Journal of Applied Psychology1978, Vol. 63, No. 3, 352-359Individual and Situational Factors in Eyewitness TestimonyBrian R. Clifford and Jane ScottNorth East London Polytechnic, London, EnglandA four-way split-plot design was used to investigate the effect on eyewitnessaccuracy of the nature of the witnessed incident (violent, nonviolent), mode ofinitial questioning (narrative, interrogative), sex of witness, and type of information probed (actions, descriptions). Groups of six male or six female nonpsychology undergraduates were each randomly assigned to one of the resultingeight conditions. Leading questions, personality and accuracy, and confidenceratings were also examined. Accuracy was poorer under the violent condition,and females performed more poorly than males in this condition. Actions werebetter recalled than descriptions. Although type of initial questioning had noeffect on later accuracy, subjects were misled by leading questions. No relationship was found between either personality and accuracy or confidence in correctness and objective accuracy. Implications for police procedure are indicated.Although our awareness of the factors thatcan potentially influence eyewitness testimony is fairly comprehensive (see Buckhout,1974, 1977; Clifford & Bull, 1978), our knowledge of how these factors interact is extremelyscant. Unless and until such interactive effectsare studied, the information that law enforcement agencies receive from psychologists willbe essentially incomplete.As an example of incomplete information,one question for which there is at the momentno answer is whether all types of criminal incident are recalled with equal accuracy. Itcould be argued that crimes range along a continuum of emotionality or arousal and that therecall abilities of witnesses to such differentincidents may closely parallel that continuum.Support for this speculation comes from asurvey of 100 cases on police files by Kuehn(1974), in which he found that the type ofcrime was a significant factor in completenessof report. Victims of robberies provided a significantly fuller report of their assailant thandid rape or assault victims. Although a number of possible explanations of this findingRequests for reprints should be sent to Brian R.Clifford, Department of Psychology, North EastLondon Polytechnic, Stratford, London E IS 2RP,England.exist, it could be hypothesized that completeness of report decreased as a function of theincreasing emotionality of the crime. Indeed,in the field of perception, Scott (Note 1) andLaughery (Note 2) have clearly indicated theadverse effects that high levels of arousal andstress can have on perception. The testing ofthis hypothesis—that accuracy of recall decreases with increasing emotionality of thewitnessed event-—is the main focus of thepresent experiment.The type of crime witnessed may also interact with the type of witness. Clues as towhose testimony can be trusted would be ofgreat benefit to police, but at the moment psychologists are providing conflicting evidence.A number of lines of evidence suggested thatfemales ought to be better witnesses thanmales. First, females have a greater opportunity to learn to encode facial features thanmales because of their greater exposure tomagazines and cosmetic literature (Cross,Cross, & Daly, 1971; Ellis, Shepherd, &Bruce, 1973). Second, females are reported tobe more socially attentive than males (Exline,1963; Witryol & Kaess, 1957), more sociallymotivated (Smith, 1966), and more field dependent, which predisposes them to betterCopyright 1978 by the American Psychological Association, Inc. 0021-9010/78/6303-0352$00.753S2FACTORS IN EYEWITNESS TESTIMONY 353process and retain socially significant stimuli,of which faces form a special subclass (Messick & Damarin, 1964; Witkin, Dyk, Faterson, Goodenough, & Karp, 1974). Third,there is some evidence from the laboratorythat females attend to faces earlier ontogenically than males (Fagan, 1972; Haaf &Brown, 1976).In the field situation, however, althoughpolice seem to prefer female witnesses (Levine& Tapp, 1973), Kuehn (1974) found thatfemale victims gave significantly less complete descriptions of assailants than did malevictims. In the laboratory, conflicting evidence also exists. Findings of no sex differencehave been obtained in recognition of schematic faces (McKelvie, 1976) and face photographs in an intentional learning paradigm(Howells, 1938) and in an incidental learningparadigm (Schill, 1966). On the other hand,however, Witryol and Kaess (1957) did finda significant female superiority with facephotographs, and Cross et al. (1971), Ellis etal. (1973), and Goldstein and Chance (1971)all found female superiority, at least with female photographs. Although face photographresearch is characterized by artificiality andproblems of extrapolation to situations outsidethe laboratory (Clifford, 1975; Clifford & Hollin, Note 3), artificiality cannot be consideredas the only