Error and Accuracy in Eyewitness Memory for Firearms
Matthew J. Sharps
This research addresses the cognitive bases for errors in firearm description and identification, a largely neglected area within the field of eyewitness memory. Consistent with the theoretical considerations driving this study, a substantial portion of participant responses were uninformative, based on general, “gestalt” concepts rather than on concepts emphasizing specific details. However, when participants were questioned in a manner that encouraged more specific “feature-intensive” processing, their responses became more generative. Surprisingly, the gender of the assailant holding a firearm did not contribute to error, for witnesses of either gender. These results provide practical information on eyewitness memory for consultants, investigators, and legal experts in investigative and courtroom proceedings.
Key Words: Eyewitness Error, Eyewitness Memory, Weapon Identification, Gestalt/Feature-Intensive Processing
The American College of Forensic Examiners Institute® provides this continuing education opportunity to fulfill 1hr of Continuing Education Credit for all certified members. Certified members are required to obtain 30 hours of continuing education credits in the 3 year recertification period to maintain their certification status.
ACFEI members can take CEs HERE
Originally published in The Forensic Examiner - November, 2015
Eyewitness memory for the weapons used in crimes is extremely important in investigation and in courtroom settings. Yet there is little information available for consultants, investigators, and legal experts confronted with this issue.
A vast corpus of research on eyewitness memory chiefly concerned with person identification – and to a lesser degree with memory for crime scenes – has revealed many factors that affect the eyewitnesses. Witness confidence (Wells, Lindsay, & Ferguson, 1979; Wells, 1993; Penrod & Cutler, 1995), the construction of line-ups (Wells, 1996), situational factors including poor lighting, lack of attention, distance between witness and perpetrator, exposure time, retention interval, and stress (Loftus, 2010; Memon, Hope, & Bull, 2003; Narby, Cutler, & Penrod, 1996; Sporer, Malpass, & Koehnken, 1996; Wells, Memon, & Penrod, 2006;), and cross-racial identification (Meissner & Brigham, 2001) have all been shown to influence eyewitness performance. This body of work has proven to be of enormous practical, as well as theoretical, importance. Eyewitness testimony is persuasive to juries (Brewer & Burke, 2002; Semmler, Brewer, & Douglas, 2012), even when the validity of eyewitness evidence is called into question (Loftus, 1974). Eyewitness identification was a contributing factor in over 75% of erroneous convictions examined by the Innocence Project (Retrieved April 15, 2013, www.innocenceproject.org).
Thus, as this brief review of research on the subject shows, approaches to eyewitness memory for persons, and even for crime scenes, has reached a high degree of sophistication. Yet, the study of eyewitness memory for weapons has remained in its infancy; we do not currently possess an understanding even of the rudiments of error types and contextual influences, such as the gender of a given perpetrator, that arise in the eyewitness processing of firearms.
The minimal information available on this subject is far from encouraging. Even under ideal viewing conditions, witnesses have been shown to have relatively poor memory for firearms: they correctly identify common handguns less than 50% of the time a mere ten minutes after observation (Sharps, Barber, Stahl & Villegas, 2005). Larger weapons that have more attention-drawing features were recalled at a higher level, but still only 71% of the time under ideal viewing conditions.
What is the nature of accuracy and error in weapon description and identification? When law enforcement questions witnesses, the questions tend to go from general and global questions to more specific questions regarding the features of a given individual or situation (Fisher & Schreiber, 2007). In theoretical terms that have been shown to apply to memory for person and for the details of crime scenes (Sharps, Hess, Casner, Ranes, & Jones, 2007), as well as to memory for vehicles (Villegas, Sharps, Satterthwaite, & Chisholm, 2005), the general questions taking place in the initial phases of questioning tend to lead to more “gestalt” responding, in which the crucial features of given crime elements may be ignored. The more specific questions asked during later phases, however, may promote “feature-intensive” processing, wherein those crucial features are more likely to be recovered (Sharps & Nunes, 2002; Sharps, 2003, 2010). Higher levels of item-specific, feature-intensive information tend to reduce false memories (McCabe, Presmanes, Robertson, & Smith, 2004). Thus, especially within the tradition of Cognitive Interview recommendations, in which more specific inquiries follow the more general questions involved in open-ended narration (Fisher & Schreiber, 2007), better weapon recall may be more likely to occur with “feature-intensive” questioning.
But is this the case? One of the most important and established facets of eyewitness memory, lies in the tendency of memories to reconfigure themselves with time. Bartlett (1932) showed that memory is not static, but tends to change in the direction of brevity and personal belief. His original research was confirmed by more modern work (Ahlberg & Sharps, 2002; Bergman & Roediger, 1999).
These reconfigurative tendencies can explain the types of errors eyewitnesses commonly make. In a systematic evaluation of typical eyewitness errors in person and scene memory (Sharps, Janigian, Hess, & Haward, 2009), the most common mistakes required little theoretical explanation (errors of the physical appearance and clothing of the perpetrator; slightly less, on average, than two such errors per witness). However, the second most common type of error (one and one quarter per witness, on average) were not so straightforward. These were errors of the imagination, in which witnesses simply made things up, with no physical bases for the errors. Witnesses confabulated material, which they reported as fact, based on memories influenced by their imaginations (Sharps, Janigian, Hess, & Hayward, 2009). Other types of important errors (environmental features, etc.) were far less likely to be made than those based solely in imagination and in the type of reconfiguration first demonstrated by Bartlett (1932).
