86ROSEANDDRICKAMER Ropartz,P.Demonstrationofanincreaseoflocomotor activity ofgroupsoffemalemicetotheodorofagroupofstrange males.C.R.Hebd.SeancesSer.DSci.Natur.,1968,267, Scott, J.W.,Pafaff,D.W.Behavioral and electrophYsiologicai offemalemicetomale urine odors.Physiology and BehaVIor, 1970, 5, 407-411. Stern,J.J.Responsesofmaleratstosexodors.Physiology Bulletinof thePsychonomicSociety 1975,Vol.5(1),86-88 Behavior, 1970, 5,"519-524. Vandenbergh,J.G.Maleodoracceleratesfemalesexual maturation in mice.Endocrinology, 1969, 84, 658-660. Whitten,W.K.Pheromones andmammalian reproduction.In.E. McLaren(Ed.),Advancesinreproductivephysiology.Vol.I, NewYork, Academic, 1966. (Received for publication September 20,1974.) ELIZABETHF. LOFTUS Universityof Washington,Seattle,Washington98195 and GUIDOZANNI New School for Social Research, NewYork,NewYork 10011 Twoexperimentsarereportedinwhichsubjectsviewedafilmof anautomobileaccidentandthen answeredquestionsabouteventsoccurringinthefilm.Relativetoquestionscontaininganindefinite article(e.g.,Did you see abroken headlight?), questions which contained adefinite article (e.g.,Did you seethebrokenheadlight?)produced(1) feweruncertainor"Idon'tknow"responses,and(2) more "recognition"of events that never,infact,occurred. The results,which are consistent with the viewthat questionsaskedsubsequenttoaneventcancauseareconstructionin one's memory of that event, have important implications forcourtroom practices and eyewitness investigations. Anautomobileaccidentisahighlyc,omplexand suddeneventoftenlastingonlyafewseconds.Isour perception,recollection,andverbalizationofsuchan incidentanidenticalcopyoftheoriginalevent?Most researchersinthefieldof humanmemorywouldagree thattheanswertothisquestionis"no."Thereare numerouswaystoinfluence(andoftendistort drastically)therecollectionsof awitness.Onerelatively easywayistovarythemethodbywhichthe recollectioniselicitedortovarytheforminwhich questions areasked about the recollection. Muchof theresearchinthisareahasindicatedthat whenpeopleareforcedtoanswerspecificquestions, ratherthantoreportfreely,theirreportsaremore completebut lessaccurate(Cady,1924; Gardner,1933; Marquis,MarshallOskamp,1972;Marston,1924; Whipple,1909). Furthermore, the accuracy of an answer toaspecificquestioncanbenoticeablyinfluencedby thewordingof thequestionitself.Bychangingoneor twowordsinaquestion,clear-cutvariationshavebeen showntoappearinasdiverseareasasasubject's report ofhypnoticexperiences(Barber,1969)andinhis estimatesofthespeedofamovingvehicle(Loftus& ThisresearchwassupportedbytheUrbanMaSs TransportationAdministration,DepartmentofTransportation, GrantNo. WA-ll-0004.Requestsforreprintsshouldbesentto ElizabethF.Loftus,PsychologyDepartment,Universityof Washington, Seattle, Washington 98195. 86 Palmer,inpress).Thewordingof aquestionissuchan importantmatter,thatarecentbookintendedtohelp potentialquestionnairedesigners(Oppenheim,1966) devotesanentirechaptertothetopicofquestion wording. Thepresentresearchdemonstratestheinfluenceof verysmallchangesinthewordingofaquestionina situationinwhichsubjectsviewedafilmofan automobileaccidentandthenansweredquestions about eventsthatdidanddidnotoccurinthefilm, For some ofthequestions,theEnglisharticlethe(thedefinite article)wasused,asin"Didyouseethebroken headlight?"Forotherquestionsthearticlea(the nondefiniteorindefinitearticle)wasusedresultingin questionslike"Didyouseeabrokenheadlight?" Previousresearchonthedefiniteandindefinitearticle hasbeenequivocalastowhetherthere isa differencein influencebetweenthetwo.Muscio(1915)concluded thatthemorereliableformof question wasone that did notusethedefinitearticle,whereasBurtt(1931) reported that a and the are about equally suggestive. Whatisthedifferencebetweentheanda,andwhy shoulduseof these articlesproducedifferential behavior