factor, because the female superiority effect still appears in situations inwhich the initially witnessed person appearedlive (Bahrick, Bahrick, & Wittlinger, 1975) oron videotape (Lipton, 1977). The present experiment attempts to disentangle the complexity, and to investigate the generality, ofthe sex difference effect in witnesses by havingmales and females recall actions and descriptions contained in emotional and nonemotional videotaped incidents.Personality may be a second individualdifference factor with predictive power. Ifarousal interferes with memory retrieval inthe short term (Kleinsmith & Kaplan, 1963,1964; Walker, 1958) and if introverts aremore chronically aroused than extraverts (H.J. Eysenck, 1967), then extraverts should bebetter witnesses than introverts, at least initially, and especially when the witnessedevent is arousing. Given that no publishedwork exists on this specific relationship, aminor aim of this study was to test the aboveprediction.A third area in which less than clear evidence has been provided is whether eyewitness testimony is poor overall or only in certain respects. A study carried out by Tichnerand Poulton (1975) in which they asked observers to detect crimes and their perpetratorsin a videotaped street scene found that perception was better for actions than for people.The applied point here, if it can be replicatedin a memory experiment, is that witnessesshould give more accurate action testimonythan description testimony, and this wouldcarry implications both for the efficient allocation of police resources in furthering inquiries and for identification as a prosecutiontool.The elicitors of testimony themselves havebeen implicated in the generation of faulty recall as a result of their overall questioningprocedure and the way they ask specific questions. Early research suggested that free report produced more accurate testimony thaninterrogative report (Gardner, 1933; Marston,1924), and free report followed by interrogative report was more beneficial than the opposite sequence (Cady, 1924; Snee & Lush,1941). However, later research (Lipton,1977; Marquis, Marshall, & Oskamp, 1972)has cast doubt on these statements as generalrules. Given that this whole area has important implications for the initial stages of policeprocedure, the present study looks at the effectof both narrative and interrogative report onlater, structured testimony, and as an adjunctto this, the effect of leading questions, whichhas assumed special importance of late in eyewitness testimony research (see Loftus, 1975;Miller & Loftus, 1976).To summarize then, this study set out toinvestigate the effect on accuracy of testimonyof different types of witnessed events, the sexof the witness, different modes of initial questioning, and the probing for recall of physicaldescriptions and physical actions. Additionalinterest lay in the relationship of personalityto accuracy and the susceptibility of subjectsto leading questions.354 BRIAN R. CLIFFORD AND JANE SCOTTMethodSubjectsForty-eight subjects, 24 male and 24 females, tookpart in this experiment. All were undergraduate volunteers, with an age range of 18 to 30 years, whowere from numerous nonpsychological disciplines,having been drawn at random from the North EastLondon Polytechnic, London, England, male andfemale subject pools.Measurement of VariablesThe main dependent variable was accuracy oftestimony computed by allocating 1 point for everycorrect answer given in a final questionnaire, whichprobed memory for both physical actions and physical descriptions, with a maximum of 40 points (20 forphysical descriptions and 20 for physical actions).Testimony accuracy was also computed for initialreports (narrative or interrogative) by dividing thetotal number of correct items mentioned by the totalnumber of items mentioned. Additional interest layin the relationship of accuracy of testimony to (a)the witnesses’ certainty of correctness (the latter assessed by confidence ratings given by each witness foreach answer on a S-point scale: 1 = perfectly sure, 5= not at all sure) and (b) the person’s standing onthe introversion-extraversion dimension, assessed bythe Eysenck Personality Inventory (EPI). The lasttwo sources of interest were how susceptible witnesseswere to four leading questions embedded in the finalquestionnaire and whether male and female witnessesrated the violent film differently on a 2-point scale(1 — slightly violent, 2 = very violent). (A normativestudy containing 30 subjects indicated that a thirdcategory “not at all violent” was redundant, sinceno one had used it.) In all cases tests of difference orrelationship in the dependent variables were performed with the rejection region for the statisticaltests set at p < .05.VideotapesTwo black-and-white videotapes were made: oneof a violent incident and one of a nonviolent butsimilar incident. In both cases the same three semiprofessional actors and scenario appeared in