IN SUPPORT OF A DIVERSITY OF METHODS: EYEWITNESS MEMORYIN ACTUAL CASES OF ROBBERY AND FRAUDbyPatricia Ann TollestrupB.Sc., The University of Lethbridge, 1988A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THEREQUIREMENTS FOR THE DEGREE OF MASTER OF ARTSinTHE FACULTY OF GRADUATE STUDIESDepartment of PsychologyWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAAugust 1994© Patricia Ann Tollestrup, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives, It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)__________________________Department of Ii)crik iay-”The University of British ColumbiaVancouver, CanadaDate 44 tf99DE-6 (2/88)ABSTRACTMuch of our knowledge of eyewitness memory is built on laboratory-based studies. In recent years this has caused a controversy, the resolution ofwhich seems to be an acceptance of the need to learn about eyewitness memoryfrom a variety of methods such as field studies, case, and archival research.The thesis reports an archival analysis of actual Royal Canadian MountedPolice case files of robbery and fraud. Previous research on the configurationof robberies was replicated. The majority of robberies are victim-only crimesthat do not involve bystander eyewitnesses. Robbery victims and witnessesprovided more details (10.96 and 9.37 respectively) regarding the perpetrator’sappearance than fraud victims (2.11). Less than 10% of robbery victims andwitnesses, but almost 75% of fraud victims were unable to describe theperpetrator. All eyewitnesses tended to overestimate age and underestimateheight and weight. Identification outcomes were analyzed according to theevidence category of the perpetrator (Confession, Implicating and None). Thelikeithood that the police suspect was indeed the actual perpetrator is assumedto be highest in the Confession condition and lowest in the None condition. Inthe Confession condition, the police suspect was selected by 84.6% of robberyvictims, 55.55 of robbery witnesses and by 22.7% of fraud victims.Identification accuracy was adversely influence by the average delay betweenthe crime and attempting an identification. Delay was a confound in theseanalyses as on average, victims of robbery attempted their identificationssooner than both robbery witnesses and fraud victims. Limited support was11found for the weapon focus phenomenon. Weapon presence (in robbery casesonly) did not influence descriptions (amount or accuracy). Weapon presencedid not significantly adversely influence identification accuracy, but was closeto obtaining significance (p = .061). The present analysis demonstrated thefeasibility as well as the utility of archival research.111TABLE OF CONTENTSAbstract iiList of Tables vIntroduction 1Method 55Results 63Discussion 83References 103Appendix A (ANOVA tables) 116ivList of TablesPageTable I Number of details in eyewitness descriptions 68Table 2 Accuracy of age, height and weight estimates 70Table3 Identification accuracy by evidence category 74Table 4 Identification accuracy and delay 76Table 5 Number of descriptive details and weapon presence 79VLike horse and cart, the study of eyewitness memory has been coupledwith and propelled by a debate about the external validity or thegeneralizability of the research (for an excellent historical review see Cutshall,1985). Since the 1970’s we have amassed a great deal of knowledge abouteyewitness memory, most of which has had an explicit agenda to assist thetriers of fact. There are many books (e.g. Loftus, 1979; Ross, Read, & Toglia,1994; Wells & Loftus, 1984; Yarmey, 1979), hundreds ofjournal articles andseveral meta-analyses (e.g. Deffenbacher, 1986; Shapiro & Penrod, 1986)dedicated to eyewitness memory. In addition there are several journals (e.g.Law and Human Behavior) that deal largely, but not exclusively, witheyewitness issues. Yet the bulk of our knowledge on eyewitness memory isderived from laboratory studies. In what has become the standard laboratorymethod, an event, usually depicting some kind of crime, is presented touniversity students via slides, video or staged live. Memory for the event istested by a variety of methods, such as questionnaires, interviews, and the useof photospreads or the occasional live lineup. It is uncommon for subjects inthese experiments to be involved in the event, although there are someexceptions (e.g. Hosch & Bothwell, 1990; Kassin, 1984; Read, Yuille &Tollestrup, 1992;) and ethical considerations rightly prevent subjects frombeing seriously victimized, aroused and/or engaged to the extent thateyewitnesses to violent crimes can be. Subjects are often debriefed about thestaged nature of an event immediately after its occurrence, thus eliminatingany consequences of their recall or recognition decisions for either themselves1or the “perpetrator.” A few studies have kept subjects blind about theirparticipation in an experiment on eyewitness memory right through theidentification task (e.g. Murray & Wells, 1982), but subjects in all laboratoryexperiments know they are taking part in some kind of psychologicalexperiment. Until recently, very few researchers had taken experimentalmethodology out of the laboratory to conduct field studies (e.g. Brigham,Maass, Snyder & Spaulding, 1982; Fun & Shepherd, 1986). In field researchan event is staged before subjects who are often drawn from a population otherthan university students. They generally learn about their participation in anexperiment only after they have attempted to identify the confederate. Likelaboratory studies the event is limited in the extent to which it can engage andarouse the subjects. Only a handful of case (Cutshall & YuilIe,1989; Yuille &Cutshall, 1986) and archival studies (Kuehn, 1974; Sporer, in press) of actualeyewitnesses have been conducted. Both case and archival research canaddress the influence of arousal on eyewitness memory, as well as point out thevariety of contexts in which eyewitnesses to crimes find themselves. Theysuffer from an almost complete lack of control and are costly and timeconsuming. At the outset, it is acknowledged that these categories of researchare fuzzy and that placement of a particular study into one of the categories isoccasionally less than straight forward.The dependence on laboratory studies has created a controversy withrespect to the generalizability of eyewitness findings. Many psychologists havetestified in court about the weaknesses of eyewitness accounts and this2testimony has depended on laboratory findings. It remains unknown whetherthe exponential increase in eyewitness memory research has been accompaniedby an increase in the prevalence of expert testimony but there is somesuggestive evidence. At least one prominent psychologist has written a bookabout her experiences as an expert witness for the defense (Loftus & Ketcham,1991) and there are a number of books (e.g. Blau, 1984; Brodsky, 1991;Kagehiro & Laufer, 1992) and journals (e.g. Expert Evidence) on the topic ofexpert testimony. Regardless, the debate about generalizability and whether ornot psychologists have a sufficiently proper and adequate data base abouteyewitness memory from which to draw their expert testimony was ignited byan exchange between McCloskey and Egeth (1983) and Loftus (1983). Sincethat time extreme positions have been advocated with some arguing that onlystrictly controlled laboratory studies can lead to generalizable findings (Banajiand Crowder, 1989) and others taking the position that only archival researchcan provide us with meaningful information (Koneem & Ebbesen, 1986).While those holding these more extreme viewpoints do offer some validcriticisms of other modes of research, much of the fire has been taken out ofthe debate by a growing recognition of a need for a multi-method approach tothe study of eyewitness memory (Christianson, Goodman & Loftus, 1992,Davies, 1989, 1990; Yuille & Wells, 1991). For example, Davies (1990) hasargued that “no one research method can of itself provide a reliable data basefor legislation or advocacy. Rather, problems need to be addressed from a3number of perspectives each of which makes a different compromise betweenecological validity and methodological rigor” “(p. iv).One of the most studied topics in the area of eyewitness memory and theone seemingly most testified about by expert witnesses in court, is eyewitnessidentification. There have been hundreds of studies conducted on this topic, alleither laboratory or field based. To the best of my knowledge there has beenonly one attempt at examining identification by actual eyewitness to crimes,quite some time ago (Borchard, 1932), the purpose of which was todemonstrate the existence of mistaken eyewitness identifications. To be fair,the task of studying actual eyewitness identifications is fraught withdifficulties, not the least of which is a lack of ground truth against which tocategorically determine the accuracy of eyewitness identifications.Nonetheless, it is time that a start be made to address the imbalance betweenlaboratory-based and archival knowledge about eyewitness identification. Inthe spirit of contributing to the multi-method approach, the present thesisinvolved an archival examination of case files from the Richmond, B.C.detachment of the Royal Canadian Mounted Police (RCMP). As such, thepurpose of this thesis was to determine how frequent eyewitnesses are asked tomake an identification, and to the extent that it is possible, examine theaccuracy of eyewitness identification and indicate factors that might influenceidentification performance. In addition, the thesis provides data on the amountof detail and accuracy of perpetrator descriptions provided by eyewitnesses. Itis hoped that data from this research and others like it will help to clarify issues4regarding the generalizability of laboratory and field research and in doing sosolidify and expand the data base on eyewitness memory.Just what is the current state of our knowledge on eyewitness memory?The following review of eyewitness research is organized according tomethodology with laboratory research presented first, followed by fieldresearch, with case studies and archival research concluding the review. Aprogression of quality is not intended in the presentation order of the researchapproaches. In the following pages, the major findings pertaining to eyewitnessidentification and descriptions as well as the strengths and weaknesses of eachof these areas will be reviewed. When a topic such as cross-racialidentification has been looked at by a variety of methods, conclusions regardingthe topic will be offered in the last section of the review in which the researchwas conductecLLaboratory ResearchThere are two distinct laboratory-based approaches to the study ofeyewitness memory. The standard method, described earlier involvespresenting an event to students and testing their memory by questionnaires,interviews, and the use of photospreads or the occasional live lineup. Mosttypically, subjects passively watch the event and are not physically oremotionally involved in or by it. Unless specified, the reader is to assume thestandard approach was employed.The second method, photo recognition, does not present events andinstead involves showing subjects slides or photos of faces. Often a number of5faces are serially presented and subjects must later recognize these among newfaces. Although much of this research was designed to learn about the neuralstructures and processes involved in face recognition and does not have amandate of forensic relevance, some research germane to eyewitness memoryhas been conducted with this paradigm. Most of the findings from the photorecognition research that have forensic relevance have been supported byresearch in other paradigms.Wells’ (1978) seminal paper on the distinction between system andestimator variables offered a classification system that allowed researchers toorient their research in ways that would be more immediately applicable to thecriminal justice system (CJS). Estimator variables refer to aspects of thecrime or its investigation over which the criminal justice system has no controland whose influence in any particular case must be estimated. Most estimatorvariables are associated with the crime itself such as the characteristics of theeyewitnesses and perpetrators, the seriousness of the crime, and illumination..Estimator variable research is aimed at finding factors that can help the CJSincrease in some contexts or reduce in others its reliance on an eyewitness’stestimony. In contrast, a system variable is one over which the CJS can exertsome control. Examples include lineup presentation and construction, and thetype of questions police ask eyewitnesses. The goal of system variable researchis generally to improve eyewitness performance. Wells suggested that researchinto system variables would offer the most immediately beneficial findings tothe CJS and encouraged research into this area. The distinction between6estimator and system variables is used to organize the laboratory and fieldresearch on eyewitness memory.Estimator VariablesIndividual DifferencesSeveral individual difference variables have been examined in relation toeyewitness memory. In comparison to older children and adults, youngchildren are less accurate in answering objective questions and they offershorter free recall accounts (Leippe, Romanczyk & Manion, 1991). There is awhole separate area that deals with the eyewitness memory of children, most ofwhich is focused on children as victims of sexual abuse. The present thesis isconcerned only with adult eyewitness memory. Yarmey, Jones and Rashid(1984) compared the recall and identification performance of young and oldadults and found that young adults are more accurate in answering specificquestions and they offer more complete, but less accurate free recalls than olderadults. Young adults are less likely to make false alarms on a target-absentlineup than older adults.The degree of self-monitoring has been related to identification accuracy(Hosch & Cooper, 1982; Hosch & Platz, 1984; Zimmerman, 1982). High selfmonitors are concerned about behaving “correctly” or appropriately in socialsituations than low self-monitors. High self-monitors tend to be more accurateeyewitnesses. Self-monitoring scores have also been related to susceptibility tobias in lineup presentation (Hosch, Leippe, Marchione & Cooper, 1984). Four7of the five studies on the relation between scores on the Benton FacialRecognition Test (BFRT) and identification accuracy summarized by Hosch(1994) report a strong and positive correlation. Two studies have examinedthe relationship between observational ability and eyewitness memory.General observational accuracy was positively related to eyewitness recall of anevent (Boice, Hanley, Shaughnessy, & Gansler, 1982). Identification accuracywas positively related to memory for the event in one study (Boice et a!, 1982)and was negatively related to memory for peripheral details of the room inwhich the event occurred in another (Wells & Leippe, 1981).In Shapiro and Penrods (1986) meta-analysis of factors affecting facialrecognition, individual difference variables demonstrated a small butsignificant effect on both hits and false alarms. The result of the metaanalysis and the earlier reported findings suggest that there may be some paperand pencil tests that the police could give to an eyewitness after a crime to helpthem determine the general ability of the eyewitness to identify faces and recallthe details of the crime. However, none of these methods have been used onactual eyewitnesses, and only the relation between self-monitoring andidentification accuracy has been tested on subjects in a field study. This aside,there are practical issues regarding the hypothetical use of these tests by policethat have been left unaddressed by the psychologists proposing them. Policedepartments may not have the time, space and/or money to administer andscore the tests. Moreover, they may not want to routinely administer tests thatcould potentially end up assisting in the defense of a suspect. A closer liaison8with the potential consumers of research on eyewitness memory, policeofficers, lawyers, judges, etc., could lead to more usable research as well asmake it easier for these consumers to accept the findings by feeling that theyplayed a role in their discovery.Situation VariablesWith the exception of arousal, discussed in a separate section, there hasbeen little research into situation variables (aspects of the crime itself or theparticipants) that could influence eyewitness memory. Clifford and Holim(1981) found that subjects were able to accurately identify the main perpetratorif there was one or three perpetrators, but performed at chance level if therewere five. Poor illumination did not affect recall accuracy, but did adverselyaffect identification accuracy (Yarmey, 1986). The appearance of theperpetrator has been shown to influence identification accuracy. Guilty orunpleasant looking perpetrators are more likely to be correctly identified thanperpetrators who were rnnocent or pleasant looking (Yarmey et al., 1984). Thisfinding is reminiscent of the typicality effect obtained from photo recognitionstudies (Courtois & Mueller, 1981; Vokey & Read, 1992) in which atypicalfaces are easier to recognize than typical faces. Finally, MeKelvie (1988) useda photo recognition design to demonstrate that recognition of bespectacledfaces was poorer than faces without glasses primarily due to higher false alarmson bespectacled faces.The effect of race (subject and target) has been extensively examined,mainly with Caucasian and Black faces, and it has generally been concluded9that people are better at recognizing faces from their own race than from others(e.g. Ayuk, 1990; Brigham & Barkowitz, 1978).Finally, there have been some investigations into the effects of alcohol(Read, et al. 1992; Yuille & Tollestrup, 1990) and marijuana (Yuille,Tollestrup, Marxsen & Porter, under review) on eyewitness memory. Thesestudies were conducted after archival research (Yuille, 1986) indicated thateyewitnesses to crimes were occasionally under the influence of intoxicants.Subjects in Yuille et a! witnessed a live event and were interviewed twice;immediately after the event and a week later. Marijuana did not affectaccuracy of recall, nor did it have a significant effect on identificationaccuracy. However, the study lacked adequate power to detect any differencesin identification accuracy or choosing rates. On the target-absent photospreadmarijuana subjects were twice as likely than placebo subjects to chose thephoto of an innocent foil. The effects of marijuana on the amount recalledappeared to be temporary as subjects who had consumed the drug recalled lessthan those who had consumed a placebo, but only on the immediate interview.By the second interview, when presumably the effects had worn off, placeboand marijuana subjects performed equivalently.In contrast to marijuana, the effects of alcohol were not temporary. Subjects inthe Yuille and Tollestrup (1990) study consumed alcohol or a placebo andwatched a staged event. Some subjects were interviewed immediately and allsubjects were interviewed a week later. Alcohol suppressed the amountrecalled during the immediate interview and both the amount and accuracy of10recall after a 1-week delay. Like marijuana, alcohol did not affectidentification accuracy except on target-absent lineups where alcohol subjectsmade numerically more selections of innocent foils than placebo subjects.Finally, the study demonstrated the advantage of an immediate interview;subjects who were interviewed twice recalled much more after a week delaythan those for whom the 1-week interview was their first.The second alcohol study also examined the effects of arousal (Read eta!. 1992). It employed a mock theft paradigm in which subjects committed asimulated theft. Arousal was manipulated by varying subject’s perceptions ofthe likeithood and consequences of getting caught in the act of thievery. Aconfederate interrupted the theft which allowed for an examination of theeffects of alcohol and arousal on identification accuracy (in the second studyonly). Subjects were interviewed only once, a week after the simulated theft.In the first study reported in Read et al. (1992), alcohol had an adverse effecton the amount and accuracy of recall, with the greatest impairment being in therecollection of the appearance of the confederate intruder. In the second study,subjects consumed less alcohol and recalled fewer details about the appearanceof one of two targets. Alcohol also impaired the accurate recall of the sequenceof actions during the theft, something which was not analyzed in the first study.With target-present photospreads, alcohol had an adverse effect onidentification accuracy only if arousal was low. This suggests that higherarousal enabled subjects to overcome the adverse effects of alcohol with11respect to identification and that perhaps actual eyewitnesses who are similarlyintoxicated and aroused might also be able to identify the perpetrator.These last studies of the effects of drug use were undetaken becausearchival research had revealed drug use played a role in actual crimes. As suchthey illustrate how archival research can provide fruitful avenues for researchon situation variables. Each different type of crime such as fraud and assault islikely to have a different complex of situation variables such as number ofperpetrators, familiarity with the perpetrator, and the number of non-victimeyewitnesses. Archival research can also aid in determining how much of thelaboratory context overlaps with that of actual eyewitnesses.ArousalThis topic has been examined from a variety of angles. In early efforts,some studies manipulated arousal by means such as electric shock or whitenoise that were external to the to-be-remembered material (Brigham, Maass,Martinez & Whittenberger, 1983; Deffenbacher, 1983). These studies reportdetrimental effects of arousal. Most would agree that this method does notprovide a close enough approximation to the situation faced by actualeyewitnesses and have found other ways to approach the topic. Currently,arousal is typically operationalized by the use of violent or non-violent materialor by looking at differences between victim and non-victim subjects.Additionally, there are some who employ the mock theft paradigm describedearlier in which subjects commit a simulated theft and arousal is manipulatedby varying the threat of apprehension. Further lines of research have pursued12the qualities of memories for arousing events and the weapon focusphenomenon.Event Violence. The outcomes of research on the effects of arousal,operationalized as event violence, on eyewitness memory includes the wholegamut of possible findings, from detrimental through to facilitative. Tworeviews of this literature, almost a decade apart, drew very differentconclusions regarding the discrepant findings and the effects of arousal oneyewitness memory (Deffenbacher, 1983; Christianson, 1992).Deffenbacher (1983) reconciled the different findings by fitting them tothe oft-cited Yerkes-Dodson inverted u-shape curve which demonstrates thatperformance improves with low to moderate levels of arousal, but is impairedunder high levels of arousal (Yerkes & Dodson, 1908). Deffenbacher inferredthat the studies which showed either a facilitative or no influence of arousalhad induced lower levels of arousal than the studies which showed adetrimental influence. Like many other researchers, (e.g. Loftus, 1979;Yarmey, 1979) Deffenbacher assumed a relatively simple relationship betweenarousal and eyewitness memory and cautioned that the level of arousal faced byactual eyewitnesses to violent crimes would be very high, thus their memoryfor the event would be poor.With many more studies under our collective belt, Christianson (1992)concluded in his review that the lack of consensus in the field is more apparentthan real and reflects the different foci of various researchers. For example,some focus on eyewitness accuracy and persistence over time (Reisberg et al,131988), and others (Wagennar & Groeneweg, 1990) focus on errors and thedecline in memory over time (for a more detailed analysis of these two foci,see Cutshall, 1985). In terms of the effects of arousal on eyewitness memory,he concluded that there “are no real grounds for a simple relationship betweenintense emotion and memory” (p. 302). Two variables, retention interval (e.g.Christianson, 1984; Heuer & Reisberg, 1992) and type of detail (e.g. Burke,Heuer & Reisberg, 1992; Christianson & Loftus, 1987, 1991) interact witharousal to paint a complex picture that is more supportive of Easterbrook’s(1959) theory of arousal and memory than that of Yerkes and Dodson (1908).Essentially, Easterbrook’s theory predicts that with high arousal, more attentionis focused on the source of the arousal. This leads to an accurate, detailed andpersistent memory for the material upon which attention was focused, alsoreferred to as the central details. Information outside of this focus of attention,referred to as peripheral detail, is remembered in less detail, is less accurateand less persistent.Eyewitness Role. Several studies have examined differences betweenvictims and non-victims (‘witness’ in this discussion), all of which assume thatthe role ofvictim is more arousing that the role of witness and all involved alive staged theft. In terms of identification accuracy, three studies show nodifference between victims and witnesses (Hosch & Bothwell, 1990; Hosch &Cooper, 1982; and Hosch, et al 1984) and one showed that victims wereinferior to witnesses (Kassin, 1984). The effects of eyewitness role onproviding description are less clear with two studies showing no difference14(Hosch & Cooper, 1982 and Kassin, 1984) and one showing superiordescriptions provided by victims (Hosch & Bothwell, 1990). These studiesdiffer on many dimensions which makes it difficult to draw any firmconclusions about laboratory-based differences between victims and witnessesor the effects of arousal.Mock theft paradigm. In study 1 of Read, et a!. (1992), there was noeffect of arousal on either the amount or accuracy of recall and identificationaccuracy was not examined. In the second study, the event was changedslightly to increase the amount of arousal experienced by all subjects, butparticularly those in the high arousal condition. With the modified event, higharousal subjects recalled more correct information than low arousal subjects.Subjects were asked to identify two targets, the confederate intruder and abystander whom subjects encountered before receiving instructions on how tocommit the crime. The intruder was considered to be central to the event andthe bystander was considered to be peripheral. Identification accuracy of thesetwo targets was the same for low arousal subjects, but high arousal subjectsaccurately identified the intruder more often than the bystander, suggestivelysupportive evidence for Easterbrook’s theory of the effects of emotion onmemory.Oualities of memory of arousing events. Christianson and Loftus (1990)have surveyed students about their memories of traumatic events. They reportthat these memories are quite vivid and the more intense the emotion of theevent, the more central, but not peripheral details are recalled. In addition,15most people with memories of traumatic events think about and frequentlydiscuss the event with others. There is often a specific detail that isremembered best. This well-remembered detail is equally likely to bedescribed as central or peripheral to the event. Most describe this well-remembered detail as being concrete (e.g. “an average leather jacket”), butover one-third recalled a thought or a feeling (e.g. “my heart felt like it wassinking”).Weapon Focus. According to the notion of weapon focus, eyewitnessespay more attention to a weapon than the face of the person holding the weapon,which results in poor memoiy for the face. The laboratory studies that haveexamined this topic have found support for weapon focus in terms of thegreater attention paid to the weapon (Loftus, Loftus & Messo, 1987) and thepredicted effects on both recognition and recall (Kramer, Buckhout & Eugenio,1990; Maass & Kohnken, 1989; Tooley, Brigham, Maass & Bothwell, 1987).Although the absolute effects of arousal on eyewitness memory are notyet entirely clear, there is substantial evidence to support the notion thatattentional focus is narrowed under arousal. The weapon focus phenomenon isperhaps the clearest example of this narrowing of attention to the source ofarousal. Yuille and Tollestrup (1992) speculated that attention need not befocused on the source of arousal. For example in cases of extreme arousalpeople may direct their attention inward, largely in self-protection, and wouldstore little of the event. They may end up with amnesia, or like some subjectsin Christianson and Loftus (1990) they may remember their emotional state16most clearly. The present study was able to address attentional narrowing onlyin the form of weapon focus; the eyewitness statements were not as completeas those obtained from laboratory subjects. For example, statements providedby actual eyewitnesses generally did not contain the “gist” or action details ofthe crime, which are considered central details in lab research. Even if theyhad, it would have been extremely difficult, if not impossible to determine theaccuracy of those details.The potential difference in arousal experienced by the typical laboratoryeyewitness and various eyewitnesses to actual crimes has been one of the mostpersistent issues in the generalizability debate (e.g. Yuille & Tollestrup, 1992).Archival research (Yuille, 1986; Yuille & Cutshall, 1986) has demonstratedthat there are several discrepancies between the role of the typical laboratoryeyewitness and an actual eyewitness to a violent crime (Yuille & Tollestrup,1992). The usual role of the laboratory eyewitness is that of an uninvolvedobserver to a relatively non-arousing event. In actual forensic contexts,eyewitnesses are cast in a multitude of roles that exhibit varying degrees ofsimilarity to the typical laboratory eyewitness. For example, violent crimessuch as robbery and assault (sexual and non-sexual) typically involve only thevictim. So, the modal eyewitness to a violent crime is a direct participant in anarousing and threatening event. Not all eyewitnesses are involved in arousingevents. Victims of fraud and witnesses to minor traffic accidents, for example,probably have levels of arousal that are similar to the typical laboratoryeyewitness. Low levels of arousal are also experienced by secondary17eyewitnesses, those who did not witness a crime but who might have interactedwith a suspect and can tie him or her to the crime (for example, a store clerkwho sold a unique piece of clothing left at the crime scene). The debate aboutgeneralizability could benefit from information about actual eyewitnesses thatdemonstrate various degrees of similarity to the typical eyewitness. Certainlya fair and valid test of the generalizability of laboratory based findings is to seeif they are replicated with actual eyewitnesses who were in similar contexts.The present study examined cases of robbery and fraud. It is assumed thatvictims of robbery bear the least similarity to the typical laboratory eyewitness,fraud victims bear the most similarity and witnesses of robbery are somewherein between. Additionally, the comparison between actual eyewitnesses torobbery and fraud will hopefully address some issues regarding the effects ofarousal.Accuracy of description and identificationFramed from a psychologist’s point of view, the issue is one concerningthe relation between the skills of recalling and describing the appearance of aperson from memory and the skill of recognizing that person in a photo lineup.Only a handful of studies have set out to examine this question directly.Wolfskiel and Brigham (1985) reported that subjects who gave accuratedescriptions were not more likely than those who gave poor descriptions toidentify one of two targets. In a photo recognition study Wells (1985) used 88target faces and found a small but significant correlation between descriptiveaccuracy and identification accuracy. However, the relation was due to the18fact that faces which were better described were also better identified. Otherstudies have reported incidental findings that support some relation betweenaccuracy of description and identification. Buckhout, Alper, Chem,Silverberg, and Slomovits (1974) reported that subjects who successfullyidentified a confederate thief were more accurate at estimating his weight thansubjects who did not successfully identify the confederate. Kassin (1984)reported a significant positive correlation between identification accuracy andscores on a police description form (r= .37) and Hosch and Bothwell (1990)reported a significant positive correlation between identification accuracy andfree recall description accuracy (r = .31). This topic has been examinedfurther in field studies.Confidence and accuracy of identificationIs an eyewitness’s stated confidence in her/his identification indicative ofthe accuracy of that decision? There have been two meta-analyses of this topic.First, Wells and Murray (1984) reported an average correlation betweenaccuracy and confidence of only .07. Bothwell, Deffenbacher and Brigham(1987) analyzed a sample of studies that were homogeneous in terms of thetask expected of the subjects, the measure of memory accuracy and the mannerof calculating the accuracy-confidence correlation. The Wells and Murrayanalysis did not meet these criteria. Bothwell et a! reported an estimatedcorrelation of .25 with a 95% confidence interval of .08 to .42. Exposureduration was tested as a possible moderator variable and found to influence theaccuracy-confidence correlation. Longer exposure durations allowed for19greater predictability of accuracy from confidence. Several other studies haveidentified more moderator variables. Higher accuracy-confidence correlationsare obtained when subjects are in a state of “retrospective self-awareness,”(Kassin, 1985; Kassin, Rigby & Castillo, 1991), among subjects who made aselection from a lineup (Fleet, Brigham & Bothwell, 1987;) as opposed to thosewho did not make a choice (choosers vs. non-choosers), and when the targetperson is distinctive looking, attractive or both (Brigham, 1990). Finally,Cutler and Penrod (1989) performed a meta-analysis of mne studies in whichsubjects had made two confidence assessments, one before their identificationattempt, and one after. Post-identification confidence ratings correlated muchmore strongly with accuracy than did pre-identification confidence ratings.The confidence-accuracy relationship is relatively easy to study and there havebeen several field studies which have reported on this relationship.DelayLike the fields of botany and biology, psychology has a ‘field season,’ awindow of time during which data must be collected. As most subjects areuniversity students, the field season for studying the effects of delay oneyewitness memory is generally limited to an academic term of eight months orless. The police generally collect statements when they arrive at the scene of acrime or within a few days, so the issue of the effects of delay on recall havenot been too much of a concern for psychologists. Furthermore, although thereare often lengthy delays between the crime and a court appearance,eyewitnesses are often permitted to review their statement. The present review20is concerned only with the effects of delay on identification accuracy. Shapiroand Penrod’s (1986) meta-analysis reported a modest detrimental effect of delayon hits (d .43) and a smaller effect on false alarms (d = .33) andDeffenbacher’s (1986) meta-analysis reported similar results. Laughery andWogalter (1989) reviewed the effects of delay on identification accuracy andfound that five studies showed no adverse effect of delays from 48 hours to twoweeks and five studies which reported a decline in identification accuracy afterdelays from a few days to 11 months. Obviously more research intomoderating factors such as length of exposure, and target distinctiveness areneeded to clarif’ the effects of delay on identification accuracy.System Variable ResearchIn contrast with estimator variable research which typically examines theeffect of a variable on both recall and recognition, system variable researchoften manipulates variables related to only one or the other. System variablesrelated to recall (providing a statement) are presented first, followed by thoserelated to recognition (identification). The present research was unable toaddress most of the system variables. The case files usually did not indicatehow the eyewitness account was collected and copies of lineups were sealed inevidence envelopes and could not be viewed. Yuille (1986) reported that ridealongs with the RCMP showed that they employ an acceptable and uniformprocedure for obtaining statements from eyewitnesses. And with regard toidentification procedures, the officers I spoke with were aware of the problems21of constructing and administering lineups and in the interest of insuring thattheir cases had the best chance of being prosecuted, they had adopted many ofthe procedures suggested by system variable research regarding lineups.Recall VariablesCognitive Interview. One of the most important recent developmentsregarding collection of eyewitness statements has been the Cognitive Interview(Fisher & Geiselman, 1992) which encourage eyewitnesses to use manyretrieval paths and to increase the feature overlap between retrieval andencoding contexts. The Cognitive Interview, compared with standard policeinterviews has been shown to increase the amount of details an eyewitnessrecalls without adversely affecting the accuracy of those details (Fisher,McCauley, & Geiselman, 1994; Geiselman et al., 1984).Rehearsal. To the extent that eyewitnesses go over the event in theirmind or speak about it to others, rehearsal must also be considered an estimatorvariable. It is also possible for police to exert some influence on the timing andtype of rehearsal that eyewitnesses to particularly violent crimes are apt toengage in (Yuille & Tollestrup, 1992). Schooler and Engstler-Schooler (1990)reported an alarming finding that verbalizing a description of a perpetrator, aform of rehearsal, lead to poorer identification accuracy than not describing theperpetrator. However, this finding has not ever been replicated and even if ithad been a reliable result, it is unlikely that police could afford to not askeyewitnesses for description of the perpetrator.22Read, Hammersley, Cross-Calvert and McFadzen (1989) found thatmental rehearsal of an event resulted in little memory loss over a week delaycompared to no rehearsal. With respect to rehearsal’s effects on identificationaccuracy, they found that rehearsing the event immediately after it’s conclusion,compared with no rehearsal lead to an improvement in identification accuracyif the target’s face was unchanged in appearance, but was reduced if the facechanged slightly from viewing to test. If rehearsal was delayed by ten minutes,identification accuracy was improved even if there were changes in theappearance of the face.Suggestibility and post-event misinformation. Loftus and her colleagueshave largely used slide sequences to demonstrate that subjects are influencedby question wording. For example, higher speed estimates were offered bysubjects who were asked how fast a car was going when it smashed than bysubjects who were asked the question with the word hit (Loftus & Palmer,1974). Similarly, Loftus and Zarini (1975) demonstrated that subjects weremore likely to incorrectly answer a question about an item not seen in the eventif the definite article the was used than if the indefinite article a was used.Zanni and Offerman (1978) however, failed to replicate this effect of definitearticles. Again with slide sequences, Loftus, Miller and Burns (1978) showedthat when inconsistent post-event information is given to a subject, theaccuracy of responses to a questionnaire decreases significantly. Christianson,Sweeney and Ochalek (1983) used a live event to demonstrate the impact ofpost-event information. Information that the principal man in the event was a23truck driver lead to larger weight estimates than infonnation that he was adancer. Also, subjects who heard the man referred to as a young mancompared to those whom heard him referred to as just a man provided youngerage estimates. This post-event misinformation phenomenon is still underdebate as to how the misinformation affects the original memory. Somecontend that misinformation actually impairs the original memory, othersbelieve that the misinformation leads to confusion as to what actually happenedduring the event (see Belli, 1989, 1993 for reviews of these differenthypotheses).Identification VariablesPre-identification procedures. Police often ask eyewitnesses to help themget a better idea of the suspects appearance. There are two of these preidentification methods used. In the first, mugshot viewing, the eyewitnesssearches through hundreds of photos of people who have committed crimes tofind the perpetrator. The other method, composite construction, involvescreating an image of the perpetrator. Eyewitnesses may be asked to do bothand if so, the mugshot viewing usually precedes the composite construction.Much of the research into the use of mugshots has focused on the impact onsubsequent identification accuracy that such a practice may have. Subjectswho choose a mugshot photo from a set not containing the perpetrator oftenremain committed to that photo at an identification task even when theperpetrator’s photo is present (Brigham & Cairns, 1988; Brown, Deffenbacher& Sturgill, 1979; Gorenstein and Ellsworth, 1980). Another approach has24examined the possible interfering effects of mugshots and has proposed thatviewing so many faces may corrode the original memory and lead to reducedaccuracy at an identification task. Davies, Shepherd and Ellis (1979) foundthat subjects who viewed mugshots were less accurate in their identificationsthan either a control group that waited an equivalent amount of time beforeattempting an identification or a group who viewed the same mugshots butrated them for attractiveness instead of searching for the targets. Lindsay,Nosworthy and Martynuck (1992) point out that the potential dangers ofmugshots apply only to the extent that viewing the mugshots is seen as anidentification technique rather than an investigative tool. They report thatsubjects are reasonably accurate at picking out a confederate’s photo from up to700 mugshots and they generally do not select many incorrect photos.As with mugshots, much of the concern over composite drawings hasrevolved around possible interfering effects. Interference is measured bycomparing the identification accuracy of subjects who previously constructed alikeness of the target against the accuracy of subjects who did not. Wogalter,Laughery and Thompson (1990) determined that the amount of verbal activityinvolved in construction of the face does not influence subsequentidentification accuracy or quality of the constructed face. They also noted thatthere seems to be a trade-off between image quality and accuracy of laterrecognition. Methods such as sketching and Mac-A-Mug which involvemanipulating a great deal of facial detail produce the highest quality images,but subsequent recognition suffers. In contrast, methods such as Identi-kit and25Photo-fit result in lower quality images, but may improve later recognition. Inthe present research, it was noted if any pre-identification procedures wereemployed, but too few eyewitnesses were asked to look at mugshots orconstruct a composite image to draw any conclusions regarding possibleinterfering effects or to evaluate the quality of composite images.The present research was able to note the frequency and what type of preidentification procedures were used, but the frequency was too low to permit anexamination of the effects of such procedures.Lineup Construction. Bias in the construction of a lineup, for example,by tilting the suspect’s photo (Buckhout, Figueroa & Hoff, 1975) or by dressingthe target in the same clothing worn during the event (Lindsay, Walibridge &Drennan, 1987) leads to higher identification of the suspect than non-biasedconstruction. There are more subtle ways in which the construction of a lineupcan affect identification accuracy. Wells, Leippe and Ostrom (1979)introduced the concept of functional size of a lineup which refers to the numberof viable members of the lineup and is contrasted with the nominal size whichrefers to the number of the people in the lineup. The functional size of a lineupis reduced to the extent that the other members of the lineup (distractors) areeasily ruled out as not being suspected by the police. In an extreme example, ifa suspect is described as being black and the eyewitness is shown a lineup withsix white men and only one black man, the nominal size is 7, but the functionalsize is 1. There is a problem however, if the lineup members too closelyresemble the suspect. For example, if the lineup administrator goes to great26lengths to find distractors that match as many features of the suspect aspossible, then the lineup is said to have lost propitious heterogeneity and theidentification task becomes confusing and difficult (Luus & Wells, 1991).Wells and his colleagues (e.g. Wells & Turtle, 1986; Wells, Seelau,Rydell, & Luus, 1994) have been instrumental in drawing up guidelines for theconstruction of fair lineups. One of the most widely accepted practices is togive a description of the suspect to people who have never seen him and thenask these mock witnesses to “identify” the suspect solely on the basis of thedescription (Wells et al. 1979). If the lineup is fair, then the choices of themock witnesses will be distributed equally across all lineup members.Lineup Administration. Several studies have demonstrated thatsequential presentation of a lineup in which pictures are presented one at a timeand a yes/no judgment must be made for each picture, significantly reducesfalse alarms and does not affect correct identifications when compared to thetraditional presentation method in which all lineup members are presentedsimultaneously (e.g. Cutler & Penrod, 1988; Lindsay & Wells, 1985; Melara,DeWitt-Rickards & O’Brien, 1989).The instructions an eyewitness