onthepartof eyewitnesses?Onthistopic, anumberof psychologistshaverecentlyhadsomethingtosay(e.g., Anderson&Bower,1973;Brown,1973;Chafe,1972; Maratsos,1971;Osgood,1971);mosthavemadethe pointthatif a speaker hasalready seen a particular item, andassumeshislistenerisalsofamiliarwithit,hewill usethearticlethe.Forexample,whenayoungman wantstoborrowthefamilycar,hesays "Can I havethe familycartonight?"Thesenotionshavealreadybeen successfullyembodied into anelaborate modelof human memorycalledHAM(Anderson& Bower,1973); when HAMseesanounprecededbytheitlooksforthe referentofthatterminitsmemory.Incontrast,when HAMseesanounprecededby a itassumesthatanew memberofthenounclassisbeingintroducedintoits memory. Toreturntotheexampleof accidentsandtheora brokenheadlight,considerfirstthequestion,"Did you seeabrokenheadlight?"Twoquestionsareimplicitly beingaskedhere:(1) Wastherea broken headlight? and (2) if therewas,didyouseeit?a subject decides that theanswertoQuestion1is"yes,"hecanthenask himself Question 2, and heshould befairlycertain of his response.Theproblemthatarisesforasubjectisthat filmedaccidentsoccurinthe spaceof seconds making it nearlyimpossibletobecertainofQuestion1,and makingitlikelythatthesubjectwillrespond"don't know" much of thetime. Incontrast,thesecondquestion,"Didyouseethe brokenheadlight?"canbetranslatedintothenearly equivalent,"Therewasabrokenheadlight.Didyou happentoseeit?"Thus,asubjectwhoisinterrogated withthedefinitearticledgesnotneedtoanswer QuestionEffectively,theansweris"yes."Heneed onlyanswerQuestion 2,and,aswasthecasewiththe indefinitearticle,atthispointhecanbefairlycertain abouthisresponse.Accordingtothisanalysis,fewer "don'tknow"responseswouldbeexpected. Furthermore,ifasubject'srecollectionstendto conform,forsomereason,towhathebelievesactually didoccur,thenthedefinitearticlemay lead to a greater "recognition"of events,evenwhentheyneverinfact occurred. Method Onehundredgraduatestudentsparticipatedinthis experiment, in groups of various sizes. Allsubjects were told that theywereparticipatinginanexperimentonmemoryandthat theywouldbeshownashortfilmfollowedbyaquestionnaire. The content of the filmwasnotmentioned. Thefilmitself depicted a multiple car accident.Specifically, a carmakesarighthandturntoenterthemainstreamof traffic; thisturncausesthecarsintheoncomingtraffictostop suddenly,causingafivecarbumpertobumpercollision.The totaltimeof thefilmislessthan1 min,andtheaccidentitself occurs within a 4-secperiod. A ttheendof thefilm,thesubjectsreceivedaquestionnaire askingthemtofirst"giveanaccountof theaccidentyouhave justseen."Whentheyhadcompletedtheiraccounts,aseriesof specificquestionswasasked.Sixcriticalquestionswere embeddedinalisttotaling22questions.Halfthesubjects receivedcriticalquestionsintheform,"Didyouseea... ?" and theotherhalfofthesubjectsreceivedthemintheform,"Did EYEWITNESSTESTIMONY87 youseethe... ?"Threeofthecriticalquestionspertainedto itemspresentinthefilmandthree to items not present. Subjects wereurgedto report only what theysaw, and did soby checking "yes""no"or"Idon'tknow."Eachsubjectreceiveda pe;mutation of the questions. Results Table1presentsthepercentageof "yes,""no," and "Idon'tknow"responsesforboththe"the" and"a" subjects.Whetheranitemwasactuallypresentornot, subjectsinterrogatedwith a wereovertwiceas likely to respond"I don'tknow."Subjectsinterrogatedwith the tendedtocommitthemselvestoa"yes"or"no" response. Anotheraspectof thesedata areworthyof mention_ First,whenasubjectisqueriedaboutanitemthatwas notpresentinthefilm,"yes"responsesareparticularly interesting.A"yes"responseindicatedasubject reportedthathesawsomethingthatwasnot,infact, present.Usingtheindefinitearticleresultedinfalse "yes" responses 7% of the time. With the definite article, however,false"yes"responsesoccurred15%ofthe time-over