which,infact,wastheonlyitemthatshowedadecitfortheextremeposition).Thisisshownintherightmostcolumn(whichsumsthepercentofchoicesofthetopandbottomitems).Allbutoneitem(D5Carmela)ofthe16individualitemsinTable3(andallbut4itemsofthe28itemsinTable2)showthatanindividualitem,nomatterhowFigure1:Themeanpercentofchoicesmadewhenanitemwasonthetoporbottomvs.inthemiddleofitsfoodcategory,sortedbycategorytype.*indicatesp.05,twotailed. AppetizersEntréesDessertsSoft drinksTotal Extremes% choice 010203040506070 ** popularorunpopular,benetsfrombeingplacedatthebeginningorendofitscategorylistratherthanatitsmid-dle(2-tailedsigntest,p=0.0005andp=0.0002,respec-tively).Althoughtheindividualitemsshowthatadvan-tagetorangeupto64%,nonewassignicant.Atthecategorylevel,theadvantagerangedupto57%,withtheGrandTotalbeingasignicant56%overall(p.001by2-tailedsign-test).Eveninthecategoryof“unpopularitems”(namelythoseorderedsoinfrequentlythattheywereleftoutofTable3),theadvantagewas54%.ThecategoryresultsarevisuallydisplayedinFigure1.Althoughthispaperisconcernedexclusivelywithmiddle-vs.-edgespositioneffects,thedataaffordanop-portunitytocheckforprimacyandrecencyeffectsaswell.None,however,wasfound:50.5%ofthechoiceswerefortheitemsatthetophalfoftheircategoryvs.49.5%forthebottomhalf.Moreover,theTotalper-centsshowastrikingsymmetryaroundthemiddle(seethebottomlineofTable3).Itempopularityasafunc-tionofdistancefromthemiddlewas21%–20%(rstvs.last);7%–7%(secondvs.penultimate);5%–6%(M++vs.M--);17%–16%(M+vs.M-).Thesedataalsoshow,somewhatsurprisingly,thattheedgeadvantagewasnotlargerforrst/lastpositionvs.middlethanforsecond/penultimatepositionvs.near-middle(noticehowitemsoftheformerandthelatterkindinterleavethroughoutthetable,whichisorderedbythemagnitudeoftheeffect).Theenhancedpopularityoftherst/lastpositionswas55%(outof674observations),andthatofthesecond/penultimatepositionswas57%(outof JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus336Table2:Numberofordersasafunctionofitemposition. Numberofordersasafunctionofitemposition ItemnameN1st2ndM++M+M-M--PULast A1EnsaladaVerde5514-1-1110--20A2Caprese10129--2223--27A3Empanada5016--1012--12A4Colorada347--78--12Appetizertotal24066--5053--71 E1Cumbia82--32--1E2Milonga9-32--22-E3Lambada47--12,132--10,12--E4Tango15-45--33-E5Candombe144--31--6E6Friconne6317--1615--15E7Samba24-58--47-E8Salsa2--0,1--0,1--E9Meringue28-106--66-E10Rumba309--67--8Entréetotal19122,2732222112,1428251510,131830 D1Cookiesice-cream214--64--7D2Choc-chipice-cream24-75--66-D3Strawberryice-cream26-73--97-D4Cappuccinoice-cream164--34--5D5Carmela306--89--7D6Nicoletta69-2211--1620-D7Chaja33-85--713-D8DulcedeLecheice-cream217--54--5Dessertstotal2402144242221384624 S1OrangeJuice5018--1211--9S2Sprite5-1,1----2,1-S3CocaCola7521--1614--24S4Fanta30--11--1S5Sodawater2-0,0----1,1-S6Lemonade10534--2423--24Softdrinkstotal2334,373-1,1-5349--3,258 Grandtotal90426,30192661,14513,141531485310,13643,2183 1Here(andintheothertables),-occursinacellthatwasnotrepresentedbyaposition.2Doubleentriesinacelloccurwhentheitemappearedinthesamepositionintwodifferentmenuorders. JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus337Table3:Itempopularityinpercentsasafunctionofitemposition. Itemposition Itemname1st2ndM++M+M-M--PULast%extreme