receives regarding a lineup can alsoinfluence eyewitness accuracy. Eyewitnesses viewing a sequentially presentedlineup are responsive to non-verbal suggestion from the administrator of thelineup such as leaning forward or pausing at a particular photo regardless ofwhether the photo these behaviors were directed at was the target (Smith,Pleban & Shaffer, 1982). Failing to inform eyewitnesses that the offender may27not be present (e.g. Cutler, Penrod & Martens, 1987; Malpass & Devine, 1981;Wamick & Sanders, 1980) can dramatically inflate false identifications ofinnocent suspects. Kohnken and Maass (1988) used identical instructions asMalpass and Devine (1981) and additionally manipulated whether or notsubjects were informed about the staged nature of the event and subsequentidentification. They successfully replicated Malpass and Devine’s findings, butonly with informed subjects which suggests that eyewitnesses may be lesssusceptible to biased instructions than previous research had indicated.Finally, the general topic of the accuracy of eyewitness identificationsneeds to be addressed. About one in every eight articles reviewed in this thesiscommented on the great range of eyewitness identification accuracies in theliterature and then went on to outline and explore a different explanation forthis state of affairs. The range is impressive; from a low of 14.1% (Buckhout,1975) to highs of over 80% (e.g Malpass & Devine, 1981). Given all of thevariables that can affect identification performance such as the distinctivenessof the target, various factors related to lineup construction and administration,and delay between exposure to the target and identification just to name a few,it seems hasty to draw any general conclusions other than it is multiplydetermined and that we know of some of the factors that influence theidentification performance of laboratory eyewitnesses.Concluding remarks about laboratory researchThere are several advantages and strengths of the laboratory approach tothe study of eyewitness memory. Of all the methods, laboratory research is28most likely to be the least expensive and least time consuming to execute.These are not trivial matters in the publish or perish world of academics. Theprecision with which variables can be controlled is unrivaled, making afactorial design conducted in a laboratory setting a very powerful exploratorytool. Similarly, there are some factors such as drug or alcohol use that can notbe studied in situ. However, laboratory experiments are limited in the extent towhich they can arouse subjects and it is difficult to instill in subjects the beliefthat their identification will have consequences for subject and person theyidentify. Although realism is usually strived for in the event itself in that somekind of crime is staged, subjects are normally debriefed about theirparticipation in an experiment before they attempt an identification. Thus theydo not have to worry about making a false identification or letting a guiltyperson remain free and unpunished. They do not have to consider that theymight be involved in a lengthy trial and they do not risk possible retributionfrom the person they identify.This issue was illuminated by Malpass and Devine (1984) who noted that“while realism is not the only issue, and not the only strategy of importance,until we know more about the degree to which simulations actually produceresults which can be validly applied to genuine events in the naturalenvironment, realism will be an important aspect of psychological research inthis area, and its absence will be an important source of reservation about theapplicability of the research literature” (p. 86). Only a few studies havemanipulated realism by informing or not informing subjects before they made29an identification that the event was staged (e.g. Murray & Wells, 1982; Sanders& Wamick, 1981). Sanders and Wamick (1981) found no difference inidentification accuracy between informed and uninfonned subjects.Uninformed subjects in the Murray and Wells (1982) study were more likely tochoose a foil and less likely to identify the offender from a target-presentlineup. They also had lower confidence-accuracy relationship than subjectswho were informed that the event was staged. Murray and Wells (1982)conclude that “...there should be some value placed on the general plan ofcorroborating results from the informed witness procedure by using anuninformed witness procedure” (p.52). This suggestion has gone largelyunheeded and it is likely that the tremendous logistical and proceduralproblems involved in keeping subjects uninformed in a laboratory experimenthas contributed to this state of affairs. Even when subjects have been kept inthe dark through the identification procedure about the staged nature of theevent, the fact remains that the staged event must be relatively innocuous.Murray and Wells (1982) staged a theft of a T.V. game and Sanders andWamick (1981) staged a cheating episode. Also, the perceived status ofcampus security (Murray & Wells, 1982) and the proctor of an exam (Sanders& Wamick) as lineup administrators may be different from that of a policeofficer. Field research lends itself more readily to keeping subjects uninformedthroughout the entire procedure but there are still limits on the events that canbe staged for an unsuspecting public and the role in which the lineupadministrator can be cast.30Field ResearchThe majority of these studies take place outside of the confines of thelaboratory or class room; those that do not were published as an integratedseries of laboratory and field research. Similarly, most of the subjects in thesestudies are from a population other than university students and most were notinitially aware of their participation in an experiment. Field research onestimator variables is presented first, followed by the field research on systemvariables.Estimator VariablesSeveral field studies have adopted a procedure first employed by Brighamet al. (1982) so the gist of it is reported here first. A confederate orconfederates are sent into convenience stores to pose as customers. Theinteraction is engineered to be memorable such as paying for an item withpennies or buying a soda with a money order. To insure eye contact and tomake the interaction longer, the confederate asks for directions to a distantlocation or inquires about a product not sold in that state. Some time later,other members of the research team return posed as law interns who arelooking for someone who might have been in the store within the past 24 hours.They show the clerk a photo lineup and administer any questionnaires theyhave and then, after all the data has been collected, they inform the clerk thatshe/he has been in an experiment. This procedure is referred to here as the lawintern procedure.31Individual DifferencesHosch and Platz (1984) used the law intern procedure to examine therelation between self-monitoring scores and identification accuracy. Theyfound a significant correlation, with high self-monitors being more accurate atidentif’ing the confederate from a target present lineup. The correlation in thefield was roughly twice those obtained in laboratory studies, suggesting that atleast for non arousing events, self-monitoring scores are related toidentification accuracy.Flin and Shepherd (1986) examined people’s ability to estimate the size(height and weight) of others. They were particularly interested in learninghow height and weight estimates were influenced by these same characteristicsin the person making the estimate, the person being described and a contextperson asking for these estimates. There was an overall tendency tounderestimate height. All subject’s estimates of height were influenced by theactual height of the target. The underestimates of both women and menincreased with the height of the target, so that the greatest underestimatesaccompanied the tallest targets. The subject’s own height also contributedsignificantly to estimates of the target’s height such that the taller subjects(female and male) produced more accurate estimates. The height of thecontext person did not have any effect on height or weight judgments bywomen or men.On average, men underestimated, and women overestimated weight.The actual weight of the target influenced the estimates of both women and32men. Weight estimates showed regression toward the mean with heavierestimates of targets who were below average weight and lighter estimates oftargets who were beyond average weight. The subject ‘s own weightinfluenced only the males estimates, with heavier men providing more accurateestimates. These results suggest that the police might be aided in determiningthe height and weight of a perpetrator by obtaining the height and weight of theeyewitnesses providing these estimates. Additionally, researchers employinglive events who ask their subjects to estimate the size of the target(s) shouldreport the target’s height and weight.Situation VariablesPerhaps because of its social significance, cross racial identification isone of only a few topics that has been studied in three different researchparadigms. Brigham et al. (1982) developed the law intern procedure that keptsubjects ignorant of their participation in an experiment until after the datawere collected. Convenience store clerks were engaged by two confederates(one white, one black) in separate, unusual but non-threatening transactions.Two hours later, other members of the research team posed as “law interns” andhad the clerks examine two target-present lineups (one for each confederate).After their identification decision the clerks were asked to rate their confidencein terms ofhow willing they would be to testify in court about it, which helped,along with their nebulous connection to the CJS, to instill the belief that theiridentification would have consequences for the “suspect” and also perhaps forthemselves.33The white clerks who rated themselves as the having the most cross-racial experience were more accurate and reported higher confidence in theiridentification of the black confederate than the less experienced white clerks.Unfortunately, cross-racial effects amongst blacks could not be examined asthere were too few black clerks in the sample. Across all confederates,identification accuracy was related to distinctiveness and to distinctiveness andattractiveness combined. This finding replicates the typicality effect (e.g.Vokey & Read, 1992) demonstrated in photo recognition research.Accuracy of description and identificationPigott, Brigham and Bothwell (1990) used the law intern procedure toexamine the relation between accuracy of description and identification. Thebank-teller subjects in this experiment provided a description by means of astandard police suspect identity chart that listed 16 features of appearance.Each feature was accompanied by a list of descriptive nouns from which tochoose. There was no relation between accuracy ofdescription andidentification. It is possible that the three laboratory-based studies thatincidentally found support for a relation between description and identificationaccuracy just happened to employ targets that were easily described and thus,as Wells (1985) reported, more easily identified. However, two of the threethat found a significant correlation employed a live, staged theft of an itembelonging to the subject (Hosch & Bothwell, 1990; Kassin, 1984), and the third(Buckhout et al, 1974) employed a theft staged live in front of a classroom.All of these events would have been more arousing and engaging than the34events employed in the studies that did not find support for a relation betweendescriptive and identification accuracy. Of the four studies that failed todemonstrate a relationship, the field study ofPigott et a! (1990) employed themost engaging event; the target tried to cash an obviously altered money orderwith bank teller subjects. This event was successful in arousing the suspicionof the tellers, several followed the confederate out of the bank to record hislicense plate, but it may not be the same as having one’s own purse stolen orwatching a brazen theft followed by a vigorous pursuit. Wolfskeil andBrigham (1985) used a live target, but he simply stood in front of the subjects,and Wells (1985) used only photos. Of course it remains to be seen if thisdifference in arousal induced by the to-be-remembered event plays any part inthe pattern of results obtained thus far. There were other differences that mayplay a role as well, such as the manner in which the descriptions werecollected (free recall vs. a checklist), length of exposure to the target and delaybetween exposure and identification. However, I could not discern anypattern between these variables and whether or not the study found acorrelation between descriptive and identification accuracy.Unconscious TransferenceRead (1990) has been very inventive and entrepreneurial in finding fieldsettings. He has utilized the captive audience of shopping mall clerks, andgone door to door soliciting the help of an unsuspecting public. He has alsobeen very adept at weaving the techniques and findings of laboratory-based andfield research into a comprehensible whole. Read, Tollestrup, Hamrnersley,35McFadzen and Christensen (1990) tried to demonstrate the phenomenon ofunconscious transference in a series of lab and field studies. Unconscioustransference refers to an eyewitness’s misidentification of an innocent personfor a perpetrator because of the witness’s exposure to the (innocent) person inanother context. This phenomenon was demonstrated in an actual eyewitnessby Houts (1956) and as Read et al. (1990) note “. ..is a phenomenon that has, byvirtue of its frequent reference, been reified as a danger in person identificationfor which we should ever be vigilant” (p. 3). The only experiment todemonstrate unconscious transference (Loftus, 1976) used photographs seenonly for 2 s, rather than a live event and more accurately demonstrates thatpeople are poor at recalling the circumstances of encountering photos of faces.Read Ct al. (1990) conducted a series of five field and laboratory-based studiesthat involved over 700 subjects, five retention intervals, seven intervalsbetween exposure to the perpetrator and the innocent person, two levels oflineup similarity (of distractors to the perpetrator), four different innocentpersons, and four different perpetrators. Despite persistent efforts, Read et al.were only able to demonstrate unconscious transference when they returned toa controlled laboratory setting, when the perpetrator and innocent person weresimilar in appearance and with a lineup in which the similarity of otherdistractors to the perpetrator was low, thus making the innocent person the onlyreasonable choice. This topic might be a good example of something that canhappen to actual eyewitnesses, but only under a relatively rare set ofcircumstances. The present research was unable to address this phenomenon.36ArousalYuille, Davies, Gibling, Marxsen and Porter (in press) employed a veryunusual and exciting paradigm in their research on eyewitness recall. Theymade use of the Police training Centre in Hendon England which employs arealistic village setting. Recruits are sent into the village in groups of four tofive with one designated as the “on patrol” officer, another videotapes theevent, and the rest are observers. The recruits encounter a variety of situations,that range from routine and easy to deal with to unusual and difficult. Theauthors took advantage of these naturally occurring events that have thevariable of participation “built-in” and which also permit manipulation of thedegree of arousal experienced and the frequency and timing of recollections ofthe event. One half of the recruits were interviewed twice; once a week afterthe event and again after 12 weeks while the other halfof the recruits wereinterviewed only after 12 weeks. Manipulation checks confirmed that thestressful event was experienced as more stressful and more difficult than thenon-stressful event by both participants and observers.Recruits who were interviewed twice offered more details in their firstinterview than their second and those in the low stress roleplay recalled more,but were less accurate than those in the high stress role play. The loss of detailfrom the first to the second interview occurred only for recruits who saw thehigh stress role play. On average, recruits in the high-stress roleplay suffered a22% decrease in the amount of information from the first to the final interview,and recruits in the low stress roleplay offered the same amount of information37in both interviews. Finally, for recruits who were interviewed twice,participants recalled more than observers in the delayed recall. In analyses ofonly the recalls provided at 12 weeks, the advantage of an early interview wasobvious. Subjects who were interviewed twice remembered more and weremore accurate in their final interview than were those who were onlyinterviewed once. Again, recalls of the low stress event were more detailed,but less accurate than the high stress recalls.The effects of stress as demonstrated in this study fit Easterbrook’stheory of the effects of stress on attention. The high-stress subjects may havefocused their attention on fewer, more relevant details. But why should they bemore accurate than low-stress subjects and why should they alone suffer decay?The pattern of memory loss across high and low stress groups does not fit withthe remarkable memory hypothesis of Yuille and Tollestrup (1992). Onepossible explanation is that the events depicted in the role plays did not merittoo much consideration after they were over. Furthermore, recruits were askednot to discuss the event with others.There are many differences between this study and typical laboratorystudies. The first recall was a week after the incident and most lab studies donot wait so long. The interview was “standard police protocol” in whichsubjects first described the event in free recall and then answered standardquestions designed to elicit specific details. Most lab studies examine amountand accuracy of free recall and questions separately, however in this study they38were combined. But, like lab studies subjects knew they were participating in aresearch study and that they might be questioned later.Confidence and AccuracyThe relationship between confidence and accuracy is often much higherin field studies than in lab studies. Brigham et al. (1982) reported an overallcorrelation between confidence and identification accuracy of r =.50. Pigott etal. (1990) reported that their confidence-accuracy correlation among those whomade a choice from the lineup was .42. and Krafka and Penrod (1985) report acorrelation of .52 among choosers who reinstated the context of theirinteraction with the perpetrator prior to viewing the lineup. However, in a labstudy (Murray & Wells, 1982) which kept some subjects ignorant of theirparticipation in an experiment until after they had made an identificationdecision on a lineup, the uninformed subjects had lower confidence-accuracycorrelations than subjects who knew they were taking part in an experiment.As these uninformed subjects were essentially in a field study, it can’t beunequivocally stated that the confidence-accuracy relationship is higher in thefield than in the lab. The confidence-accuracy relation could not be examinedin the present study because it was not standard police procedure to obtainconfidence judgments from eyewitnesses after their identification.The effects ofdelayKrafka and Penrod(1985) report there was no effect of delay (2 vs 24hours) for subjects who viewed target present photo spreads, but for those39viewing target absent arrays, the proportion of false choices after 24 hours wasmuch greater (52.4%) than after two hours (15%). This finding is interestingin light of a pilot study reported in Brigham et al. (1982) which found thatclerks exposed to an event of similar content and duration performed at chancelevel on a target present lineup after 24 hours. Brigham et al. did not usetarget-absent lineups. What could account for one study finding a devastatingeffect of only a 24 delay and the other finding no effect of the same delay,especially since they employed similar events? One possible explanation isthat the specificity of the questions posed to the clerks at identification tasksdiffered between these two studies. The clerks in both studies were asked tolook at the lineup to see if they recognized anyone who might have been in thestore in the past 24 hours. The clerks in the Kraflca and Penrod study were toldwhat this person probably did while in the store, and clerks in the Brigham etal. study were not. Thus, the clerks in one study received more informationabout the identity of this person than those in the other. A second explanationcould lie in the fact that the confederate in the Krafka and Penrod study, theone demonstrating no effect of a 24 hour delay, drew attention to hisappearance during the interaction. The confederate in the other study did not.The events in these studies are more akin to fraud where eyewitnesses areunaware that a crime has occurred until some time later when the cheque isreturned NSF or the credit card is reported stolen or invalid. The present thesiswas able to address the effects of delay on identification accuracy for both40fraud and robbery eyewitnesses as well as some variables, such as weaponpresence, that moderate this relationship.System Variable ResearchPose ChangeAn interest in the generalizability of photo recognition studies on theeffects of pose change on identification accuracy prompted Logie, Baddeleyand Woodhead (1987) to conduct a combination of laboratory and field tests.First, they employed standard facial recognition research to demonstrate therobust effect that recognition accuracy is influenced by the type ofpose at study(e.g. Davies, Ellis & Shepherd, 1978; Woodhead, Baddeley & Simmonds,1979) More information at study (frontal, three-quarter and profile poses)leads to better performance and ifjust a single pose is used, three-quarter leadsto the best performance. They also demonstrated that a change in pose fromstudy to test leads to poorer performance than if the pose remains unchanged.Then, in a laboratory study subjects saw the target in a live staged event andwere tested on lineups composed of photos in one of the three pose types. Asin the photo recognition work, the three-quarter pose lead to better recognitionperformance than the profile pose, but did not differ from the frontal pose.The shorter delay between the event and test lead to better identificationperformance, but had no effect on the pattern of results for pose. These resultssuggest that there is no need to alter current police practice of composinglineups of frontal poses.41In field research, subjects were given photos of six confederates indifferent poses to examine for 15 minutes. Subjects then followed the samewalking route as the confederates, except in the opposite direction. They wereto note the time and location as well as attire of any confederates they spotted.Only half of the subjects reported seeing one or more confederate and of these“sightings”, false recognitions greatly outweighed correct