twiceasoften. Toteststatisticallyforthedifferencebetween interrogationwith a and with the, a single scoreforeach subjectwasgenerated."Yes"responseswereassigneda valueof+1,"Idon'tknow"responseswereassigneda valueof 0,and"No'responseswere. assigneda valueof -1.Asubject'smeanscore,then,reflectedhis confidencethattheitemswerepresent.Thedifference betweenthe"confidencescores"fortheaandthe subjectswassignificantbyaMann-WhitneyU test, 2.98, P < .01. Totestforthedifferencebetweenitemswhichwere andwerenotpresent,twomeanscoresforeach subject weregenerated,oneforitemswhichwerepresentand theotherforitemswhichwerenot.Again,"yes" responsesreceivedavalueof1,"Idon'tknow" responsesavalueof 0,and"No"responsesavalueof -1.Asbefore,asubject'smeanscoreforaparticular typeofitemexpressedhisconfidencethatthoseitems werepresent.TwoWilcoxonmatched-pairssigned-ranks testsrevealedthatboththesubjectswhohadbeen interrogatedwithaandthesubjectswhohadbeen interrogatedwiththeweremoreconfidentaboutitems thathadbeenpresent,Z= 3.97,p.001, andZ= 2.52, p.01, respectively. Method Inorderto besure that the resultsobtained were not peculiar tothefilmorsubject popUlationused inExperiment I, a second experimentwasconducted.Experiment IIwasidenticalto Experiment Iinallrespectsexcept two.First, a different subject populationwasused;60people betweenthe agesof14and20 wererecruitedfroma public library. Second, a different filmwas used.Thefilmshowninasmallroomatthelibrary,depicted a minorcollisionbetweena man whowasbackingout of a narrow spaceinasupermarketparkinglot and a womanpedestrian who 88LOFTUSANDZANNI wascarryingalargebagof groceries.Thetotaltimeof the film waslessthan4min, andthe accident itself occurred with a 2-sec period. Subjectsviewedthefilmandthenansweredaquestionnaire. Sixcriticalquestionswereembeddedinalistof 22; 3 pertained toitemspresentinthefilmand3toitemsnotpresent.The definitearticlewasused for30subjects; the indefinite article for the other 30. Results Table1presentsthepercentageof "yes,""no,"and "Idon't know"responsesforboththe"the" .and subjects.TheeffectsofExperiment 1arereplicated. Whenanindefinitearticlewascontainedinaquestion aboutanitemthatwasnotpresentinthefilm,"yes" responsesoccurred6%ofthetime.Whenthedefinite articlewasused,"yes';responsesoccurred20%of the time."Idon't know"responsesoccurred,overall,more oftenwhentheindefinitearticlewasused(47.5%vs. 15.5% forthe definite article). DISCUSSION Theresearchreportedhererequiredsubjectstoviewa filmof atrafficaccident andthen to answerquestions aboutthe film.A majorfindingwas thatquestionscontainingan indefinite article ledtomanymore"Idon'tknow"responses.Thephrase"a brokenheadlight"couldrefertoanyof anumberof headlights whichasubjectmight have seen.Since it is impossible to inspect allof the headlights carefully in thetimeallowedforviewingthe accident,itisimpossibletobesurethatnoheadlightwas broken. In thiscase,then,the subjectmust deal withuncertainty aboutwhetherabrokenheadlightactuallyexistedatall,anda largernumberof "uncertain"or"don'tknow"responsesresult. Ontheotherhand,"Didyouseethe broken headlight?"more stronglyimpliesthe existence of a specific broken headlight, and thesubjectneednotdealwiththeuncertaintyaboutwhether the brokenheadlightexisted.Thisfindingmay alsoindicate that itiseasiertobeconfidentthatyouhavenot seen some specific itemthantobe confidentthat youhavenot seen any instance of a generalclass of items. Asecondresultisthatquestionscontainingadefinitearticle resultedinagreaternumber of falserecognitions ("recognition" of eventsthat hadneveroccurred). At leasttwoexplanations for thisfindingarepossible.One isthat the definite articleproduces abiasfavoringa"yes"or"no"response;inotherwords,the changesasubject'scriteriaforhowmuchobjectiveevidencehe needstosay"yes"or "no." The other isthat the definite article