D7Chakha-2415--2139-64A1EnsaladaVerde25--2018--36621D6Nicoletta-3216--2329-61S3CocaCola28--2119--3260E9Meringue-3621--2121-57E10Rumba30--2023--2757A3Empanada32--2024--24562A4Colorada21--2124--35562A2Caprese29--2223--2755S6Lemonade32--2322--2355D2Choc-chipice-cream-2921--2525-54S1OrangeJuice36--2422--1854D3Strawberryice-cream-2712--3527-54E6Friconne27--2524--2451E7Samba-2133--1729-50D5Carmela20--2730--2343 Appetizertotal28--2122--3057*Entréetotal17121115137.99.41653Dessertstotal8.818109.28.816191056Softdrinkstotal31--2321--2556 Grandtotal217.3517165.97.12056* 1Hereandelsewhere,irregularitiesinsums(e.g.,25+36=61,buttableshows62)areduemerelytoroundingerrors.2Here,andelsewhere,apparentties(e.g.,betweenA3andA4)werebrokenbythenextdigit,notshown.*indicatessignicanceatthe.05levelorbetter,2-tailed.228observations;thedifferenceisnotsignicant).3Study2ClearandunambiguousastheresultsofStudy1are,theyarenonethelesshypotheticalchoices,made—forbetterorforworse—undercontrolled,butarticial,conditions.Incontrast,Study2wasrunontherealchoicesofrealcus-tomersinaTelAvivcafé.Themenu,naturally,listedprices(inNewIsraeliShekels;seeAppendix).Theman-agementcooperatedwiththestudymanipulationsintworespects.First,theyagreed,forthestudyperiod,toalter-nate(acrossdays)theusualmenuwithoneidenticaltoitineveryrespectexceptfortheorderofsomemenuitems.Second,theyrecordedcustomerordersforourbenet,asdetailedbelow.Thecaféisasmalltown-centercoffee-shop,open7daysaweek,from8amtilloneortwohoursaftermid-night,andcateringprimarilytostudentsandyoungpro-fessionals.Itconsistsofseventablesandacounter,andoffersawideselectionofhotorcolddrinksbasedoncof-fee,tea,orice-cream,andservedwithorwithoutalco-hol,aswellasaselectionofsodasandfreshjuices.Italsoservesdessertssuchascakesandice-cream.Therewerealtogetherabout60listingsonthemenu,someofwhichstandformultiplepossibilities(e.g.,“coffee”canbehadinacaffeinatedanddecaffeinatedversion;“ice-cream”comesinmanyavors;somecanneddrinkscanbehadinadietversion;etc.).Thestudyfocusedononly3categories(whichappearassuchonthemenu):Coffee JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus338Table4:20menuitemsandthenumberoftimestheywereorderedintwomenuversions.A=Alcoholiccoffee;S=Softdrinks;D=Desserts. BaseMenuFreqI/OmenuFreq A1FrangelicoQuarto2A28A2Kahlua/GrappaQuarto3A12A3FrangelicoEspresso4A417A4IrishCream17A35 Alcoholiccoffeetotal26-32 S1Mineralwater51S378S2SanPelegrino85S287S3Coke/DietCoke50S145S4Sprite/DietSprite41S65S5Nutdrink5S58S6Passionfruitdrink6S446 Softdrinkstotal238-269 D1Croissant18D535D2Brownie16D424D3Coffeecake29D335D4Bananabread28D211D5riedfruitcake29D19D6Cookieplatter20D108D7Carrotcake39D92D8Tiramisu2D84D9Chocolatesoufe2D742D10Fruitsalad7D626 Dessertstotal190-196 withalcohol—4items;Softdrinks—6items;Desserts—10items.Anexactreplicaofthemenu,translatedintoEnglish,appearsintheAppendix.Duringtheperiodofthestudy,allordersmadefromthesecategorieswererecorded,separatelyanddiscreetly,bythewaiters.Ordersplacedwithoutresorttothemenu(e.g.,bysomeregularcustomers)werenotrecorded.Thestudytookplaceinsummer,thoughnotonadailybasis.Eachformwasgivenon15days,alternating,withexactlythesamedistributionoverthedaysoftheweek(thuscontrollingforpossiblesystematicvariationsindaysoftheweek).TheBaselinemenuisthecafé'sstandardmenu.TheInside-Outversionchangedonlythepositionsoftheitemsinthestudy's3targetcategories,exchangingitemsonthetwoendsofthecategorywithitemsfromthemiddleofthecategory,asshowninTable4.3.1MethodParticipants.Participantsweretheself-selectedclientelewhoorderedfromthethreetargetcategoriesduringtheperiodwhenobservationswerecollected.Wecannotsayexactlyhowmanycustomerswereinvolved,onlyhowmanyorderswereinvolved(459fromtheBasemenu,and492fromtheI/O[Inside-Out]menu).Somecus-tomersmayhaveorderedmorethanoneitemduringasinglevisit,andsomemayhavebeenrepeatcustomers,butnorecordsweremadeofthesepossibilities.Design,stimuliandprocedure.Thereweretwover-sionsofthemenu,inwhich20itemsoutofthe60inthemenudifferedonlyintheirorderwithintheircate-gory(seeTable4).Datawerecollectedwithnopartic-ularprotocol.Weareawarethataskingthecafé'swait-ers(whonecessarilywerenotblindtothemanipulation,butwereblindtothehypothesis)torecordthedataisapossiblesourceofnoise,addingtothenaturallyoccur-ringnoisefromhavingnocontroloverthecustomersandtheirchoices.However,itishardtoimaginehowanybi-ases,includingthosethataretimesensitive(e.g.,moresloppinessattheendoftheday,duetowaiterfatigue,orlesssloppinessattheendoftheday,duetowaiter“warmup”),mightinteractwithourvariableofinterest.3.2ResultsTable4liststhecategoriesanditemnamesthatwerema-nipulatedinStudy2.Itshowsthetwoorderingsofthemenuside-by-side,andthenumberoftimeseachitemwasrequestedduringthestudyperiod.Table5wasderivedfromTable4inthesamemannerasTable3wasderivedfromTable2.Thus,itdoesnotshowthehypothesis-irrelevantitems(S2,S5,D3,D8),anditdoesnotshowitemswhosetotalnumberofordersoverthestudyperiodfellunder24(A1,A2,A3,S6,D9,D10—atotalof56orders).Thelatterwerenonethelessincludedinthecategory-levelpercents,shownatthebot-tomofthetable,andinallanalyses.Frequencyofcus-tomerorderswasreplacedbypercentofallordersfromthatcategory,anditemsarelistedinthetablefromthatshowingthemostbenetforanextremelocation(D1Croissant),tothatshowingtheleastbenet(D4Bananabread).Onlyoneitemofthe10itemsinTable5(D4;A4istied),andonly2ofthe18itemsinTable4(D4andD10;A1,A4andD8aretied),showanadvantagetoamid-dleposition(2-tailedsigntest:p=.04;p.007,respec-tively).Atthecategorylevel,thisadvantagerangesupto JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus339Table5:Itempopularityinpercentsasafunctionoftwoitempositions. PopularityinpercentsasafunctionofitemItemN1st2ndM++M+M-M--PULast%extreme D12767--33----67S312861--39----61*D227-5941-----59D646----43--5757D564----45--5555S487----48--5353S19653--47----53D781-----4852-52A434----50--5050D452-4654-----46Atotal5817--8.636--3855Stotal32240--3014--1656*Dtotal31617131212913141054Grandtotal696285.75.620145.96.31555* *indicatessignicanceatthe.05levelorbetter,two-tailed.56%,withagrandmeanof55%—justatadlowerthantheadvantagefoundinStudy1(56%).Atthecategorylevel,onlySoftdrinkswassignicant.Allinall,Study2showedmuchthesameadvantagetobeingplacedatthebeginningorendofamenucategoryaswasshowninStudy1.Inthisstudy,therewasalargergainwhenanitemmovedfromtheexactmiddletotheextremeend(55%),thanwhenitmovedfromthenear-middletothenear-end(51%;n.s.).Table5alsoseemstoshowanadvantagetobeinglistedinthetophalfofacategory(59%ofthechoices;p0.0001,2-tailedsigntest).However,wedidnotuseaMirrororderinghere,sotheeffectofsideisconfoundedwiththenatureoftheitemsthemselves,andthuscannotbeattributedtoposition.Therefore,thereisnopointincheckingforsymmetryinTable5aswedidinTable3.ThecategoryresultsarevisuallydisplayedinFigure2.4DiscussionMenuconsultantsmaybewronginsomeoftheirrecom-mendations(Kincaid&Corsun,2003;Reynolds,Merritt&Pinckney,2005),butapparentlytheyarenotwrongontheparticularonestudiedhere.Intwostudies,oneelicit-inghypotheticalchoicesandoneobservingrealchoices,wefoundthatplacingmenuitemsatthebeginningorendoftheircategoryincreasestheirpopularitybyabout20%(namelythegainfrom45%ofthetimewhenanitemap-pearedinthemiddleofitscategory,to55%ofthetimewhenitappearedatoneoftheendsofitscategory).Thiseffectdependedneitheronthekindoffoodsinthecat-egory,noronitssize(4items,6items,8itemsor10items—albeitthetwoweresomewhatconfounded).Wecannotofferasatisfyingexplanationforwhymenuchoiceswoulddifferfromthemanyothercontextsinwhichdifferent,usuallyevenopposite,biaseswerefound,surveyedinourintroduction.Indeed,wefoundnotonesinglestudythatshowedanadvantagetobeingbothrstandlastoverbeinginthemiddle.Wearedis-missing,ofcourse,thevastliteratureontheserialposi-tioneffect,because“position”thereistemporal,notspa-tial,andthedependentvariablesarerelatedtomemory,nottochoice.Wemustalsodismisstheresultsinourcompanionpaper,althoughRozinetal.(2011)alsofoundanedgeadvantage—placingfooditemsateithersiderowofathree-rowfooddisplay,ratherthaninitsmiddlerow,enhancedtheirpopularity.ButRozinetal.hadacon-vincingphysicalexplanationfortheirresults:theitemsinthemiddlewerephysicallyhardertoaccess,requiringalongerreachunderaplasticshield(“SneezeGuard”).Alas,theiraccountcannotbeappliedtochoicefromamenu.Fromtheotherstudiesofpositioneffectsinsimultane- JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus340Figure2:Themeanpercentofchoicesmadewhenanitemwasonthetoporbottomvsinthemiddleofitsfoodcategory,sortedbycategorytype.*indicatesp.05,twotailed. Coffee w/AlcoholSoft drinksDessertsTotal Extremes% choice 010203040506070 ** ouschoice,weshallalsoputasidethosethatinvolvehide-andseekstrategies(e.g.,Rubinstein,Tversky&Heller,1986;Attali&Bar-Hillel,2003),becauseamenuiscer-tainlynotasetofoptionshidingone“correct”optionforthechoosertodiscover,butratheradisplayinvitingthechoosertosuithim-orher-selfonly.Finally,weshallputasidethosewherealloptionsarethesamebutfortheirposition,whichisnotthecaseformenus.Oftheremain-ingstudies,perhapstheclosestistheballotvotingstudy(Koppell&Steen,2004),andthetwoconsumer-choicestudies(Nisbett&Wilson,1977;Valenzuela&Raghu-bir,2009).Alas,thesedidnotndconsistentpositioneffects(theeffectsfound,respectively,wereprimacy,re-cency,and“centerstage”),andcorrespondingly,didnotofferconsistentaccounts.Moreover,theyalsofoundtheliteratureinconsistent,andofferedtheiraccountsspecula-tively(e.g.,Koppell&Steen:“theliteratureiscontradic-tory,withnoclearpatterninthendingsacrossstudies”,p.268;Nisbett&Wilson:“Preciselywhythepositioneffectoccursisnotobvious.Itispossiblethatsubjects[were]“shoppingaround”,holdingoffonchoiceofearly-seengarmentsontheleftinfavoroflater-seengarmentsontheright”,p.244;Valenzuela&Raghubir:“priorre-searchexaminingtheeffectofphysicalpositionofprod-uctsinanarrayhasfoundinconsistenteffects([referencelistfollows]),andisdividedastowhypositioneffectsoccur....evidence[forofferedaccounts]...islacking...”,p.185).Atthistime,thetopicofpositioneffectsinsimulta-neouschoiceisfarfrombeingwellunderstood(seeBar-Hillel,2011),andadditionalresearchisclearlycalledfor.Althoughourresultsmaypresentlylackanexplana-tion,webelievetheyarerobustenoughtowarrantcon-dence,especiallywithaneyetoreal-worldapplicationratherthantheory.Webelievethatwehavepresentedenoughevidencetorecommendanudge:Putthefoodyouwanttoencourageattheextremesofthemenulist-ings.Thisrecommendationappliestothelistingswithincategory;wehavenotstudiedwhetheritissimilarlypos-sibletonudgepeopleacrosscategoryboundaries.Ifany-onewhowishestoadoptthisnudgetonobesityremainsskeptical—itisridiculouslyeasyandcheaptotestitintheirspeciccontext:changemenupositions,andsee.Nudgescanbeusednotonlytopromotehealthierfoodchoices,butanyotheragendaaswell(higherearnings;fasterturnoverformoreperishablefoods;etc.).Itisuptoustonudgetonobesity.Rozinetal.(2011)showinquantitativedetailhowevennegligiblysmalleffectscanaccumulateovertimetilltheyaresignicant.Theyalsoaddressthevariouscaveatsthatcanberaisedagainstat-temptstoaffectfoodintakebyasinglenudge.Theiranal-ysisappliestoournudgeaswell.And,liketheirs,itcanallbedonedirtcheapandwithminimaleffort.ReferencesAttali,Y.&Bar-Hillel,M.(2003)Guesswhere:Thepo-sitionofcorrectanswersinmultiple-choicetestitemsasapsychometricvariable.JournalofEducationalMeasurement,40,109–128.Bar-Hillel,M.(2011,forthcoming)Location,location,location:Positioneffectsinsimultaneouschoice.InBrun,W.,Keren,G.,Kirkebøen,G.,&Montgomery,H.(2011).PerspectivesonThinking,Judging,andDe-cisionMaking.Oslo:Universitetsforlaget.Christenfeld,N.(1995)Choicefromidenticaloptions.PsychologicalScience,6,50–55.Garner,D.&Wooley,S.(1991).Confrontingthefailureofbehavioralanddietarytreatmentsforobesity.Clini-calPsychologyReview,11,729–780.Kincaid,C.S.&Corsun,D.L.(2003)Areconsultantsblowingsmoke?Anempiricaltestoftheimpactofmenulayoutonitemsales.InternationalJournalofContemporaryHospitalityManagement,15,226–231.Koppell,J.G.S.&Steen,J.A.(2004)Theeffectsofballotpositiononelectionoutcomes.TheJournalofPolitics,66,267–281.Nisbett,R.E.&Wilson,T.D.(1977)Tellingmorethanwecanknow:Verbalreportsonmentalprocesses.Psy-chologicalReview,84,231–259.Main,B.(1998)Menuitemreplacement.InW.R.Mar-vin(ed.),50ProvenWaystoBuildMoreProtableMenus.GigHarbor,WA:HospitalityMasterspress.Mann,T.A.,Tomiyama,J.,Erika,W.,Lew,A.,Samuels,B.,&Chatman,J.(2007).Medicare'ssearchforef- JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus341fectiveobesitytreatments:Dietsarenottheanswer.AmericanPsychologist,62,220–233.Panitz,B.(2000)Doesyourmenuattractorrepeldin-ers?In:Readingbetweenthelines:Thepsychologyofmenudesign.RestaurantsUSA,8,81–87.Reynolds,D.,Merritt,E.A.,&Pinckney,S.(2005)Un-derstandingmenupsychology:Anempiricalinvesti-gationofmenudesignandconsumerresponse.Inter-nationalJournalofHospitality&TourismAdministra-tion,6,1–10.Rozin,P.,Scott,S.,Dingley,M.,Urbanek,J.K.,Jiang,H.,&Kaltenbach,M.(2011)NudgetonobesityI:Minorchangesinaccessibilitydecreasefoodintake.JudgmentandDecisionMaking,6,323–332.Rubinstein,A.,Tversky,A.,&Heller,D.(1996)Naïvestrategiesincompetitivegames.In:W.Guth(ed.),Un-derstandingstrategicinteraction—EssaysinhonorofReinhardSelten.394–402.Berlin:Springer-Verlag.Shaw,J.I.,Bergen,J.E.,Brown,C.A.,&Gallagher,M.E.(2000)Centralitypreferencesinchoicesamongsimilaroptions.TheJournalofGeneralPsychology,127,157–164.Thaler,R.H.,&Sunstein,C.R.(2008).Nudge:Im-provingdecisionsabouthealth,wealth,andhappiness.NewHaven,CT:YaleUniversityPress.Valenzuela,A.&Raghubir,P.(2009)Position-basedbe-liefs:Thecenter-stageeffect.JournalofConsumerPsychology,19,185–196. JudgmentandDecisionMaking,Vol.6,No.4,June2011Positioneffectsinmenus342Appendix:ReplicaofmenuusedinStudy2