identifications. Posedid not affect the rate of false recognitions. In an attempt to improveperformance, they replicated the experiment, with subjects searching for onlyone target instead of six. With this modification, subjects were much moreaccurate at detecting the confederate. Once again, there was no effect of poseon identification accuracy. As this scenario is more akin to someoneresponding to a missing person ad or a wanted poster, these results suggest thatthe current practice of displaying front-posed photos of missing or wantedpersons need not be altered. Further tests in which subjects see a live targetand recognition is tested by using photos of different poses are needed tocomplete the picture of the effects of pose change. In this scenario though andin the context of an actual crime, it is conceivable that subjects would see a livetarget from a variety of angles, frontal, three-quarter and profile, and thuswould not be seeing just one “pose” but several.Context ReinstatementKrafka and Penrod (1985) modified the law intern procedure by havingthe confederate drawing attention to his appearance by commenting on howmuch he had changed compared to the photo on his driver’s license which he42produced in order to purchase an inexpensive item with a traveler’s cheque.Before viewing the lineup, half of the clerks in the context reinstatementcondition (CR) were asked to recall the transaction, to visualize theconfederate’s face, and were presented with photo copies of his non-photoidentification and a cheque signed by the confederate. The other half in the nocontext reinstatement condition (NCR) did not receive any of these instructionsor materials. Context reinstatement is one of the components of the cognitiveinterview and this experiment demonstrated its potential for becoming anadvantageous system variable. Importantly, context reinstatement did not altersubject’s willingness to choose. It only affected lineups that contained theconfederate’s photo, in which case, the identification accuracy in the CRcondition was nearly double that of the NCR condition. And finally, thecorrelation between confidence and identification accuracy among CRchoosers (r(19) = .52) was the only significant accuracy-confidence correlationfound in this study.The principal advantages to field studies is that subjects are often drawnfrom a non-university population and they usually don’t know they areparticipating in an experiment which makes it easier to make them believetheir identifications will have consequences. However, because theexperiment has to be taken to the subjects in field research, it is more timeconsuming and costly than laboratory research and often suffers from higherattrition rates. A slight loss of control has to be suffered in field research, butfactorial designs are still feasible. Events in field research can’t approximate43very high levels of arousal, and judging from the events in the currentcollection of field studies, it may be even more difficult to come up with fieldevents that are within ethical limits, or at least within the comfort zone of theresearchers than it is to design laboratory events. None used an event in whicha subject was victimized, and I’m not sure that it would be possible to do so in afield study. This makes field research a good analog for fraud, a lessglamorous, but certainly still costly crime. For example, credit card fraud inthe United States amounted to one billion dollars in 1982 and it is predictedthat shortly, these losses will climb to two billion dollars (Caminer, 1985).Several field studies on the eyewitness abilities of children have takenadvantage of naturally occurring stressful events such as dentist appointmentsor innoculations (e.g. Peters, 1991). While these experiences might be able toinform us about memory for arousing events, they are not good analogs forstudying adult eyewitness memory.Case Studies and Archival ResearchAlthough the methods in this section are diverse, the one element theyhave in common is an examination of the memory of actual eyewitnesses. Incontrast to other types of research which have focused on identification, thisresearch method has focused on eyewitness descriptions.In the only study involving actual eyewitnesses to manipulate a variable,Fisher, Geiselman and Amador (1989) tested the effectiveness of trainingpolice officers in the cognitive interview. Experienced detectives in therobbery division of Miami Florida took part in this pre-test post-test design.44The cases selected for inclusion met the following criteria: serious enough torequire an in-depth interview by a detective (as opposed to those only requiringthe initial interview by the uniformed officer at the scene of the crime); at leastone eyewitness had sufficient chance to observe the crime; and eachinterviewed eyewitness had to be fluent in English and be cooperative. Theinterviews were primarily with victims of commercial robbery or purse-snatching. All of the detectives recorded several pre-traimng interviews usingthe standard police procedure. One-half of the officers were trained in theCognitive Interview, with the remainder forming the control group. Post-training interviews of the trained and control groups were recorded. Allinterviews were transcribed and scored by people blind to the officer’s trainingstatus. The scoring consisted of counting the number of relevant, objectivestatements made by the eyewitness in the interview. The scored statementsprimarily concerned physical descriptions of the perpetrator and relevantactions, clothing, weapons, vehicles, objects taken and conversations.The trained detectives obtained 47% more information in their post-training interviews than their pre-training interviews. A comparison of theamount of information obtained by trained versus control officers indicated thatprior to training the two groups of officers elicited an equivalent amount ofinformation, but in the post-training interviews, the trained officers elicited63% more information than the control group. Importantly, a comparison ofthe interview conducted at the scene of the crime by a uniformed officer andthe interview conducted by the trained detectives in terms of the amount of45same, different and new information revealed that the increase between preand post-trained interviews was attributable to new information. Training inthe cognitive interview was effective in eliciting more information and did notcontaminate or discredit the eyewitness by increasing the amount ofinformation that was different or changed between the initial interviewconducted by the (untrained) uniformed officer and the trained detective.Although an examination of the effects of the cognitive interview onaccuracy of recall was not possible, the authors were able to look atcorroboration in cases involving two or more eyewitnesses. Over 94% ofcorroborable statements were corroborated (between eyewitnesses) and therewas no difference between the pretrained and the posttrained interviews.Thus, this test of the cognitive interview replicated laboratory study findingsthat the cognitive interview boosts the amount of information recalled withoutadversely affecting the accuracy of the recall.Kuehn (1974) examined the ability of victims of violent crimes todescribe the appearance of their assailants. His sample consisted of 22 rapes,15 assaults, 61 armed robberies and 2 murders whose victims lived long enoughto provide a description. All of the cases involved a single account from avictim who was unacquainted with the perpetrator. His analysis consisted ofwhether the victim’s description of the perpetrator included each of thefollowing nine physical descriptors: race, sex, age, height, weight, build,complexion, hair color and eye color. The study did not involve an assessmentof the accuracy of descriptions. Although not expressly stated, it was casually46implied that the police force used a form to collect standardized descriptionsand that this was the source of the nine traits.Only four victims were unable to describe their assailant. The meannumber of traits in the victims descriptions was 7.2, the mode was 8 and over85% of the victims described six or more traits. Descriptions most oftenincluded the assailant’s sex (93%) and least often included eye color (23%).The remaining traits were listed in more than 70% of descriptions.There was plenty of evidence in Kuehn’s data to suggest that arousal has anegative effect on the completeness of descriptions. Victims of robbery,possibly overall the least arousing crime in his sample, provided fullerdescriptions than assault or rape victims. Injury had an overall negative effecton completeness. Injured victims of all crime types provided less completedescriptions than uninjured victims. Nonetheless, injured victims of robberystill provided more complete descriptions than injured victims of rape orassault. Male victims of robbery or assault provided fuller descriptions thanfemale victims of these crimes. Also, injured males provided more completedescriptions than injured females. It is possible that the female victims ofassault were more frightened than the males, so this arousal could underlie thisgender difference.Descriptions provided in crimes involving white victims and black orwhite suspects were more complete than descriptions provided by blackvictims. Kuehn raises the possibility that the police were not particularlydiligent in assisting black victims, but as there were only six black victims, the47result could also be an artifact of the small sample size. In contrast tolaboratory studies, the presence of a weapon was not related to completeness,neither was victim intoxication.In a more recent archival study, Sporer (in press) also examinedeyewitness descriptions of perpetrators and factors that affect the number ofdetails in these descriptions. His sample consisted of 139 descriptions fromeyewitnesses to bank robberies, forcible rape, indecent exposure and violationsof public safety. All of these cases involved at least one person description andan identification attempt by an eyewitness. Sporer argued that selecting casesin which an identification was attempted “. .implies that all cases can beconsidered major violations, because otherwise police would not have gonethrough the trouble of constructing identification parade&’ (p. 9). The accuracyof person descriptions was not assessed, nor was any attempt made to examineidentification procedures or eyewitness identification performance. Sporer’ssample included 52 victims of the crimes, 12 non-victim witnesses and 75“incidental witnesses’ who did not directly witness the crime but who had beenquestioned by the police about the appearance of the perpetrator whom theyhad observed on a different occasion. Unlike Kuelm (1974) his analysis wasnot restricted to nine physical appearance traits, but encompassed whole persondescriptions (clothing, jewelry, physical appearance etc.) and appears to havebeen based on free recall.The overall average number of descriptive details was 9.71. Victims andwitnesses offered an average of 13.17 and 11.67 descriptive details48respectively, significantly more than the 7.79 offered by incidental witnesses.The total sample of person descriptions was composed of 30% details aboutfacial features, 31% about clothing, a combination of age, height, stature andrace accounted for 22.4%, and the remaining details were about thepersonality, jewelry, dialect, smell and disguise of perpetrators.Sporer examined the effects of a number of variables on the amount ofdetails in eyewitness descriptions. There was no influence of gender for any ofthe eyewitness types in his sample. Subsequent analyses exclude incidentalwitnesses. Alcohol had an adverse effect on the amount of detail indescriptions. These results and those on gender differ from those of Kuehn(1974) who failed to find an adverse influence of alcohol, but did report agender difference in favor of males. As one would expect, descriptions forcrimes that occurred in good and medium levels of illumination included moredetails than those that occurred in low levels. As with Kuehn’s study, therewas no support for a weapon focus phenomenon.Using information available in police files, Sporer operationalizedarousal in four (likely non-independent) ways; the degree of injury to theeyewitness, whether or not the eyewitness engaged in self defense, the degreeof threat to life, and anxiety. A main effect of arousal for each of the fourvariables was reported. However, for all four variables a graph of the numberof details provided at each level showed a U-shape with similar amounts ofdetail provided at the lowest and highest levels and fewer details offered atmiddle levels. There was no report of any tests conducted to determine if the49amount of detail at middle levels of arousal was significantly lower than thatprovided at low and high levels.Cutshall (1985) combined archival and case analyses. She first employedthe archival method to learn about the demographic characteristics of peopleinvolved in violent crimes and the configuration (number of victims, witnessesand perpetrators and their relation to one another) of these crimes. Then, arobbery case which had plenty of eyewitnesses and sufficient evidence toreconstruct the crime was selected for an examination of the amount andaccuracy of information contained in eyewitness’ recollections of the incident.This research was the first to assess the accuracy of eyewitness memory in situ.The archival analysis presented in Cutshall (1985) demonstrated thatmurder, robbery and assault present different patterns of eyewitnesses. Forexample, in non-sexual and sexual assaults, most often the victim is the onlyeyewitness. If other people are present during the assault, they often know thevictim and/or the perpetrator. In contrast, about half of the robbery casesinvolved additional eyewitnesses and they rarely knew the victim or theperpetrator. This simple descriptive data is often cited as reason to conductcase and archival research. The more we know about the contexts to which wewant to generalize laboratory findings, that is the more we know about thevarious contexts that actual crimes take place in, the more adequately we willbe able to determine if our laboratory studies match these real world contextsand how important a match or mismatch of context is.50Cutshall’s case analysis involved a gun shooting incident. This particularcase was selected for three reasons. There was sufficient evidence to assess theaccuracy of both descriptive and action details. There were enougheyewitnesses to compare their accounts, and finally, the death of the perpetratorclosed the case which allowed the witnesses to be interviewed withoutinfluencing the outcome of a trial. The police interviewed 21 eyewitnesseswithin two days of the incident. Four to five months later, 13 of thesewitnesses participated in a research interview (the victim of the crime declinedparticipation). This research was able to address questions regarding theamount of detail in eyewitness accounts, the accuracy of these accounts,consistency of eyewitness accounts over time, the effects of stress, andsuggestibility of eyewitnesses. Identification however, was not an issue as theperpetrator was killed in the shooting.Details in the eyewitness accounts were classified as either action, peopledescription or object description. As there was no appreciable difference inaccuracy between the two interviews, only the results from the police intervieware included here. Descriptions of objects were most accurate (88.53%) andpeople descriptions were least accurate (75.57%). Many of the errors inperson description came from inaccurate estimates of age, height and weight.Fifty percent of these estimates offered during the police interview wereincorrect. It was not possible to assess if Fun and Shepherd’s (1986) fmdingson providing estimates of height and weight were replicated.51The research interview was more broad in scope than the police interviewand witnesses actually provided more details four to five months later than theydid at the time of the crime. This precluded an analysis of the effects of delayon the amount of information, but did allow for a consideration of the effects ofdelay on the consistency and accuracy of information. Of the details in theresearch interview that had also been offered to the police, just over 80% wereconsistent (whether right, wrong, or unclassifiable). The overall accuracy ofdetails repeated in the research interview was comparable to the overallaccuracy from the police interview (79.4% and 82.1% respectively). Thissuggests that these eyewitnesses were able to maintain an accurate memory ofthe event over a period of five months. In addition, witnesses were resistant tosuggestive questioning about the perpetrator’s car, a peripheral object duringthe event.Five witnesses who had contact with either of the two main players in theshooting or with a weapon reported the greatest amount of stress, thus stressand proximity to the action were confounded. The accuracy of this higherstressed group was greater in both the police and the research interview. Thequestion about whether or not it was the arousal or the opportunity to view thatlead to this group being more accurate is partly moot in that these two factorswill almost always be confounded in natural events with a third, participation(being a victim).Another, less deliberate approach to case analysis is to wait foropportunities to present themselves, usually crimes reported by the news media52or referred to a psychologist by police or some other player in the criminaljustice system. Christianson and Nilsson (1989) report the course of amnesiaand successful memory recovery of a woman who was raped while jogging.Identification was not an issue once her memory returned as the perpetratorconfessed and as well, she had covered her eyes and did not see his face.One of the most important strengths of case and archival research is thatthey study actual eyewitnesses. So far, the research has focused on violentcrimes, mainly because the other methods of research can only induce mild tomoderate levels of arousal, thus leaving the high end of arousal unexplored andunexplorable by these methods. Case and archival research can provideinformation about the context of eyewitnesses in various crimes that can beused to direct and inform laboratory research.On the down side, there is an almost complete lack of control. Banajiand Crowder (1989) note that “._the multiplicity of uncontrolled factors innaturalistic contexts actually prohibits generalizing to other situations withdifferent parameters” (p.1189). This statement seems a bit extreme, but is notwithout merit. Another issue that makes generalizability a bit moreproblematic is that instead of dealing with a known event in a laboratory,researchers doing case and archival work have to settle for less than an absolutedetermination of truth or accuracy. For example, a very special set ofcircumstances were required in Cutshall’s case analysis so that the actualmechanics of the crime could to be “triangulated” in order to evaluate theaccuracy of eyewitness memory. Data are very difficult to analyze with53traditional statistical tools. And finally, both case and archival research requirea phenomenal amount of time, work, and patience. Simply getting through thered tape to gain access to the police files in this project took well over one fullyear.The present research was able to overcome and accomodate many of thedifficulties of archival research. The goals of this study were to provide dataon the amount and accuracy of details in eyewitness descriptions ofperpetrators, to determine the the frequency of identification attempts in casesof robbery and fraud, to learn about how often eyewitnesses select the policesuspect when viewing a lineup, and to learn about the influence of delay andarousal on recall and recogntion.54METHODData Source and Description of SampleA member of the Richmond detachment of the Royal Canadian MountedPolice (RCMP) supplied a list of all cases of robbery committed during theyears 1987 through 1989 and all cases of fraud committed in the year of 1989.This list contained 119 cases of robbery and 66 cases of fraud. Some case filesrepresented requests for assistance from other police departments, unfoundedcases (usually a false tip of criminal activity) or fraud between businessassociates who knew each other. These types of files, involving 42 robbery and45 fraud cases, did not contain any information regarding eyewitnessdescriptions or identification attempts and thus were not suitable for analysis.The final sample contained 77 cases of robbery and 21 cases of fraud.The author read each case file and recorded information about the crimeitself, eyewitness accounts of the crime, and eyewitness identifications. Anelaboration of the types of data collected under each of these three categories,the coding thereof follows. First, the reader should be aware of a few stylisticconventions that have been adopted to aid smoothness of expression. Severalrobbery cases involved multiple perpetrators. I have adopted the convention ofusing only the singular throughout the thesis. All perpetrators of robbery andthe majority in fraud cases were male. Therefore, the male gender is used torefer to perpetrators. Finally, the term ‘eyewitness’ is used when a distinctionbetween victims and non-victims is unnecessary. When finer distinction is55required, ‘victim’ refers to individuals who directly interacted with theperpetrator and ‘witness’ refers to individuals who were present during all orpart of the crime but who did not directly interact with the perpetrator.Descriptive Information about the CasesThe data collected about the crime itself included the date, the amountstolen, number and type of eyewitnesses (victim or witness) and perpetrators,demographic data about the participants (age, sex, race), their relation to eachother, and in cases involving an arrest, the police description of the chargedsuspect and whether a confession was offered.Eyewitness AccountsIn lab, field, and to some extent case research, the amount and accuracyof eyewitness recall is based on descriptions of people (the perpetrator andvictim), objects (weapons, details of other objects in the event) and actions (ofperpetrator and eyewitnesses if relevant). In contrast, eyewitnesses in thepresent study generally offered statements regarding what the perpetrator(s)looked like, and if they said or did anything threatening such as waving a gunabout and shouting “nobody move and nobody gets hurt.” Other actions of theperpetrator, and objects not carried, driven or worn by the perpetrator were notof interest to the police. Consequently, the present thesis can offer informationpertaining only to eyewitness descriptions of perpetrators. Because otherresearch methods consider eyewitness recall differently, the extent to whichcomments can be made about the generalizability of those findings on the basisof the present fmdings is limited.56Scoring of Eyewitness AccountsTwo people reviewed the eyewitnes& accounts and coded the data for agross count of the number of clothing and physical appearance details. Itemssuch as age, hair color, facial hair, and complexion as well as