leadsasubjecttoinferthattheobjectwasinfactpresent, causingforsomeareconstructionintheiroriginalmemoryfor theevent.Whilethepresentdatacannotdifferentiatebetween responsebiasandreconstructivememoryexplanations,arecent study(Loftus&Palmer,inpress)doesnotindicatethat questionsaskedsubsequenttoaneventcancausea reconstructioninone'smemoryofthatevent.Inthatstudy subjectsviewedfilmsof automobile accidents andthen answered questionsabouteventsoccurringinthefilms.Thequestion "Abouthowfastwerethecarsgoingwhentheysmashedinto eachother?"elicitedahigherestimateofspeedthan"About howfastwerethecarsgoingwhentheyhiteachother?" Furthermore,onaretest1weeklater,thosesubjectswho receivedtheverbsmashed weremorelikelytosay"yes"to the question"Didyouseeanybrokenglass?"eventhoughbroken glassdidnotexist inthe accident. Theimplicationof theseresultsforcourtroomexaminations, policeinterrogations,andaccidentinvestigationsisfairlyclear Table1 Percentageof"Yes,""No,"and"Idon'tknow"Responses toItemsthatWerePresentandNotPresentintheFilm PresentNot Present Response"a" Experiment I Yes1720157 No60297255 I don't know23511338 Experiment II Yes1815206 No62286956 I don't know20571138 cut.Themainaimofinterrogationsconductedbyattorneys beforethecourt,forexarnple,istoprovideinformationabout eventswhichhaveactuallytakenplace.Differentformsof questionscanbeconsciouslyusedtoelicit desired answers from awitness, andalsotocreate a desiredinfluence upon the jury. In thepresentresearch,theindefinitearticleelicitedmorefalse responses.Questionswhicheitherbyformor content suggestto thewitnesstheanswerdesiredor"lead"himtothatdesired answerarecalled"leadingquestions'inthe courtroom, andthe existenceofrulesforexcludingthem(e.g.,SupremeCourt Reporter,1973)isad.;finiterecognitionoftheirpowerof suggestion.Whileanattorney canseemingly easily"sense" when toobjecttoaleadingquestionaskedby anotherattorney,the definitionofleadingisalongwayfrombeingprecise.Any completedefinitionmusteventuallyconsiderthesubtle suggestibility that individual words can carry with them. REFERENCES Anderson,J.R.,Bower,G.H.Humanassociativememory. Washington,D.C:V. H. WinstonSons,1973. Barber,T.X.Hypnosis:ascientificapproach.NewYork:Van Nostrand Reinhold,1969. Brown,R.Afirstlanguage:theearlystages.Cambridge,Mass: HarvardUniversityPress,1973. Burtt,Legal Psychology,1931. Cady,H.M.Onthepsychologyof testimony.American Journal of Psychology,1924, 35, 110-112. Chafe,W.L.Discoursestructureand human knowledge. In J.B. CarrollR.R.Freedle(Eds.),Languagecomprehensionand theacquisitionofknowledge.Washington,D.C:V.H. WinstonSons,1972. Gardner,D.S.The perception and memoryof witnesses.Cornell LawQuarterly,1933, 8, 391-409. Loftus,E.F.&Palmer,J.C.Reconstructionofautomobile destruction:anexampleoftheinteractionbetween language and memory. Journal of Verbal Learning and VerbalBehavior, Maratsos,M.P.Theuseofdefiniteandindefinitereferencein youngchildren.PhDdissertation,HarvardUniversity,1971. Marquis,K.H.,Marshall,J.,Oskamp,S.Testimonyvalidity as a functionofquestionform,atmosphere,anditemdifficulty. Journal of Applied Social Psychology,1972,2, 167-186. Marston,W.M.Studiesintestimony.JournalofCriminalLaw and Criminology, 1924, 15, 5-31. Muscio,B.The influenceof the formof question.British Journal of Psychology, 1915, 8, 351-389. Oppenheim,A.N.Questionnairedesignandattitude measurement.NewYork:BasicBooks, Inc.,1966. Osgood,C.E.Wheredosentencescomefrom?InD.D. SteinbergL.A.Jakobovits(Eds.),Semantics:An interdisciplinaryreaderinphilosophy,linguistics,and psychology.Cambridge,England:Cambridge UniversityPress, 1971. SupremeCourtReporter,1973,3:Rulesof Evidence forUnited States Courts andMagistrates. Whipple,G.M.Theobserverasreporter:Asurveyofthe psychologyoftestimony.PsychologicalBulletin,1909,6, 153-170. (Receivedfor publication October 71, 1974.)