eachfilm. The scenario basically involved two policemensearching for a criminal (who was never seen) andeventually finding him with the reluctant help of athird person. In the nonviolent film this reluctancegave rise to a verbal exchange between the threepeople and a number of weak restraining movementsby one of the two policemen. In the violent filmthis reluctance eventuated in physical assault on thethird person by one of the two policemen, with fourblows being exchanged (three by the policemen, oneby the third man). The weak restraining movementsand the exchange of blows defined the nonviolent andthe violent films, respectively. These critical sequenceswere identically placed in the middle of their respective sequences, thus rendering the beginning andend of the two films identical. The violent filmlasted for 1 min 4 sec, and the nonviolent film lastedfor 1 min 3 sec. The verbal content was identical forthe two films.ProcedureA 2 X 2 X 2 X 2 split-plot factorial design (Kirk,1968) was used. The factors were type of film (violent, nonviolent), type of initial questioning (narrative, interrogative), sex of witness, and type of probeditem (physical descriptions, physical actions). Thefirst three variables were between-subject factors, thelast was a within-subject factor. Four groups of sixmales and four groups of six females were randomlyassigned, with the constraint of sex, to the resultingeight conditions.The procedure consisted of five parts. In Part 1,half of the subjects watched the violent film, halfwatched the nonviolent film. In Part 2, all subjectswere given the first half of the EPI to complete. InPart 3, half of the subjects in each condition (violent and nonviolent film) gave a narrative, free report of the film’s content and were asked to stressactions and descriptions. The other half of the subjects answered IS questions (interrogative report)concerning actions and descriptions. This phase ofthe experiment lasted 2 minutes. In Part 4, all subjects completed the second half of the EPI. In Part S,all subjects answered a 44-item questionnaire. Fifteenminutes were allowed for this part of the experiment.Subjects in the violent film condition were then required to indicate on a 2-point scale how violent theyconsidered the film to be.The narrative and interrogative reports were balanced for content as far as possible by using asquestions in the interrogative questionnaire thoseaspects that were recalled spontaneously in narrativereport by nonexperimental subjects in an extensiveseries of pilot studies. This was done to equate thetwo groups on rehearsal, because interest lay ininitial recall mainly as an independent variable ratherthan as a dependent variable.ResultsTable 1 summarizes the mean accuracyscores obtained from the main questionnaire.A four-way split-plot analysis of variance wasused to analyze these data. The factors weretype of film (violent, nonviolent), type of initial report elicited (narrative, interrogative),sex of witness, and type of informationprobed (actions, descriptions). The completeanalysis of variance can be found in Table 2.FACTORS IN EYEWITNESS TESTIMONY 355Two main effects were highly significant,type of film and type of probed information.Recall was much more accurate for nonviolentfilm, detail than for violent film detail, andactions were much better recalled than descriptions. The Type of Film X Sex of Witness interaction was also significant. Furtheranalyses of this interaction (simple main effects analysis) revealed that although bothmales, ^(1,40) = 21.52, p< .001, and females, F(l, 40) = 50.36, p < .001, differed intheir accuracy of recall in the violent and nonviolent conditions, with recall being poorer inthe violent conditions males and femalesdiffered significantly only in the violentconditions, with females performing worsethan males, F(l, 40) = 7.08, p < .025.The only other first-order interaction toreach significance was the Type of Film XType of Information Probed interaction. Further analysis revealed that although recall ofactions and descriptions differed significantlyin both violent, F(l, 40) = 14.43, p < .001,and nonviolent film conditions, F(l, 40) =54.94, p < .001, the violent film conditionhad a much greater effect on recall of actions,Table 1Mean Recall Accuracy of Actions andDescriptions in Violent and NonviolentIncidents by Male and Female WitnessesTable 2Analysis of Variance of Accuracy Scoresin the Final QuestionnairePhysicalType ofreportDescriptionsM SDActionsM SDViolentNarrativeMalesFemalesInterrogativeMalesFemales7.05.86.66.51.11.2.81.57.47.08.17.61.01.01.11.1NonviolentNarrativeMalesFemalesInterrogativeMalesFemales8.58.27.78.31.31.41.5.89.910.59.810.51.4.81.11.0SourceBetween subjectsType of film (A)Type of initialquestioning (C)Sex of witness (D)AX CAX DC X DA X C X DErrorWithin subjectsType of probedinformation (B)A XBB X CB X DA X B X CA X B X DB X C X DA X B X C X DErrordf471 11111140481111111140MS F111.80 69.01*.18