personalityjudgments (e.g. nasty), odor (e.g. he smelled of cigarettes) and gait (e.g. He hada loping kind of walk) were coded as physical appearance details. Anymention of the perpetrator’s attire including hats, jewelry, glasses and bags (e.g.gym or army-type bags, purses) was categorized as clothing details.Within these categories, the amount of detail was calculated by theprocedure developed by Yuille and Cutshall (1989). This procedure parsesdescription into object-adjective sets such as ‘curly black mustache’ or ‘plaidpants’. A point is assigned to each object and to each non-redundant adjective.The phrase ‘curly black mustache’ would be assigned 3 points (1 each forblack, curly and mustache) and the phrase ‘plaid pants’ would be assigned 2points (1 each for plaid and pants). Any mention of an absence of facial hair orglasses (e.g. ‘He didn’t have a mustache’ or ‘He wasn’t wearing glasses’) wasconsidered to contain as much information as a mention of the presence ofthese features and was awarded points accordingly (one point in each example).In cases where an eyewitness provided a range for estimates of height, weightor age (e.g. ‘He was between 20 and 25 years old’), the midpoint of the rangewas calculated if it was five inches or less, ten pounds or less or five years orless. Ranges that exceeded five inches, ten pounds or five years wereconsidered less precise and were scored with only half of a point. Vague57descriptions such as “he was tall” or “he was middle-aged” were also assigned ahalf of a point unless accompanied by a number estimate in which case themore precise number estimate was assigned points, and the vague descriptionwas not. For example, if an eyewitness said that a perpetrator was “really tall -probably between 6’5” and 6’7”, the height estimate would have been recordedfor scoring purposes as 6’6” and would have been assigned one point. Anyqualifiers in a statement such as “I’m not sure, but he may have had Nikerunners on” were disregarded. Inter-rater reliability for scoring the amount ofclothing detail and phsyical appearance detail were .99 and .97 respectively.The accuracy of eyewitness accounts could be evaluated only in caseswhere a suspect was arrested. In addition, to account for the possibility thatthe police may not always charge the guilty party, the accuracy of eyewitnessaccounts is considered only in cases in which the perpetrator confessed and/orwas apprehended at the scene of the crime. In cases involving an arrest, thepolice measured the suspecfs height and weight and recorded his age, hair andeye color, the presence of facial hair and any unusual marks such as tattoos,scars, or birthmarks. Even if a suspect was apprehended at the scene of thecrime, the files did not contain a description of his attire or disguise and nocase files contained any information relevant to assessing the accuracy ofweapon description. Thus, the assessment of the accuracy of eyewitnessdescriptions is confmed to the above elements of physical appearance.For the statistics of height and weight, a signed difference score wascalculated by subtracting the measured value of the charged suspect from the58estimated value provided by the eyewitness. The accuracy of age estimateswas calculated by determining the suspect’s age at the time of the crime andthen subtracting that age from the eyewitness’ estimate. A negative scorereflects an underestimate and a positive score reflects an overestimate. Recallof hair color and facial hair was scored as accurate if it was consistent withwhat was recorded at the time of arrest. Hair color was conservatively scoredso that even a difference between the eyewitness and police descriptions thatcould reflect different categories of hair color were scored as incorrect. Forexample, if the police recorded hair color as blonde and the eyewitnessdescribed it as mousy brown, the eyewitness’ description was scored incorrect.With the exception of height, it must be acknowledged that a perpetrator’sappearance could change quite dramatically between the crime and hissubsequent arrest. He could put on or lose weight, dye his hair, and grow orshave facial hair. It was impossible to determine if such changes had occurred,and as such, one will have to take the results on the accuracy of eyewitnessdescriptions with a grain of salt, the size of which will no doubt be individuallydeterminedEyewitness Identification AttemptsThe data collected about eyewitness identifications included whether ornot an eyewitness was asked to identify a suspect, the type of identification task(e.g. photo spread, live lineup, chance encounter, etc.), the date of the attempt,the outcome of the identification attempt and the eyewitnesses pre- and post59identification confidence levels. It was also noted if the eyewitness viewed anymug shots or helped a police artist make a composite picture of the perpetratorprior to the identification attempt. Virtually all identification attempts weremade with lineups, so the cumbersome phrase ‘identification task’ is replacedwith ‘lineup.’The consideration of the accuracy of eyewitness identifications faced twoobstacles. The first was to determine whether the lineup contained a photo ofthe perpetrator. This obstacle was overcome by classifying cases according tothe degree of certainty that the lineup did indeed contain the guilty party.Police files generally do not include court decisions as to the guilt or innocenceof a suspect. Even if they did, that would not guarantee that the right personwas charged. However, some cases involved evidence that pointed directly tothe guilt or innocence of a particular person. In some cases, the suspectconfessed to his deeds and/or was apprehended at the scene of the crime. Inother cases there was evidence that implicated the suspect’s guilt or innocence.For example, possession of marked bills, or the identification (driver’s license)of the victim. Finger prints matching the suspect were frequently found at thescene of the crime or in an abandoned vehicle used during the commission ofthe crime. There was evidence in only one case which supported theinnocence of the suspect; a mismatch between finger prints found at the sceneof the crime and those of a suspect. Finally, many case files contained noevidence that supported the guilt or innocence of the suspect. These threeevidence conditions (confession/apprehension, implicating and none) reflect60degrees of certainty regarding the presence of the perpetrator in the lineup. Incases involving a confession or an on-the-scene-apprehension the likelihoodthat the police suspect in the lineup was the actual perpetrator is quite highalthough not absolute. Some confessions may not have been genuine, or somemay have been obtained under duress. In cases involving implicatingevidence, the likelihood that the police suspect in the lineup was the actualperpetrator is perhaps not quite as great as in cases involving a confession orimmediate apprehension, but nonetheless is still quite high. Finally, we canbe least certain of the presence of the perpetrator in lineups from cases inwhich there was no evidence. Some of the lineups in this category were acollection of photos of people known to commit a certain type of crime. Inthese types of “fishing expeditions” the police generally did not a have aparticular suspect in mind and the lineup, much like a mugshot inspection, wasan attempt to generate one.The second obstacle could not be overcome and thus served to limit theanalysis. The actual decision in identification attempts that did not result inselection of the police suspect could not be determined reliably. The police inthis detachment did not use a uniform reporting system for eyewitnessresponses to identification tasks. Identification outcomes were entered in thefiles in a variety of ways such as “negative results”, “unable to identify policesuspect”, no in the lineup”, “pointed out suspect and one other as looking likeperp”, “positive ID”, and “weak ID”. Sometimes it was clear from commentsmade by an eyewitness at the time of the identification attempt whether or not61they had rejected the photospread entirely (e.g. “there’s not even one close”) butfor the most part, no reliable distinction could be made between outcomes inwhich an eyewitness rejected the photospread or failed to select the policesuspect. Identification attempts in which an eyewitness selected only the policesuspect were coded as “positive.” All other outcomes had to be coded simply“negative” and these include false alarms and rejections. This recordingprocedure suggests that police do not consider the lack of an identificationdistinct from a misidentification and certainly not as informative as a positiveID, despite Wells and Lindsay’s (1980) argument suggesting thatnomdentifications can be just as informative as positive identifications.62RESULTSDescriptive InformationRobberies are considered first. There was a total of 122 perpetrators (M= 1.58 per case; Range: ito 6), 81 victims (M = 1.05 per case; Range: ito 2),and 83 witnesses (M = 1.07 per case; Range: 0 to 33). All perpetrators ofrobbery were male. Police descriptions of charged individuals indicated thatthe average age of robbery perpetrators was 20.87 years (range: 15 to 40 years).There were 35 female and 45 male victims of robbery1. The average age offemale and male victims was 36.81 and 35.85 years respectively. The age ofthe female victims ranged from 18 to 78 years; males ranged from 11 to 82years. There were 44 female and 28 male witnesses to robbery. The averageage of female and male witnesses was 29.51 and 27.18 years respectively.Their ages ranged from 11.5 to 47 years in the case of females and from 11.5 to58 years in the case of males.The majority of robberies (n 48; 62.3% of the sample) involved a singlevictim and no other witnesses. There were 28 (36.3%) cases in which therewas at least one witness. Finally, one case (1.3%) involved two victims and nowitnesses. The identity of the perpetrator was generally unknown toeyewitnesses of robbery. In only two cases (2.6%) were victims able to supplythe police with the identity of the perpetrator. Five robberies (6.5%)werecommitted by a perpetrator who seemed familiar to the eyewitnesses (5victims, and 1 witness). These eyewitnesses thought that the perpetrator lived63in the area or had been in the vicinity prior to the robbery. When witnesseswere present during a robbery, they were frequently acquainted with the victim.Witnesses were co-workers, friends, or family members of the victim in 16(57.14%) of the 28 robberies that had witnesses. These data on the typicalconfiguration of a robbery (number of eyewitnesses, relationships betweenparticipants) replicates that found in an archival study reported by Yuille(1986).In the 21 cases of fraud there were 44 perpetrators (M = 2.09 per case;Range 1 to 13) and 134 victims (M = 6.38; Range: 1 to 24). All but two of thefraud cases involved a single perpetrator. The average number of perpetratorsin the fraud cases was elevated by two cases in which police suspected thatseveral perpetrators were at work. I considered each of the incidents in thesetwo cases to have been perpetrated by a different individual (n = 12 and 13).The police charged an equal number of males and females with fraud (n=4 each). The average age of charged suspects was 26.88 years and rangedfrom 20 to 35 years. Most victims of fraud were female (90 female, 39 male)2.The average age of victims where recorded (n 9 female, 5 male) was 25.86years and ranged from 17 to 58 years.Unlike robbery, any potential eyewitnesses present during thecommission of a fraud (including the victim) are usually unaware that a crimeis taking place. Hence they do not have a reason to attend to the event orremain on the scene like witnesses to robbery. Once the crime has beendetected, it is difficult to locate eyewitnesses unless they work in the store or64place of business where the frauds most often occurred. The fraud files thatwere examined involved only victims. Most fraud victims within a case wereinvolved in separate instances, and it is assumed that within a case the victimswere strangers to one another. By virtue of the selection criteria, no victims offraud knew the identity of the perpetrator and unlike robbery, no fraud victimsreported feeling a sense of familiarity with the perpetrator.Amount of detail in eyewitness descriptions ofperpetratorsMost eyewitnesses to robbery were able to provide some informationabout the appearance of the perpetrator. However, this was not the case forfraud victims. Only eight robbery victims (9.88%) and eight witnesses (9.6%)were unable to describe the perpetrator(s). Almost three-quarters of fraudvictims (n = 97; 72.4%) offered no information regarding the appearance of theperpetrator. Whereas some fraud files contained a clear indication that thevictim was unable to describe the perpetrator (e.g. that the clerk didn’t evenremember the interaction), there generally was no consistent indication in thefiles regarding whether the victim could not describe the perpetrator or was notasked to.A closer look at the circumstances of the crimes in which eyewitnesses torobbery were unable to describe a perpetrator revealed that most often justbeing a victim was enough to render one unable to describe a perpetrator.None of these victims had been verbally threatened, none of the perpetratorswore a disguise, only two victims were involved in crimes committed with a65weapon and only two were involved in multi-perpetrator crimes. The scenariofor witnesses who were unable to describe the perpetrator was quite different,especially regarding weapon use and number of perpetrators. One had beenwitness to a crime involving a verbal threat of death or injury, three witnessed acrime in which the pepetrator wore a disguise, all but two were involved incrimes committed with a weapon, and all but one were involved in multi-perpetrator crimes.Because some robbery cases [n = 27) involved multiple perpetrators,there were a total of 116 possible descriptions from victims and 100 fromwitnesses of robbery. One-way ANOVA’s indicated a significant effect of typeof eyewitness on the total amount recalled, amount of clothing and physicalappearance details [all F’s (2,349)> 54; p’s <.0001]. Scheffe tests at p = .05were conducted. Fraud victims provided an overall average of 2.11 detailswhich was significantly less than the 10.96 provided by victims of robbery andthe 9.37 offered by robbery witnesses. Fraud victims also offered fewer detailsregarding the clothing of the perpetrator (0.34) than victims (4.03) andwitnesses (4.35) of robbery. Victims of robbery offered more detail regardingthe physical appearance of the perpetrator (6.9) than both robbery witnesses(5.02) and fraud victims (1.76).The previous analysis included eyewitnesses who were unable or werenot asked to provide a description of the perpetrator as well as eyewitnesseswho provided descriptions so scant that they could not have been very useful. Adescription of only two details regarding physical appearance (for example, that66the suspect is a white male) does not radically reduce the population ofsuspects and thus does not offer much helpful information to the police. Adescription of four units regarding physical appearance (for example, thesuspect is a white male with brown hair and is about 6’4” tall) is a bit moreuseful. The above analysis was repeated on a reduced sample of eyewitnesseswho provided a minimum of four details regarding the physical appearance ofthe perpetrator. Table 1 presents these results.In this sample, victims (n=73) and witnesses of robbery (n = 48) offeredan average total of 14.71 and 14.79 details respectively and fraud victims (n =28) offered an average total of 9.11 details. One-way ANOVA’s indicated asignificant effect of type of eyewitness on the total amount recalled, amount ofclothing and amount of physical appearance details. [all F’s (2,146) > 5.80; p’s<.0 1J Post-hoc multiple comparisons (Scheffe at p = .05) indicated that victimsand witnesses of robbery offered more total details than fraud victims. Victimsand witnesses of robbery also provided more clothing details(4.94 and 6.41respectively) than fraud victims (1.54). Finally, Victims of robbery offeredmore details (9.76) regarding the physical appearance of the perpetrator thanfraud victims (7.57).Accuracy of eyewitness descriptions of perpetratorsIn cases where a suspect was charged, his/her age, height, weight, hairand eye color were recorded by the police. This permitted a comparison67Table 1. Mean number of descriptive details provided by eyewitnesses whoprovided at least four details regarding the suspect physical appearance.Type of EyewitnessType of Detail Robbery Robbery FraudVictims Witnesses Victims(n=73) (n=48) (n=28)ClothingM 4.94 6.41 1.53SD. 3.91 4.30 2.28Physical AppearanceM 9.76 8.38 7.57S.D. 3.54 3.21 1.71TotalM 14.71 14.79 9.115.94 5.92 3.1768between the eyewitness descriptions of these attributes and the measured valuesobtained from the charged suspect(s). Too few (n=3) eyewitnesses offeredinformation regarding eye color of the suspect to draw any conclusionsregarding accuracy. The accuracy of fraud victims’ descriptions could not beassessed as there were too few descriptions offerecLNo eyewitnesses (fraud and robbery) reported a different sex or race ofthe perpetrator than was recorded by police at the time of arrest. Because thesample of charged suspects might have contained some innocent suspects, onlythose cases in which the suspect confessed his or her guilt are considered. Thisresulted in the exclusion of only 25 out of 144 estimates of age, height orweight from the following analyses. The police laid charges in 23% of robberycases (n = 18 cases involving a total of 27 perpetrators) and 38% of fraud cases(n = 8 cases involving a total of 8 perpetrators). Of the cases where chargeswere laid, 67% of robbery cases (n =12 cases involving a total of 20perpetrators) and 50% of fraud cases (n = 4 cases involving a total of 4perpetrators) saw a confession offered. Too few fraud victims gavedescriptions to analyze the accuracy of their estimates in either the samplebased on suspects who confessed or in the sample based on all charged suspect.Table 2 depicts the accuracy of age, height and weight estimates provided byvictims and witnesses of robbery in cases in which the suspect confessed.The analysis of description accuracy included dependent-groups t-tests for thedifference scores’ deviation from zero, as well as independent-groups t-tests fordifferences between victims and witnesses of robbery. The results indicate that69Table 2. Mean difference scores for robbery eyewitnesstestimates of age,height, and weight of perpetrators who confessedType of eyewitnessType of Estimate Victim WitnessAgeM 2.87 3.40Md 2.75 2.25Miii -1.00 -4.0010.00 10.00S.D. 3.48 4.05n 12 24HeightM -1.90 -.476Md -1.50 --.450Mm -5.50 -6.40Max 0.00 3.00SD 1.56 2.27n 11 38WeightM -7.56 -4.78-5.13 -11.80Miii -31.60 -36.40Max 25.40 25.40S.D. 14.93 14.35n 10 24Note: Difference scores were calculated by subtracting the actual value fromthe estimate provided by the eyewitness. Negative values representunderestimates.Age = years; height = inches; weight = poundsMd = median; S.D. = standard deviation70both victims and witnesses overestimated age while only victimsunderestimated height. Victims’ age estimates were over by an average of 2.87years, t(l 1) = 2.86, two-tail p < .05, and witnesses’ by an average of 3.40 years,t(23) = 4.11, two-tail p <.05. The two groups did not differ from one another intheir degree of overestimation, (34) = -0.41, two-tail p> .05. Victims’ heightestimates fell short by an average of 1.9 inches, t(10) = -4.04, two-tail p < .05,which was significantly lower than their witness counterparts who were shortby only 0.476 inches, t(37) = -2.38, two-tail p < .05. Neither victims’ (M = -7.56 pounds) nor witnesses’ (M -4.78 pounds) underestimates of weight wereslim enough to be significant, 1(9) = -1.6 and 1(23) = -1.63, respectively, bothtwo-tail p’s> .05, nor were the two groups different from each other, t(32) = -.5, two-tail p> .05.Finally, in order to be able to compare the accuracy of estimates providedby these eyewitnesses with that reported by Cutshall (1985) estimates wereconsidered accurate if they were within 2 years, 2 inches or five pounds of theactual value. In the sample of perpetrators who confessed, 45.4% of theestimates were accurate, and in the whole sample, 47.2% were accurate.Two z-tests on proportions were conducted for descriptions of hair colorand facial hair. Hair color was consistent with what was recorded at the timeof arrest in 38.46% of victim’s and 48.28% of witnesses’ descriptions. Theseproportions do not significantly differ. Victims’ and witnesses did differsignificantly (z=-2.25) in description of facial hair. Sixty percent of victims71and 100% of witnesses’ descriptions of facial hair were consistent with whatwas recorded at the time of arrest.Identification ProceduresThere were a total of 170 identification attempts. Of these, 90.58% (n =154) were with photospreads. It appears live lmeups have fallen out of favordue the cost and the difficulty in constructmg a fair one, particularly forsuspects from ethnic minorities. Only 10 identification attempts were madewith traditional physical or live lineups. Four identification attempts were onesin which the police either brought the suspect to the victim, or the victim to thesuspect (i.e., showups). One identification attempt consisted of a fraud victimviewing a videotape of the bank activities on the day of the crime. Finally, inone case a victim of robbery encountered her assailant while out shopping andalerted the police.I could not look at most of the lineups shown in the cases analyzed in thisstudy because they were generally sealed in evidence envelopes. However, Iwas able to view some. Most photospreads shown at this detachment containedeight photographs. The photo arrays I viewed were well constructed; in allcases, the foils resembled the suspect (e.g similar hair color and length, race,facial hair, facial shape etc.) and there was nothing in the construction of thearray that suggested the identity of the police suspect. One officer was wellaware of the tactics employed by defense lawyers to discredit an eyewitness72identification. In anticipation, therefore, he first showed his imeups to otherofficers who knew nothing about the case and asked them to select the suspectand to look for bias in his imeups. This is precisely the procedure used byWells et al. (1979) to assess the fairness of a lineup and is generally viewed asa reasonable way to avoid some problems of using lineups.Frequency of Identification AttemptsAlmost 40% (n=30) ofall robbery cases and two-thirds (n=14) of fraudcases included an identification attempt. The police suspect was identified in16 cases of robbery and 10 cases of fraud. Thus, in this sample, the police hada positive eyewitness identification in 20.8% of all robberies and 47.6% of allfrauds.Identification OutcomesTable 3 presents the proportion ofpositive identifications by eyewitnesstype and evidence condidtion. Three identifcation attempts that were made ona lineup known to not contain the perpetrator (he confessed after a lineup notcontaining his photo was shown) were the only ones made on a confirmedtarget-absent lineup. All three eyewitnesses did not select the (innocent) policesuspect. These three identification attempts were not included in the followinganalyses as they were known target-absent lineups and the rest were eitherknown target-present or possibly target-present. The police suspect wasselected in just less than half of all identification attempts by victims of73Table 3. Percentage of positive identification and average delay (days)between exposure to the perpetrator and identification attempts by type ofevidence and eyewitnessType of EvidenceType of Total rowEyewitness None Implicating Confession averagesRobbery victimsPercentagePosiD 21.7 57.1 84.6 46.5n (23) (7) (13)Avg. Delay 43.22 31.71 9.38 31.11n (23) (7) (13)Robbery witnessesPercentage Pos ID 11.0 55.5 33.3n (9) (9)Avg. Delay 84.11 4.67 44.39n (9) (9)Fraud victimsPercentage Pos ID 16.7 38.9 22.7 25.47n (48) (36) (22)Avg. Delay 121.78 105.94 16.0 108.10n (32) (32) (4)Total columnaveragesPercentage pos ID’s 17.5 41.9 47.7Average delay 88.25 92.6 8.77robbery, one-third by witnesses to robbery and in one-quarter of attempts byfraud victims. Identification accuracy of perpetrators who confessed or whose74guilt was supported by implicating evidence was much higher than forperpetrators for whom there was no evidence. A chi-square test of associationacross evidence conditions revealed a significant association [x2 (2, n = 167)6.25; p <.05] between the three types of eyewitness and identificationoutcomes. Post hoc multiple comparisons (Marascuilo, 1966) did not revealany significant differences between eyewitness types.Considering only cases in which a suspect confessed, a second chi-squaretest again revealed a significant association [x2 (2, n = 44) = 12.82; p<.05]between type of eyewitness and identification outcome. The police suspectwas selected in 84.6% of identification attempts made by victims of robbery,55.5% of attempts made by witnesses to robbery, and in 22.7% of attemptsmade by fraud victims. Subsequent multiple comparisons demonstrated asignificant difference between victims of robbery and fraud [x2 {2, n = 17}5.99; p < .05].The delay between exposure to the perpetrator and subsequentidentification could be determined for all robbery eyewitnesses, but for lessthan two-thirds of fraud victims. Bearing this in mind, victims of robbery mayowe their superior performance to the fact that on average, they waited only31.11 days before attempting an identification. Robbery witnesses waited anaverage of 44.39 days and fraud victims waited for 107.87 days. Additionally,the delay between exposure and identification was shortest in cases involving aconfession ( M = 8.77 days) and longest in cases in which there was noevidence ( M = 88.25 days).75Table 4. Range and average delay between exposure to the perpetrator andidentification attempt in cases of robbery and fraud.Range Average PercentPositive D’sRobbery0-1 14 0.5 71.433-5 15 3.6 46.677-34 21 18.9 33.3338-191 14 120.21 14.29Fraud7-62 18 32.94 77.770-90 18 74.4 5.55107-154 15 131.67 20.0170-382 17 200.42 17.6576One of the clearest findings to emerge from this analysis is that the percentageof positive identifications dropped dramatically with time. Table 4 presentsthis information. Particularly in robberies, the longer an eyewitness waits toattempt an identification, the less likely s/he is to select the police suspect. Asecond striking feature of this table is that eyewitnesses to robbery and fraudwho faced the shortest delays made roughly the same high proportion ofpositive identifications despite a difference in average delay of over a month.One possible factor that might account for this is the difference in the extent towhich eyewitnesses to robbery and fraud interact with the perpetrator. Therobberies in this sample were brief and the eyewitnesses may not have beenexposed to the perpetrator for very long. The frauds in this sample weregenerally brief as well, but there were exceptions. Thirty-two fraud victims(3 5.9%) had a relatively extensive interaction with the perpetrator such asselling him items valued at over $1000 (e.g. furniture, a car, stereo equipment),handling a suspect’s fraudulent application for welfare benefits or instructing asuspect on the use of automated teller machines. These interactions were moreinvolved and likely lasted longer than the routine transactions thatcharacterized the majority of the fraud interactions. Seventy-four percent offraud victims who selected the police suspect had an extensive interaction withthe perpetrator whereas only 19.3% of fraud victims who did not select thepolice suspect had such an interaction with the perpetrator.77Presence of a WeaponThe analysis of weapon effects includes only robbery cases. Of the 77 robberycases, a total of 43 (55.84%) were committed with an actual (38 cases) orimplied weapon (five cases). Table 5 reports the average amount ofinformation recalled as a function of presence or absence of a weapon.Separate 2 X 2 ANOVA’s (victim versus witness by weapon versus no weapon)were conducted for each of the three categories of details (clothing, physicalappearance, and total). Despite the fact that we should expect victims to sufferthe most from the weapon-focus phenomenon, no interaction between weaponpresence and eyewitness role was found in any of the three analyses. Therewere, however, significant main effects for weapon presence and eyewitnessrole.Eyewitnesses in crimes involving a weapon provided an average total of11.29 details which was significantly higher than the average total of 8.3provided by eyewitnesses to weaponless crimes [F(1,212) 13.19; p <.000 1].This pattern or greater detail provided by eyewitnesses in crimes committedwith a weapon was repeated in analyses of clothing [F (1, 212) = 15.07; p<.00011 and physical appearance details [F (1, 212) = 5.42; p < .01]. Overall,victims proved significantly more total details (M = 10.97) than witnesses [(M= 9.37), F (1, 212) 7.11; p <.01) as well as more physical appearancedetails[(M’s 6.93 and 5.02, respectively), F (1, 212) = 14.04; p < .0001].78Table 5. Average number of details robbery eyewitnesses provided by weaponpresence and absence.Weapon used No Weapon usedType of Detail Victim Witness Victim WitnessClothing 5.07 4.89 2.96 2.20SD 4.27 4.51 3.16 2.89Physical Appearance 7.92 5.16 5.91 4.454.77 4.10 4.52 397Total 1298 10.95 8.88 6.65S.D 762 7.44 6.29 6.37n 59 80 57 2079The police charged a suspect in 14 cases committed with a weapon and in6 weaponless cases. The sample of perpetrators who confessed was too smallto examine weapon presence effects in these cases alone so these results arebased on all suspects charged by police. The tendency to overestimate age andto underestimate height remains, regardless of the presence or absence of aweapon. Separate 2 X 2 ANOVA’s (victim versus witness by weapon versus noweapon) were conducted for each of the three estimates (age, height andweight). There were no interactions and no significant main effects of eitherweapon presence or eyewitness role. There was a marginally significantdifference between the average errors in height estimates provided by ofvictims (M = -1.72) and witnesses [(M = -0.48), F(1, 55) = 3.24, p <0.10].Less than a quarter (n=8, 23%) of weaponless cases and over half ofcrimes committed with a weapon (n=22, 51%) resulted in an identificationattempt. Only 30.6 1% of identification attempts made by eyewitnesses tocrimes committed with a weapon resulted in selection of the police suspect. Inweaponless crimes, 73.33% of identification attempts resulted in selection ofthe police suspect. However, eyewitnesses in weaponless crimes made theiridentification attempts after a much shorter period of time than eyewitnesses incrimes committed with a weapon (M = 7.33 versus 41.47 days).All but one identification attempt in weaponless crimes were made afteronly 17 days. Almost half (n = 22; 44.8%) of identification attempts in crimescommitted with a weapon were made at delays greater than 17 days. I tried tocontrol for the different delays between weapon conditions by examining only80identification attempts made at delays of 17 days or less. In this analysis, theaverage delays were 5.15 and 4.5 days for crimes committed with and without aweapon. Even with similar delays between exposure and identificationattempt, eyewitnesses to crimes committed with a weapon were less likely tochoose the police suspect than eyewitnesses to weaponless crimes. Positiveidentifications were made in 44.44% of identification attempts by eyewitnessesto crimes mvolvmg a weapon and in 71.42% of identification attempts byeyewitnesses to weaponless crimes. An analysis of covariance withidentification outcome as the dependent measure, presence or absence of aweapon as a factor, and delay as covariate demonstrated a marginallysignificant effect of weapon [F(1,126) = 3.578; p =.061].Summary of ResultsVictims and witnesses of robbeiy offered more details than fraud victimsin their descriptions of the perpetrator. The accuracy of descriptions andidentifications was assessed according to the type of evidence (Evidencecondition) against a police suspect. Confession Evidence represents thehighest likelihood that the police suspect is the actual perpetrator, ImplicatingEvidence represents a lower likelihood and No evidence represent the lowestlikelihoocL Accuracy of descriptions provided by robbery eyewitnesses wasanalyzed only in the Confession Evidence condition. No eyewitness reported adifferent sex or race of the perpetrator than was recorded at the time of arrest.Victims and witnesses of robbery both overestimated the age of the perpetrator,81but did not differ from each other in their overestimations. Victims of robbery,but not witnesses underestimated the height of the perpetTator and victim’sestimates were less accurate than witnesses. There was no tendency to over orunderestimate weight and victim’s and witness’s estimates did not differ fromeach other.Across evidence conditions there was a significant association ofeyewitness type and identification outcome. Victims of robbery were mostlikely to identify the police suspect and fraud victims were least likely.) butthere were no significant differences between eyewitness types. In just theConfession Evidence condition, there again was a significant association ofeyewitness type and identification outcome and in this analysis robbery victimswere significantly likely to select the police suspect than fraud victims. Delaywas a confound in theses analyses as robbery victims waited the shortestamount of time to attempt their identifications and fraud victims waited thelongest.There was no adverse influence of weapon presence (robbery cases only)on either the amount of detail in descriptions of the perpetrators or the accuracyof those descriptions. In fact, eyewitnesses in crimes involving a weaponprovided more detailed descriptions than eyewitnesses in weaponless crimes.There was suggestive, but not conclusive evidence of an adverse influence onweapon presence.82DISCUSSIONThe discussion begins with comments on how specific fmdings comparewith those from other types of research and how this in turn impacts ongeneralizabiity. Suggestions for future research accompany comments onspecific findings. The discussion concludes with a reiteration of the need forthe multi-method approach.The analysis of the amount of descriptive detail provided by all robberyeyewitnesses revealed results that were similar to those obtained by Kuehn(1974) and Sporer (in press). Regarding just physical appearance (excludingclothing), the victims in Kuehn’s sample recalled an average of 7.2 details andvictims in this study recalled an average of 6.9 details. In terms of total persondescription (clothing and physical appearance), victims in Sporer’s studyrecalled an average of 13.17 details, and victims in this study recalled anaverage of 10.96 details. However, since Sporer’s sample was drawn fromcases in which an identification attempt was made, it is unlikely that his samplecontained victims who were unable to provide a useful description. Thecomparison between the 14.71 details provided by victims in this study whooffered descriptions involving four or more details and Sporer’s 13.17 detailsmight be more appropriate. Finally, the “incidental witnesses” in Sporer’sstudy were possibly in roles similar to the fraud victims. These incidentalwitnesses did not see the crime but were questioned by the police about theappearance of the perpetrator whom they had observed on a different, andpossibly less arousing occasion. Unfortunately, the circumstances of the83incidental witnesses encounter with the perpetrator such as how arousing theinteraction was or the length of time that passed between seeing the perpetratorand talking to the police about it could not be determined from Sporer (inpress). With this grain of salt in mind, Sporer’s incidental witnesses offered anaverage of 7.79 details in their descriptions of the perpetrator and fraud victimsin the present study offered 2.11 details. If only those who offered four ormore details are considered, fraud victims averaged 7.11 details in theirdescriptions.These similarities in the amount of descriptive details are striking,especially given that the eyewitnesses in Kuehn and Sporer’s studies weredrawn from a variety of crime types including assault and rape which may bemore arousing than robbery, at least for the victims. Furthermore, each of thesethree archival studies has been conducted in a different country; Kuehn in theU.S., Sporer in West Germany and the present study was conducted in Canada.One puzzling difference is the higher percentage of victims who were unable todescribe the perpetrator in the present study (9.88%) compared with Kuehn’s(4%). Sporer (in press) did not report these figures. Kuehn’s study was basedon more violent crimes, but did include robbery (armed only). While there arefactors other than memory per se that could affect the number ofvictims whowere unable to provide a description, such as how insistent the police were inobtaining a description or how willing the victim was to cooperate with theinvestigation, the issue may be simpler than that. Although not expresslyreported in either Kuehn (1974) or the present study, it can be reasonably84inferred that the methods by which the police collected descriptions differedbetween the two studies. The police in Kuehn’s study likely employed astandard form to collect their statements. If this form was given to the victimsto fill out, or the police asked questions from it, then it may have served as akind of cued recall test. Descriptions in the present study were almostguaranteed to have been obtained by free recall. Within the field of eyewitnessmemory research and beyond, in the field of cognitive psychology, there havebeen many demonstrations of the quantitative advantage of cued recall overfree recall (e.g.Lipton, 1977; Whipple, 1909). Had the victims in this studybeen faced with a form listing features as opposed to an open-ended question,then the number ofvictims who were unable to describe the perpetrator mighthave been equivalent between the two studies. As the type of recall taskemployed was hypothesized to account for differences in a variety fmdingsbetween the present study and others, future case or archival research shouldinvolve a concerted effort to determine the manner in which eyewitnessstatements were collected.Although eyewitnesses tended to describe the perpetrator as older andshorter than he actually was, the values of these over and under estimates wereslight; about 3 years and 1 inch. The field study of Flin et al (1986) alsoreported a tendency to underestimate height. The percentage of correctestimates of age, height and weight offered by eyewitnesses in Cutshall’s (1985)case study and the present study were similar; 50% and 47.2% respectively.Although these figures are not too impressive, it must be considered that they85reflect not only memory, but estimation as well. It is possible that estimates ofage, height and weight made by some eyewitnesses while the suspect is presentwould be about as accurate as those made from memory.Less than one-half of eyewitness recollections of hair color wereconsistent with what was recorded at the time of arrest. This could be anartifact of a very stringent scoring system, a change in hair color between thecrime and arrest, or as Cutshall (1985) reported, it could reflect the poor colormemory of eyewitnesses. Recollections of facial hair fared better, with mosteyewitness’ recollections consistent with what was recorded at the time ofarrest.Most discussions of the negative effects of biased lineups have centeredon lineups which do not contain the actual perpetrator and the possibility ofdrastic consequences for an innocent person who is falsely identified underthese circumstances. If the perpetrator is present in a biased lineup, the biasgenerally acts to secure his/her positive identification. However, due to the factthat most defense lawyers are well-versed in the effects of biased lineups, andcan easily discredit an identification made with such lmeups, bias in targetpresent lineups has drastic real-world consequences for the prosecution of thecase. The officers in the detachment employed in the present study hadconsidered the legal consequences of biased lineups and had taken steps toeliminate suggestion in the construction and administration of lineups.There is a well established literature describing case studies in whichwrongful conviction and imprisonment resulted from eyewitness86misidentifications (e.g., Brandon & Davies, 1973: Borchard, 1932; Wall, 1965).However, the present project represents the first large-scale analysis ofeyewitness identification in actual police cases. Perhaps the simplest featureof the identification data concerns the question raised by Konecni and Ebbesen(1986) regarding the frequency of eyewitness identification evidence in realcases. They estimated that only a very small proportion of cases that made it tocourt concerned the identification of the offender and suggested that the effortsof eyewitness researchers are therefore out of proportion to the actual roleidentification evidence plays in real cases. Goldstein, Chance, and Sneller(1989) have pointed out that even a conservative estimate that only 3% of casesinvolve identification of the offender results in approximately 77,000 suchcases in a typical year in the U.S. In the present study the police had a positiveeyewitness identification in much higher than3% of both robbery and fraud(20.8% of robberies and 47.6% of frauds). Presumably, these cases would beprosecuted, but the weight of the eyewitness identifications remains unknownbecause the police records seldom include information regarding the fate of acase once it is turned over to the prosecution. It is very possible that a positiveidentification leads to the collection of more evidence that carries the case incourt, or that the accused is encouraged to plea bargain and the case is closedat that point. Either way, this admittedly small sample suggests thatidentification is an important issue in many cases and that continued research iswarranted.87It is impossible to consider identification accuracy in the present studyoutside the context of delay. Identifications in the confession evidencecondition and by robbery victims were numerically the most accurate, and werealso made at the shortest delays. Not surprisingly then, the most impressivefigure in the identification data is the 85% accuracy ofvictims of robbery in theconfession evidence condition who on average, viewed a lineup only nine daysafter the crime. In the same evidence category robbery witnesses, whoaveraged viewing a lineup almost five days sooner than victims, were accurate55% of the time and fraud victims were accurate only 23% of the time, but itwas impossible to determine the delay for over 80% of these fraudidentifications. The robbery victims in this study were aroused, had particularreason to attend to the perpetrator and during the crime, or shortly thereafter,they may have even thought that they would have to describe and perhapsidentify him. Finally, they may have thought about or discussed the incidentseveral times afterwards. In contrast, the fraud victims were not likely to havebeen aroused, and other than completing the routine transaction, probably hadno particular reason to attend to the perpetrator, think about or discuss him later(unless he was the object of flirtation, or had some peculiar feature about himsuch as extreme height, big ears, or a snakeskin suit). If any of thesedistinctions, particularly arousal, are responsible for the different identificationperformance of robbery and fraud victims, then the poorest context in which tostudy eyewitness memory is the laboratory. There are circumstances to whichmuch of the laboratory research could be applied. For example, one could88witness two suspicious persons laden down with stereo equipment leaving avacationing neighbor’s house. In this circumstance, there would be littlewitness involvement with the perpetrator, the witness would know to payattention, and the event would be shocking, but likely only mildly arousing.These types of situations however, are not the type referred to when laboratoryresearch is being applied to actual eyewitnesses. Laboratory research is beingapplied from the (expert) witness box to eyewitnesses of rape, assault, murderand robbery, and it is being generalized to generic eyewitnesss of actual crimesby many forensic researchers.Despite similar lengthy delays, eyewitnesses in the no evidence conditionwere much less successful in selecting the police suspect than were those in theimplicating evidence condition. This is possibly due to a higher proportion oftarget-absent lineups in the no evidence condition. As a note in passing, theissue of false negatives, or failing to identify a suspect when s/he is in thelineup, has been given the back seat to the other kind of error, false positives.Most discussions and a great deal of expert testimony on eyewitnessidentification focuses of the fallibility of eyewitness memory and the weightyrisks and consequences of an innocent suspect being fingered by an erranteyewitness. It is unfortunate that this research could not shed light on the entirepicture of eyewitness identification, hits, false positives, false negatives, andcorrect rejections.Throughout the present research, the unit of analysis has been at the levelof eyewitnesses. From a police perspective, another informative unit of89analysis is at the case level. That is, in how many cases of robbery or cases offraud is there a suspect identification, as opposed to how many robbery andfraud eyewitnesses identify the suspect. As it turns out, this case/eyewitnessdistinction bears on crime solution rates. A case-wise look at overallidentification accuracy shows that slightly more than half of the robberies, butmore than two-thirds of the fraud cases in which an identification wasattempted involved a positive identification. The perpetrators of fraud usuallyhit more than one business and thus although the chances of any one eyewitnesscorrectly identifying the suspect was low, there were more opportunities forthese perpetrators to be identified.The higher arousal of eyewitnesses to actual violent crimes is generallyconsidered to be beyond the optimum level described in the Yerkes-Dodsonlaw and Easterbrooks (1959) cue utilization theory. Thus, in generalizinglaboratory findings most researchers have predicted that eyewitnesses in actualforensic contexts should perform worse than in the laboratory. However, thereis little in the present study to suggest that stress has deleterious effects oneyewitness recall. Throughout the present research it has been assumed thatvictims ofrobbery would be most aroused, witnesses to robbery would be lessaroused and fraud victims would be least aroused. A second way ofconceptualizing arousal considered robbery eyewitnesses only and assumed thateyewitnesses in armed robberies would be more aroused than eyewitnesses inunarmed crimes. Robbery eyewitnesses recalled more and were more likely toselect the police suspect than were fraud victims. This difference likely90reflects a combination of the low arousal of fraud victims, plus their lack ofknowledge that a crime was taking place and that they might have to rememberthe “customer” and a delay of several days before they were asked to describethe perpetrator. There was no difference between victims and witnesses ofrobbery in the amount of descriptive detail offered and equal numbers ofvictims and witnesses of robbery could not describe the perpetrator.With respect to the accuracy of descriptions, all eyewitnesses in thepresent study, including the few fraud victims who offered a description,presented the same pattern, albeit not significant in all comparisons, ofoverestimating age and underestimating height and weight. Victims andwitnesses of robbery did not differ from each other in their estimates of age andweight. The only significant difference between accuracy of victim’s andwitness’ estimates was on height, in which witnesses were closer to the actualvalue. It is interesting to note that although not significant, witnesses alsoprovided more accurate estimates of weight, and victim’s estimates of age weremore accurate. Actual proximity to the victim could not be coded in thisstudy, but it is possible that victims were closer to the perpetrator thanwitnesses. It is possible also, that this proximity offered them a better view ofthe perpetrator’s face from which to gage his age, but hindered or warped theirview of the perpetrator’s body.The second way of conceptualizing arousal lead to the same resultsregarding eyewitness recall. In contrast with laboratory research, the presenceof a weapon did not have a detrimental influence on the amount or accuracy of91descriptive information provided by actual eyewitness. In fact, weaponpresence lead to more detailed descriptions which is incompatible with aweapon “focus” phenomenon. The archival studies of both Sporer and Kuehnalso failed to fmd any evidence of a weapon focus phenomenon on eyewitnessrecall. This lack of replication of laboratory findings could be due to a numberof differences between the contexts of actual forensic and laboratoryeyewitnesss. Most laboratory studies of the weapon focus effect haveemployed slide sequences (only Maass and Kohiiken (1989) employed a liveevent). In addition, all of the studies employed some sort of a cued recall testof eyewitness memory that varied from a multiple choice test (Loftus, Loftus &Messo, 1987) to a set of open-ended questions (Maass & Kohnken, 1989). Incontrast, the eyewitnesses in this study usually provided a free recall and mayhave answered some open-ended questions. In this unstructured atmospherethe actual eyewitnesses often offered information that many of the laboratorystudies did not seek such as the complexion, odor or gait of the perpetrator.Furthermore, the accuracy that the two types of research examined aredifferent. In laboratory studies accuracy scores are generally a compound ofattributes such as hair color, age, and clothing description. The accuracyscores in the present research are for separate estimates of age, height andweight. Finally, the greater amount of descriptive information provided byeyewitnesses in crimes committed with a weapon might result from a morevigorous pursuit of descriptions from eyewitnesses, particularly victims, inthese cases due to their more serious nature. These differences highlight some92ways in which future laboratory research on the effects of weapon presencecould be improved to make comparisons with the effects of weapon presenceon actual robbery eyewitnesses.Unlike recall, arousal did influence eyewitness identification. Acrossevidence conditions, there was a significant association between the threeeyewitness types and identification accuracy. Victims of robbery were mostlikely to select the police suspect and fraud victims were least likely to do so,but there were no significant differences between eyewitness types. In themore restricted sample of confession evidence alone, robbery victims weremore likely to select the police suspect than fraud victims. These resultssuggest an enhancing effect of arousal. However, delay acts as a confound inthese results that favors high arousal. When arousal was conceptualized byweapon presence and delay controlled for in an analysis of covanance, arousalhad a detrimental effect on identification accuracy. The negative effect ofweapon presence on identification seems relatively robust. It has beendemonstrated in the lab where the level of arousal is low, the target person wasusually presented via slides and recognition was tested after very short delays(immediately to 20 minutes). It also was demonstrated in actual robberyeyewitnesses where the level of arousal was higher, the target person was veryreal and may have been quite close, and recognition was tested at roughly aweek to over a month later.Eyewitness researchers have assumed that the arousal experienced byactual eyewitnesses is beyond the optimum described by Yerkes and Dodson93and Easterbrook and that their performance would suffer. These result showthat arousal, at least at the levels, or in the contexts studied, differentiallyaffected recall and recognition. Three, non-mutually exclusive avenues areexplored with regards to why arousal influenced recognition but not recall: 1]assumptions about arousal levels; 2] cognitive effects of arousal; and 3] socialfactors that differ between recall and recognition.The recall evidence could be a result of three different possibilitiesregarding the inverted U-shaped curve of arousal and performance. The samethree possibilities exist if arousal is measured by weapon presence vs. absenceinstead of victim vs. witness. First, the assumption that victims of robberywould be more aroused than witnesses of robbery may have been invalid.While the victim and witness(es) to a particular crime may occasionally beequally aroused, particularly if the witness is a spouse or other family memberof the victim, it is unlikely that on average, victims and witnesses of robberyare equally aroused. Second, because there were no differences betweenvictims and witnesses of robbery, it is impossible to tell where they are on theinverted U-shaped curve. Fraud victims would be placed somewhere near thebottom of the left side of the curve, but robbery eyewitnesses could both be onthe ascending side above fraud victims, at the plateau, or on upper part of thedescending side. This might be partly due to the levels of arousal examinedwhich amount to “none” (fraud), “aroused” (robbery witnesses) and “probablymore aroused” (robbery victims), but still, if being a victim of a robbery is notsufficiently arousing to put one over the top and down the right side of that94inverted U-shaped curve then what type of crime is? Kuehn (1974) found thatvictims of robbeiy recalled more details than victims of assault or rape,implying that victims of these crimes might be over the top of the curve. Thethird possibility is that in more complex real life situations the applicability ofthe curve depends on the memory system being evaluated.In terms of the cognitive avenues to explore, the different delays betweenthe crime and recall the crime and recognition may have consequences.Because statements are taken quite soon after the crime, eyewitness recallcould possibly benefit from state dependent memory, an advantage likely notpresent at identification because it usually takes place some time later whenthe emotional state of the eyewitness would assumedly be quite different fromthat experienced during the crime. Also the short delay between the crime andproviding a statement leaves little opportunity for the memory to decay, but theconsiderably longer average delay between the crime and identification doesleave room for memory loss to occur.Laboratory research has found detrimental effects of weapon presence onboth recall and recognition and has concluded that information about theappearance of the perpetrator is simply not adequately encoded. The presentrecall data demonstrate that eyewitnesses in crimes committed with a weaponwere able to encode information about the appearance of the weapon holderand suggests that the locus of the weapon effect might not be at the encodingstage.95One possible alternative loci of the effect of weapon presence is at themaintenance phase where eyewitnesses may rehearse the event by thinkingabout it and discussing it with others. If the crime was particularly traumaticfor the eyewitness, s/he may have flashbacks, much like a victim of PostTraumatic Stress Disorder. We know from Read et al (1989) that both thetiming of rehearsal and the extent to which the appearance of the perpetratorhas changed from study to test can influence the identification accuracy oflaboratory eyewitnesses. Specifically, a combination of rehearsal immediatelyafter the event followed by a lineup in which the appearance of the perpetratorhas changed leads to poorer identification performance. It is not hard toimagine that these two factors would predominate most eyewitnessidentification situations. Note, in cases where the perpetrator is caught at orvery near the scene of the crime and an identification is attempted right away,the appearance of the perpetrator would be close to identical from exposure totest, which in addition to the brief delay, would help their identificationaccuracy.A second possible loci for the detrimental effect of weapon focus is at theretrieval phase. Perhaps eyewitnesses undergo a defensive reaction when theysee the perpetrator in a target-present lineup. In a study on the eyewitnessabilities of children, Peters (1991) videotaped children’s response from behindthe lineup. When the children would see the perpetrator they would “identify”him with body language (large eyes, shrinking away slightly) but would oftenpass on to the next lineup member without identifying the actual perpetrator.96Now, while this could reflect a dynamic of a child not feeling empoweredenough to “tattle on” an adult, it could also reflect a defensive reaction.According to Easterbrook (1959) “..it would be expected that, inperceptual tasks as in other tasks, the effect of emotion on proficiency woulddepend on the number of cues needed for adequate performance of the task.The deleterious effects of increased drive or stress would be expected to appearonly when the actual range of cue utilization fell below that required for thetask” (p. 189, italics mine). If recall involves more cues than recognition,then maybe it would suffer less from a reduction in cues than recognition. Ormaybe the attentional narrowing excludes cues that would be helpful only torecognition. The effects of attentional narrowing on identification accuracymight be mimicked in a laboratory study by presenting faces that had beenvisually degraded so that only some features were wholly visible or by askingsubjects to examine only a few features on a face. There is research whichdemonstrates that recognition accuracy is related to the number of facialfeatures attended to during study. Laboratory research (e.g. Baddely, &Woodhead, 1982; McKelvie, 1985; Sporer, 1991; Wells & Hryciw, 1984) ondepth of processing and memory for faces has demonstrated that deepprocessing when viewing a face such as making character judgments leads tobetter recognition than shallow processing such as making judgments aboutfacial features (e.g. wide or close-set eyes). Recently Bloom and Mudd (1991)demonstrated that deep processing involves inspection of more facial featuresthan shallow processing. Also, Loftus (1972) has reported that picture97recognition and number of features encoded are positively correlated. Ifattentional narrowing due to arousal leads to the inspection of fewer facialfeatures, then the influence of arousal on recognition could be at the encodingphase after all.Finally, there are other, social considerations such as when faced with alineup, some eyewitnesses may consider the consequences of misidentifying aninnocent suspect, or failing to identify the perpetrator. Also, eyewitnesses mayhave concerns about the future consequences of their identification decisionsuch as retribution and time spent in court. Future research efforts could bedirected at determining if actual eyewitnesses were indeed concerned aboutpossible future consequences about their identification decision or with makinga mistake and what kind of mistake they felt was most grave. At recall, it isless likely that these inhibitory factors would be present, although eyewitnessescould be reluctant to even get involved and elude their responsibility byclaiming they weren’t paying attention or that they don’t recall very much.This study replicated previous archival findings (Yuille, 1986) about theconfiguration of robberies in term of the high proportion of victim-only crimesand witness’ familiarity with the victims. This is an important replicationbecause much of the criticism of laboratory research has centered on theinadequacy of the role of urnnvolved bystander that most laboratoryeyewitnesses play. Any one or all of these configuration factors might affecteyewitness memory, mainly by operating on the level of arousal experienced byvictims and/or witnesses when present. For example, the presence of witnesses98might buffer victim’s anxiety, and/or it might curb or increase the perpetrator’saggression in conmiitting the crime. And, assuming the familiarity is a friendlyone, witnesses who know the victim might be more aroused than witnesses whodon’t, particularly in crimes where there is a threat to life and limb. Thesefindings make generalizations of laboratory-based research to actual victimsand witnesses of crimes difficult at best.Laboratory findings may not readily generalize to fraud either becausethe nature of the interaction is different. In contrast to laboratory eyewitnesses,the clerks are not aware that they will have cause to recall the perpetrator, thedelay between exposure and description is often greater, clerks are oftenpresented with the cheque or credit card receipt to jog their memory, and thereare often very long delays between exposure and identification. Also a veryhigh proportion of frauds do not involve identification or person description asthey are perpetrated between business associates. Some field studies however,particularly Read et at (1990) match many of the characteristics of frauds. Theidentification accuracy of fraud victims in the shortest delay period, presentedin Table 5, is compatible with that reported by Read et al who employedsimilar delays. Perpetrator descriptions were generally more detailed in Readet al than those provided by actual victims of fraud, but this could be due to thefact that the confederates in Read et al. engaged the clerk in rather unusualevents which were designed to be memorable.This study has demonstrated both the strengths and weaknesses ofarchival research. It was possible to study eyewitness memory in it’s entire and99real context and in some instances it was possible to comment on thegeneralizability of laboratory and field research. It has provided moreinformation about the contexts of robbery and new information about thecontext in which frauds occur. These differences can now be used to enrichand energize laboratory and field research. The difference between robberiesand fraud (awareness that a crime was taking place, delay between exposureand both recall and recognition) makes the fraud victims a less than perfectcontrol or low arousal group than would be acceptable by laboratory standards.However, finding a sufficient number of the low arousal, non-participatoryeyewitness such as mentioned earlier (witnessing the theft of a vacationingneighbor’s stereo) would be prohibitively time consuming. The archivalresearcher must take control or comparison groups as they come, warts and all.Similarly, the archival researcher is limited by the form the data take in the realworld. For example, in the current study, rates of selecting the police suspectcould be examined, but important questions regarding misidentification ofinnocent suspects and failing to identify the actual perpetrator had to be leftlargely unaddressed. Also, the police have a much narrower field of interestthan most researchers. The police are concerned with learning what theperpetrator looked like and if s/he said or did anything threatening such aswaving a gun about and shouting “nobody move and nobody gets hurt.” Incontrast, eyewitness researchers usually test recall of the entire event. Finally,the problem of ground truth will always accompany the archival researcher.100There are ways to keep it at bay, such as considering the weight of evidence foror against a particular suspect, but the problem will never entirely go away.No single research type can effectively deal with all of the issues relatedto eyewitness memory. The weapon focus phenomenon is a good example ofhow each type of research can be applied to a particular issue. The laboratoryis the most appropriate place to study basic processes. An alternativeexploration of other loci for the weapon focus phenomenon should beundertaken. Manipulating rehearsal between exposure and test might be a goodplace to start. Additionally, one possible explanation of the weapon focusphenomenon as demonstrated by laboratory studies is that it is a von Restorfeffect where subjects attend to the weapon because it is an unusual item. Thecontrol objects in the weapon focus studies have been usual items; a cheque,soup, a bag of chips or a magazine. Had the perpetrator held and pointed achihuahua at the clerk, a “weapon focus” phenomenon might have beendemonstrated for hand-held chihuahuas. Given that the human face carriesinformation about a person’s emotional state, in an actual armed robberysituation, I suspect that the victim and perhaps witnesses would alternatelooking at the weapon and at the perpetrator’s face, particularly if s/he wasspeaking. They also might send darting glances in the direction of friends ifpresent or at possible exits or hiding places. Depending on the limits of eyemovement monitoring technology, these assumptions could be tested in the labby having the subject victims be approached by a “perpetrator” wieldingvarious threatening (e.g. needles, snakes) unusual (e.g. chihuahuas, lego101houses) or mundane (e.g. papers, pencils) objects while monitoring their eyemovements. More of the context of an actual eyewitness situation could begained in the role play situations conducted at the Hendon training facilityutilized by Yuille et al (in press). The “perpetratorS’ could possibly brandishreal weapons in a confrontation with the officers-in-training. Archivalresearch has already demonstrated that future research should consideremploying recall tasks that are more in line with those used by police forcesinstead of relying on the more convenient questionnaire methods. Morearchival research on weapon focus is needed however to address the powerproblem hkely present in this research, and to determine if weapon presencehas similar effects on the identification accuracy of victims and witnesses.Weapon focus is not the only subject that would benefit from the multi-method approach; the entire field woulcL A combination of laboratory studies,field simulations, archival studies and case studies is needed. While this facthas been acknowledged (e.g. Davies, 1990; Yuille, 1993), there has been adistinct imbalance in the number of different types of studies. As a result, thisfield has painted a potentially distorted rather than a comprehensive picture ofeyewitnesses. There is a clear and pressing need for more research of thistype, as well as more direct studies of actual eyewitnesses of crime. This studyneeds replication and extension so that the extensive laboratory literature canbe appropriately and properly applied.102ReferencesAyuk, R.E. (1990). Cross-racial identification of transformed, untransformed andmixed-race faces. International Journal of Psychology, 25, 509-527.Baddely, A.D., & Woodhead, M. (1982). Depth of processmg, context and facerecognition. Canadian Journal of Psychology, 36(2), 148-164.Banaji, M.R. & Crowder, R.G. (1989). The bankruptcy of everyday memory.American Psychologist, 44(9), 1185-1193.Belli, R.F. (1989). Influence of misleading postevent misinformation:Misinformation interference and acceptance. Journal ofExperimentalPsychology: General, 118(1), 72 -85.Belli, R.F. (1993). Failure of interpolated test in inducing memory impairment withfmal modified test: Evidence unfavorable to the blocking hypothesis.American Journal ofPsychology, 106(3), 407-427.Blau, T.H. (1984). The psychologist as expert witness. New York: Wiley.Bloom, L.C. & Mudd, S.A. (1991). Depth of processing approach to facerecognition: A test of two theories. Journal of Experimental Psychology:Learning, Memory and Cognition, 17(3), 556-565.Boice, R., Hanley, P., Shauglinessy, P., & Gansler, D. (1982). Eyewitness accuracy:A general observational skill? Bulletin of the Psychonomic Society, Q(4),193-195.Borchard, E.M. (1932). Convicting the innocent: Errors of criminal justice. NewHaven: Yale University Press.Brandon, R. & Davies, C. (1973). Wrongful imprisonment: Mistaken convictionsand their consequences. London: George Allen and Unwin.Brigham, J.C., Maass, A., Martinez, D. & Whittenberger, G. (1983). The effect ofarousal on facial recognition. Basic and Applied Social Psychology, 4,279-293.103Brigham, J.C. (1990). Target person distinctiveness and attractiveness as moderatorvariables in the confidence-accuracy relationship in eyewitness identifications.Basic and Applied Social Psychology, 11(1), 101-115.Brigham, J.C., & Cairns, D.L. (1988). The effect of mugshot inspections oneyewitness identification accuracy. Journal of Applied Social Psychology, j,1394-1410.Brigham, J.C., & Barkowitz, P. (1978). Do “they all look alike?”: The effect ofrace, sex, experience and attitudes on the ability to recognize faces. Journal ofApplied Psychology, 8(4), 306-3 18.Brigham, J. C., Maass, A., Snyder, L. D., & Spaulding, K. (1982). Accuracy ofeyewitness identifications in a field setting. Journal of Personality and SocialPsychology, 4, 673-68 1.Brodsky, S.L. (1991). Testifying in court: Guidelines and maxims for the expertwitness. Washington, DC: American Psychological Association.Brown, E. Deffenbacher, K., & Sturgill, W. (1979). Memory for faces and thecircumstances of encounter. Journal of Applied Psychology 62(3), 311-318.Buckhout, R. (1975). Nearly 2000 witnesses can be wrong. Social Action and the2,7.Buckhout, R., Alper, A., Chern, S., Silverberg, G., & Slomovits, M. (1974).Determinants of eyewitness performance on a lineup. Bulletin of thePsychonomic Society, 4(3), 19 1-192.Buckhout, R., Figueroa, D., & Hoff, E. (1975). Eyewitness identification: Effects ofsuggestion and bias in identification from photographs, Bulletin of thePsychonomic Society, 6, 71-74.Burke, A., Heuer, F. & Reisberg, D. (1992). Remembering emotional events.Memory & Cognition, 20,277-290.Caminer, B.F. (1985). Credit card fraud: the neglected crime. Journal of Criminallaw and Criminology, 76(3), 746-763.Christiaanson. R.E., Sweeney, J.D., & Ochalek, K. (1983). Influencing eyewitnessdescriptions. Law and Human Behavior, 7, 59-65.104Christianson, S.-A., & Nilsson, L.-G. (1989). Hysterical amnesia: A case ofaversively motivated isolation of memory. In T. Archer & L.-G. Nilsson (Eds.),Aversion, avoidance, and anxiety: Perspectives on aversively motivatedbehavior. Hilisdale, NJ: Lawrence Eribaum.Christianson, S-A. (1984). The relationship between induced emotional arousal andamnesia. Scandinavian Journal of Psychology, 25, 147-160.Christianson, S-A. (1992). Emotional stress and eyewitness memory: A criticalreview. Psychological Bulletin, fl(2), 284-309.Christianson, S-A., Goodman, J. & Loftus, E.F. (1992). Eyewitness memory forstressful events: Methodological quandaries and ethical dilemmas. In S-A.Christianson (Ed.), The handbook of emotion and memory: Research andtheory (pp. 217-24 1). Hillsdale, NJ: Lawrence Eribaum.Christianson. S-A., & Loftus, E.F. (1987). Memory for traumatic events. AppliedCognitive Psychology, j, 225-239.Christianson. S-A., & Loftus, E.F. (1990). Some characteristic of people’s traumaticmemories. Bulletin of the Psychonomic Society, 28, 195-198.Christianson. S-A., & Loftus, E.F. (1991). Remembering emotional events: The fateof detailed information. Cognition & Emotion, 5, 8 1-108.Clifford, B. R. & Holin, C. R. (1981). Effects of the type of incident and the numberof perpetrators on eyewitness memory. Journal of Applied Psychology, 66(3),364-370.Courtois, M.R., & Mueller, J.H. (1981). Target and distractor typicality in facialrecognition. Journal of Applied Psychology, 66(5), 639-645.Cutshall, J.L. (1985). Eyewitness’ characteristics and memory: An in situ analysis.Unpublished master’s thesis, University of British Columbia, Vancouver, B..C.Cutler, B.L. & Penrod, S.D., & Martens, T.K. (1987). The reliability of eyewitnessidentification: The role of system and estimator variables. Law and HumanBehavior, 11(5), 233-258.105Cutler, B.L. & Penrod, S.D. (1988). Improving the reliability of eyewitnessidentification: Lineup construction and presentation. Journal of AppliedPsychology, 73(2), 281-290.Cutler, B.L. & Penrod, S.D. (1989). Forensically relevant moderators of the relationbetween eyewitness identification accuracy and confidence. Journal of AppliedPsychology, 7(4), 650-652.Cutshall, 3. & Yuille, J.C. (1989). Field studies of eyewitness memory of actualcrimes. In D.C. Raskin (Ed.). Psychological methods in criminal investigationand evidence (pp. 97-124). New York: Springer.Davies, G.M. (1989). The applicability of facial memory research. In A.W. Young& H.D. Ellis (Eds.), Handbook of research on face processing (pp.537- 562).Elsevier Science Publishers.Davies, G. (1990, September). Influencing public policy on eyewitness; problemsand possibilities. Paper presented at the Second European Law and PsychologyConference, Nuremberg, Germany.Davies, G., Ellis, H., & Shepherd, J. (1978). Face recognition accuracy as a functionof mode of representation. Journal ofApplied Psychology, 63, 180-187.Davies, G., Shepherd, J. & Ellis, H. (1979). Effects of interpolated mugshotexposure on accuracy of eyewitness identification. Journal of AppliedPsychology, 64(2), 232-237.Deffenbacher, K. A. (1983). The influence of arousal on the reliability of testimony.In S. M. A. Lloyd-Bostock & B. R. Clifford (Eds.), Evaluating witnessevidence, (pp. 235-25 1). Chichester: Wiley.Deffenbacher, K.A. (1986). On the memorability of the human face. In H.D. Ellis,M.A. Jeeves, F. Newcombe, and A. Young (Eds.), Aspects of Face Processing(pp. 61 -70). Dordrecht: Martinus Nijhoff.Easterbrook, J.A. (1959). The effect of emotion on cue utilization and theorganization of behavior. Psychological Review, , 183-201.Fisher, R.P., & Geiselman, R.E., (1992). Memory enhancing techniques forinvestigative interviewing. Springfield, IL: Charles C. Thomas.106Fisher, R. P., Geiselman, R. E., & Amador, M. (1989). Field test of the CognitiveInterview: Enhancing the recollection of actual victims and witnesses of crime.Journal of Applied Psychology, 74(5).Fisher, R.P., McCauley, M.R. & Geiselman, R.E. (1994). Improving eyewitnesstestimony with the Cognitive Interview. In D.F. Ross, J.D. Read & M.P. Toglia(Eds.), Adult eyewitness testimony: Current trends and developments (pp. 245-269). New York, NY: Cambridge University Press.Fleet, M.L., Brigham, J.C., & Bothwell, R.K. (1987). The confidence-accuracyrelationship: The effects of confidence assessment and choosing. Journal ofApplied Social Psychology, 17(2), 17 1-187.Fun, R. H., & Shepherd, J. W. (1986). Tall stories: Eyewitnesses’ ability to estimateheight and weight characteristics. Human Learning, , 29-3 8.Geiselman, R.E., Fisher, R.P., Firstenberg, I., Hutton, L.A., Sullivan, S.J., Avetissian,I.V., & Prosk, A.L. (1984). Enhancement of eyewitness memory: An empiricalevaluation of the cognitive interview. Journal ofPolice Science andAdministration, 12(1), 74-80.Goldstein, A.G., Chance, J.E. & Schneller, G.R. (1989). Frequency of eyewitnessidentification in criminal cases: A survey of prosecutors. Bulletin of thePsychonomic Society, 27(11), 7 1-74.Goldstein, A.G., Johnson, K., & Chance, J. (1979). Does fluency of face descriptionimply superior face recognition? Bulletin of the Psychonomic Society, 13(1),15-18.Gorenstein, G. W., & Ellsworth, P. C. (1980). Effect of choosing an incorrectphotograph on a later identification by an eyewitness. Journal of AppliedPsychology, 65(5), 6 16-622.Heuer, F., & Reisberg, G. (1992). Emotion, arousal and memory for detail in S-AChristianson (Ed.), The handbook of emotion and memory: Research andTheory (pp. 15 1-180). Hillsdale, NJ: Erlbaum.Heuer, F. & Reisberg, D. (1990). Vivid memories of emotional events: The accuracyof remembered minutiae. Memory and Cognition, 18, 496-506.107Hosch, H. (1994). Individual differences in personality and eyewitness identification.In D.F. Ross, J.D. Read, & M.P. Toglia (Eds.) Adult eyewitness testimony:Current trends and developments. New York, NY: CambridgeHosch, H. M. & Bothwell, R. K. (1990). Arousal, descriptions and identificationaccuracy of victims and bystanders. Journal of Social Behavior andPersonality, 5(5), 481-488.Hosch, H. M. & Cooper, D. 5. (1982). Victimization as a deterniinant of eyewitnessaccuracy. Journal of Apnlied Psychology, 67, 649-652.Hosch, H. M., Leippe, M. R., Marchiom, P. M., & Cooper, D. S. (1984).Victimization, self-monitoring, and eyewitness identification. Journal ofApplied Psychology, , 280-288.Hosch, H. M., & Platz, S. J. (1984). Self-monitoring and eyewitness accuracy.Personality and Social Psychology Bulletin, IQ, 283-289.Houts, M. (1956). From evidence to guilt. Springfield, IL: Charles C. Thomas.Kagehiro, K.K. & Laufer, W.S. (Eds.). Handbook of psychology and the law. NewYork: Springer-Verlag.Kassin, S. M. (1984). Eyewitness identification: victims versus bystanders. Journalof Applied Social Psychology, 14(6), 5 19-529.Kassin, S.M. (1985). Eyewitness identification: Retrospective self-awareness and theaccuracy-confidence correlation. Journal of Personality and Social Psychology,4, 878-893.Kassin, S.M., Rigby, S. & Castillo, S.R. (1991). The accuracy-confidence correlationin eyewitness testimony: Limits and extensions of the retrospective self-awareness effect. Journal of Personality and Social Psychology, 61(5), 698-707.Kohnken, G. & Maass, A. (1988). Eyewitness testimony: False alarms on biasedinstructions? Journal of Applied Psychology, 73(3), 363-370.Konecm, V.J., & Ebbesen, E.B. (1986). Courtroom testimony by psychologist oneyewitness identification issues: Critical notes and reflections. Law andHuman Behavior, 10, 117-126.108Krafka, C. & Penrod, S. (1985). Reinstatement of context in a field experiment oneyewitness identification. Journal of Personality and Social Psychology, 49,58-69.Kramer, T.H., Buckhout, R. & Eugenio, P. (1990). Weapon focus, arousal andeyewitness memory: Attention must be paid. Law and Human Behavior, 14,167-184.Kuehn, L.L. (1974). Looking down a gun barrel: Person perception and violentcrime. Perceptual and Motor Skills, 39, 1159-1164.Laughery, K.R., & Wogalter, M.S. (1989). Forensic applications of facial memoryresearch. In A.W. Young & H.D. Ellis (Eds.), Handbook of research on faceprocessing (pp. 519 - 555). : Elsevier Science Publishers.Leippe, M. R. (1980). Effects of integrative memorial and cognitive processes onthe correspondence of eyewitness accuracy and confidence. Law and HumanBehavior,Leippe, M. R. , Romanczyk, A. & Mamon, A. P. (1991). Eyewitness memory for atouching experience: Accuracy differences between child and adult witnesses.Journal of Applied Psychology, 76(3), 367-379.Lindsay, R.C.L. & Nosworthy, G.J. (1994). Using mugshots to fmd suspects. Journalof Applied Psychology, 79(1), 121-130.Lindsay, R.C.L., Walibridge, H., & Drennan, D. (1987). Do the clothes make theman? An exploration of the effect of lineup attire on eyewitness identificationaccuracy. Canadian Journal of Behavioural Science, 19,463-478.Lindsay, R.C.L. & Wells, G.L. (1985). Improving eyewitness identifications fromlineups: Simultaneous versus sequential lineup presentation. Journal ofApplied Psychology, 70(3), 555-564.Lipton, J.P. (1977). On the psychology of eyewitness testimony. Journal of AppliedPsychology, 90-93.Loftus, E.F. (1976). Ten years in the life of an expert witness. Law and HumanBehavior, 10,241-263.109Loftus, G.R. (1972). Eye fixations and recognition memory for pictures. CognitivePsychology, 3, 525-55 1.Loftus, E.F. (1983). Silence is not golden. American Psychologist, , 564-572.Loftus, E.F. (1979). Eyewitness testimony. Cambridge, MA: Harvard UniversityPress.Loftus, E.F., Loftus, G.R., & Messo, J. (1987). Some fact about “weapon focus.”Law and Human Behavior, 11(1), 55-62.Loftus, E.F., Miller, D.G., & Burns, H.J. (1978). Semantic integration of verbalinfonnation into a visual memory. Journal of Experimental Psychology:Human Learning and Memory, 4, 19-31.Loftus, E.F., & Ketcham, K. (1991). Witness for the defense: The accused, theeyewitness and the expert who puts memory on trial. New York: St. Martin’sPress.Loftus, E.F., & Palmer, J.C. (1974). Reconstruction of automobile destruction: Anexample of the interaction between language and memory. Journal of VerbalLearning and Verbal Behavior, U. 585-589.Loftus, E.F., & Zanni, G. (1975). Eyewitness testimony: The influence of thewording of a question. Bulletin of the Psychonomic Society, , 86-88.Logie, R. H., Baddeley, & Woodhead, A. D. (1987). Face recognition, pose andecological validity. Applied Cognitive Psychology, 1, 53-69.Luus, C.A.E. & Wells, G.L. (1991). Eyewitness identification and the selection ofdistractors for lineups. Law and Human Behavior, 15, 43-57.Maass, A., & Kohnken, G. (199X). Eyewitness identification: Simulating the“weapon effect.” Law and Human Behavior, U(4), 397-408.Malpass, R.S., & Devme, P.G. (1981). Realism and eyewitness identificationresearch. Law and Human Behavior, 4(4), 347-358.Malpass, R.S., & Devine, P.G. (1981). Eyewitness identification: Lineup instructionsand absence of the offender. Journal of Applied Psychology, 66, 482-489.110Marascuilo, L.A. (1966). Large-sample multiple comparisons. PsychologicalBulletin, 65, 280-290.McCloskey, M. & Egeth, H. E. (1983). Eyewitness identification: What can apsychologist tell a jury? American Psychologist, , 550-563.McKelvie, S.J. (1985). Effect of depth of processing on recognition memory fornormal and inverted photographs of faces. Perceptual and Motor Skills, 60(2),503-508.McKelvie, S.J. (1988). The role of spectacles in facial memory: A replication andextension. Perceptual and Motor Skills, 66, 65 1-658.Melara, R.D., DeWitt-Rickards, T.S., & OBrien, T.P. (1989). Enhancing lineupidentification accuracy: Two codes are better than one. Journal of AppliedPsychology, 74(5), 706-7 13.Murray, D.M., & Wells, G.L. (1982). Does knowledge that a crime was stagedaffect eyewitness performance? Journal of Applied Social Psychology, 12(1),42-53.Neil v. Biggers, 409 U.S. 188 (1972).Peters, D.P., (1991). The influence of stress and arousal on the child witness. In J.Doris (Eci), The suggestibility of children’s recollections (pp. 60-76.).Washington DC: American Psychological AssociationPigott, M. A., Brigham, 3. C. & Bothwell, R. K. (1990). A field study on therelationship between quality of eyewitnesses’ descriptions and identificationaccuracy. Journal of Police Science and Administration, 17(2), 84-88.Read, 3. D., Hammersley, R., Cross-Calvert, S. & McFadzen, E. (1987 OR SO)Rehearsal of faces and details in action events. Applied Cognitive Psychology,VOL< PGSRead, 3. D., Tollestrup, P. Hammersley, R., McFadzen, E. & Christensen, A. (1990).The unconscious transference effect: Are innocent bystanders evermisidentified? Applied Cognitive Psychology, 4, 3-3 1.111Read, J.D., Yuille, J.C., & Tollestrup, P (1992). Recollections of a robbery: Effectsof arousal and alcohol upon recall person identification. Law and HumanBehavior, 16(4), 425-446.Reisberg, D., Heuer, F., McLean, J., & O’Shaugnessy, M. (1988). The quantity, notthe quality, of affect predicts memory vividness. Bulletin of the PsychonomicSociety, 26, 100-103.Reisberg, D., Heuer, F., McLean, J. & O’Shaugbnessy, M. (1988). The quantity, notthe quality, of affect predict memory vividness. Bulletin of the PsychonomicSociety, 26, 100-103.Ross, D.F., Read, J.D., & Toglia, M.P. (Eds.). (1994). Adult eyewitness testimony:current trends and developments. New York: Cambridge University Press.Sanders, G.S, & Warnick, D.H. (1981). Truth and consequences: The effect ofresponsibility on eyewitness behavior. Basic and Applied Social Psychology,2(1), 67-79.Schooler, J.W., & Engstler-Schooler, T.Y. (1990). Verbal overshadowing of visualmemories: Some things are better left unsaid. Cognitive Psychology, 22(2),36-71.Shapiro, P.N., & Penrod S. (1986). Meta-analysis of facial identification studies.Psychological Bulletin, 100(2), 139-156.Smith, J.E., Pleban, R.J., & Shaffer, (1982). Effects of interrogator bias on a policetrait questionnaire on the accuracy of eyewitness identification. Journal ofSocial Psychology, 116(1), 19-26.Sporer, S.L. (1991). Deep--deeper--deepest? Encoding strategies and the recognitionof human faces. Journal of Experimental Psychology: Learning. Memory andCognition, 17(2), 323-333.Sporer, S.L. (in press). Describing others: Psychological issues. In S.L. Sporer, R.S.Malpass, & G. Kohuken (Eds.), Psychological issues in eyewitnessidentification. Hillsdale, NJ: Eribaum.Steblay, N.M. (1992). A meta-analytic review of the weapon focus effect. Law andHuman Behavior, (4), 4 13-434.112Tooley, V., Brigham, J.C., Maass, A., & Bothwell, R.K. (1987). Facial recognition:Weapon effect and attentional focus. Journal of Applied Social Psychology,17, 845-849.Vokey, J.R. & Read, J.D. (1992). Familiarity, memorability, and the effect oftypicality on the recognition of faces. Memory and Cognition, 20(3), 29 1-302.Wagenaar, W.A., & Groenweg, J. (1990). The memory of concentration campsurvivors. Applied cognitive Psychology, 4, 77-87Warnick , D.H., & Sanders, G.S. (1980). Why do eyewitnesses make so manymistakes? Journal of Applied Social Psychology, 10(4), 362-366.Wells, G.L., & Murray, D.M. (1984). Eyewitness confidence. In G.L. Wells & E.F.Loftus, (Eds.), Eyewitness testimony: Psychological perspectives (pp. 155-170).New York: Cambridge University PressWells, G. L. (1978). Applied eyewitness testimony research: system variables andestimator variables. Journal of Personality and Social Psychology, 36(12),1546-1557.Wells, G. L. (1985). Verbal descriptions of faces from memory: Are they diagnosticof identification accuracy? Journal of Applied Psychology, 70(4), PAGESWells, G.L., Leippe, M.R. & Ostrom, T.M. (1979). Guidelines for empiricallyassessing the fairness of a lineup. Law and Human Behavior, 3, 285-293.Wells, G. L., & Leippe, M. R. (1981). How do triers of fact infer the accuracy ofeyewitness identifications? Using memory for peripheral detail can bemisleading. Journal ofApplied Psychology, 66(6), 682-687.Wells, G.L. & Lindsay, R.C.L. (1980). On estimating the diagnosticity of eyewitnessnon-identifications. Psychological Bulletin, 88, 320-329.Wells, G.L. & Loftus E.F. (Eds.). Eyewitness testimony: Psychological perspectives.London: Cambridge University Press.Wells, G.L. & Hryciw, B. (1984). Memory for faces: Encoding and retrievaloperations. Memory and Cognition, ji(4), 33 8-344.113Wells, G.L., Seelau, E., Rydeli, S. & Luus, C.A.E. (1994). Recommendations forproperly conducted lineup identification tasks. In D. Ross, J.D. Read, & M.P.Toglia (Eds.), Eyewitness testimony: current trends and developments. NewYork: Cambridge University Press.Wells, G.L. & Turtle, J.W. (1986). Eyewitness identification: The importance oflineup models. Psychological Bulletin, , 320-329.Whipple, G.M. (1909). The observer as reporter: A survey of the “Psychology ofTestimony.” Psychological Bulletin, 153-170.Wogalter, M.S., Laughery, K.R., & Thompson, B.G. (1990). Face compositeconstructions: Effects on subsequent recognition performance. Unpublishedmanuscript.Wolfskiel, M. P. & Brigham, J. C. (1985). The relationship between accuracy ofprior description and facial recognition. Journal of Applied Psychology, 70,Woodhead, M.M., Baddeley, A.D. & Simmonds, D.C.V. (1979). On training peopleto recognize faces. Ergonomics, 22, 333-343.Yarmey, A.D. (1979). The psychology of eyewitness testimony. New York: FreePress.Yarmey, A. D. (1986). Verbal, visual and voice identification of a rape suspect underdifferent levels of illumination. Journal of Applied Psychology, 71(3), 363-370.Yarmey, A. D., Jones, H. P. T. & Rashid (1984). Eyewitness memory of elderly andyoung adults. In D. J. Muller, D. E. Blackman & Chapman, A. J. (Eds.), Topicsin psychology and law. Cliichester: Wiley.Yerkes, R.M., & Dodson, J.D. (1908). The relation of strength of stimulus to rapidityof habit-information. Journal of Comparative Neurology of Psychology,18,459-482.Yuille, J.C. (1993). We must study forensic eyewitnesses to know about them.American Psychologist, 48(5), 572-573.114Yuille, 3. C. (1986). Meaningfiul research in the police context. In J.C. Yuille (Ed.),Police selection and training: The role of psychology. Dordrecht, TheNetherlands: Martinus Nijhoff.Yuille, J. C., & Cutshall, 3. L. (1986). A case study of eyewitness memory of acrime. Journal of Applied Psychology, 71(2), 291-301.Yuille, J. C., & Cutshall, J. L. (1989). Analysis of the statements of victims,witnesses and suspects. In J. C. Yuile (Ed.), Credibility Assessment (pp. 175-191). Dordrecht, The Netherlands: Kluwer Academic.Yuille, J. C., Davies, G., Giblmg, F., Marxsen, D. & Porter, S. (in press).Eyewitness memory of police trainees for realistic role plays. Journal ofApplied Psychology.Yuille, J. C. & Tollestrup, P. A. (1990). Some effects of alcohol on eyewitnessmemory. Journal of Anplied Psychology, 7.(3), 268-273.Yuille, 3. C. & Tollestrup, P. A. (1992). A model of the diverse effects of emotion oneyewitness memory. In S. A. Christianson (Ed.), The handbook of emotion andmemory (pp. 201-215). Hilisdale, NJ: Lawrence Eribaum.Yuille, J. C. & Tollestrup, P. A., Marxsen, D., & Porter, S. (under review). Someeffects of marijuana on eyewitness memory.Yuille, J.C. & Wells, G.L., (1991). Concerns about the application of researchfindings: The issue of ecological validity. In J. Doris (Ed.) The suggestibility ofchildren’s recollections (pp. 118-128). Washington, DC: AmericanPsychological Association.Zanni, G.R., & Offerman, J.T. (1978). Eyewitness testimony: An exploration ofquestion wording upon recall as a function of neuroticism. Perceptual andMotor Skills, 46, 163-166.Zimmerman, S. (1982, April). Photo recognition accuracy. Presented at the meetingof the Southwestern Rocky Mountain AAAS, El Paso, TX115APPENDIX AAnalysis of Variance of total amount of detail provided by eyewitness type(robbery victims, robbery witnesses, and fraud victims).Source SS D.F. MS FBetween groups 5634.57 2 2817.28 72.63 p<.0000IWithin groups 13536.52 349 38.79Analysis of Variance of amount of clothing detail provided by eyewitness type(robbery victims, robbery witnesses, and fraud victims).Source SS D.F. MS FBetweengroups 1232.71 2 616.35 56.08 p<.00001Within groups 3835.36 349 10.99Analysis of Variance of amount of physical appearance detail provided byeyewitness type (robbery victims, robbery witnesses, and fraud victims).Source SS D.F. MS FBetween groups 1726.02 2 863.01 54.07 p < .00001Within groups 5569.88 349 15.96116Analysis of Variance of amount of total detail provided by eyewitness type(robbery victims, robbery witnesses, and fraud victims) for those eyewitnesseswho recalled more than four physical appearance details.Source SS D.F. MS FBetween groups 722.62 2 361.31 11.82 p < .00001Within groups 4461.55 146 30.55Analysis of Variance of amount of clothing detail provided by eyewitness type(robbery victims, robbery witnesses, and fraud victims) for those eyewitnesseswho recalled more than four physical appearance details.Source SS D.F. MS FBetweengroups 425.28 2 212.64 14.68 p<.000lWithin groups 2114.41 146 14.48Analysis of Variance of amount of physical appearance detail provided byeyewitness type (robbery victims, robbery witnesses, and fraud victims) forthose eyewitnesses who recalled more than four physical appearance details.Source SS D.F. MS FBetween groups 117.53 2 58.77 5.86 p<.OlWithin groups 1463.15 146 10.02117Analysis of Variance of amount of total detail provided by eyewitnesscondition (COND - victims and witnesses of robbery) and by weapon presence(WPN).Source SS D.F. MS FCOND 360.43 1 360.426 7.11 p < .01WPN 668.562 1 668.562 13.19 p<.0001CONDBYWPN 5.14 1 5.14 101 p>.10Within (error) 10745.47 212 50.69Analysis of Variance of amount of clothing detail provided by eyewitness type(victims and witnesses of robbery) and by weapon presence.Source SS D.F. MS FCOND 6.34 1 6.24 .391 p> .10WPN 240.24 1 240.24 15.047 p<.000lCONDBYWPN 3.52 1 3.52 .221 p>.l0Within (error) 3384.85 212 15.99Analysis of Variance of amount of physical appearance detail provided byeyewitness type (victims and witnesses of robbery) and by weapon presence.Source SS D.F. MS FCOND 271.80 1 271.80 14.04 p<.0001WPN 107.26 1 107.26 5.42 p < .05CONDBYWPN 17.17 1 17.17 .887 p>.10Within (error) 4102.98 212 19.35118Analysis of Covariance of weapon presence with delay on identificationoutcomes.Source SS D.F. MS FCovariateDelay 2.69 1 2.69 12.83 p < .0001Main EffectWeapon .751 1 .751 3.58 p<.1OWithin 26.43 126 .210119