How do these types of errors operate on eyewitness memories of weapons used in crimes? What are the types of errors typically made in this area, and how do they vary in regard to gestalt or feature-intensive questioning? Do these errors vary with the gender of the perpetrator holding the gun? In physical confrontations, women may use firearms to “equalize” typical differences in size and strength between the sexes (Felson, 2002). Female homicide offenders are more likely to use guns against male victims than female victims (Kleck & McElrath, 1991). However, there is virtually no evidence in the literature on eyewitness memory for a given gun wielded by a woman as opposed to a man. In earlier work, we found differences in witness memory regarding the physical characteristics of female and male assailants (Sharps et al., 2009), and the surprising vision of a woman pointing a gun at a victim might detract from an eyewitness’s ability to give a coherent account of the gun itself.
These were the questions driving the present work, the intent of which was to provide consultants and investigators with accurate information on four critical issues:
Twenty-eight women (mean age 19.36, SD = 2.06) and nine men (mean age 21.00, SD = 2.45) participated. Gender ratio reflected the population from which these respondents were drawn, the psychology subject pool of a central California university. All respondents provided informed consent and received course credit for participation.
Respondents viewed a realistic crime scene, projected by means of a Lafayette 91010 projection tachistoscope onto a white screen, from distances of eight to fifteen feet. The scenes employed were taken from a stimulus set created with law enforcement consultation and supervision, in order to increase realism and accuracy, and have been used and standardized repeatedly in previous research (Sharps, 2010).
The scenes, set in a cluttered driveway, depicted a male or female “assailant,” full-body in profile, pointing a 9mm Beretta semi-automatic pistol at a “victim.” Safety precautions were of course observed, and the scenes were photographed under experienced law enforcement supervision in order to create a safe depiction of an ecologically valid crime scene (Sharps, 2010).
Participants were provided with no information on what they were about to see, although they did engage in a standard modified Snellen test (Sharps et al., 2009; Sharps, Herrera, Dunn, & Alcala, 2012) of visual acuity, demonstrating that all had the necessary visual abilities needed to resolve the smallest relevant details in the scene (visual acuity of approximately 20/40 minimum). Participants were then shown the image described above, with an exposure of five seconds, standard for previous research with these materials (Sharps et al., 2009) and providing ample time for full encoding of this realistic situation (Montejano, 2004; Moore, 2006).
The participants then performed a filler task designed to prevent rehearsal of the information seen; this task involved the provision of basic demographic information (name, address, telephone number) similar to that which might be elicited by an investigator or police dispatcher.
After 10 minutes, a reasonable interval between initial call for service and questioning by arriving police officers (Montejano, 2004; Moore, 2006), participants were asked to fill out a questionnaire addressing what they had seen. They were first asked to describe the scene generally; although not important for the present inquiry, this was intended to simulate typical police questioning procedures (Fisher & Schreiber, 2007). Interestingly, participants did not provide details of the weapon in this open-ended narration, typical of the initial phases of the Cognitive Interview (Fisher & Schreiber, 2007).
Participants were then asked if they saw a gun in the scene, and if so, how they knew it was a gun. This was the gestalt manipulation, designed to elicit responses of the type characteristic of initial, more global police interview questions (Fisher & Schreiber, 2007). They were then asked feature-intensive questions, designed to elicit more item-specific detail, and were asked to describe and identify the gun by type, as well as to detail the reasons behind their descriptions. In a procedure simulating a typical police interview, we were able to examine the errors generated under earlier gestalt and later feature-intensive conditions of questioning, a course of questioning typical of that used under real-world police field conditions.
Three response types were initially examined. These were as follows:
The “useless” responses were further subjected to a standard content analysis method (Sharps et al, 2009) designed to evaluate the nature of these responses. All protocols were evaluated by three researchers, resulting in a complete consensus on the three specific categories of the responses, described below in the Results section. There were no errors that fell outside these categories.
When respondents were asked the “gestalt” questions, requiring them to state only whether or not they saw a gun and how they knew it was a gun, the difference between correct, useless, and incorrect responses was significant: F (2,34) = 103.34, p < .001. There were significantly more correct responses than incorr0ect answers, and more useless responses than incorrect answers. However, the difference between the number of correct and useless responses was not significant (standard paired t-tests after repeated-measures ANOVA, p < .05; see Table 1). No significant differences were observed in the responses of those who saw the male, versus the female, assailant.
In contrast, when respondents were asked the “feature-intensive” questions, requiring them to state the type of gun employed, and to detail their reasons, the pattern changed. The difference between correct, useless, and incorrect responses was still significant: F (2,30) = 8.89, p = .001. However, under feature-intensive conditions, the level of accurate response significantly exceeded the number of useless and inaccurate responses, which were statistically indistinguishable (paired t-tests after repeated-measures ANOVA, p < .05). Again, no significant differences were observed between male and female perpetrators.
Neither correct nor incorrect responses varied with question type. However, more useless responses were generated by gestalt than by feature-intensive questioning (paired t-test, p < .05).
Three types of “useless” responses were generated. These were found in content analysis to be of three types: “circular” responses (e.g., it was a gun because it looked like a gun); “inferential” responses (e.g., it was a gun because it looked like a crime was occurring); and “contextual” responses (e.g., the scene was well-lit, or the view was good). The difference in prevalence for these three types of responses was significant for gestalt questioning (F (2, 34) = 11.163, p < .001), but not for feature-intensive questioning (F (2,30) = 0.21, p = .811). Under gestalt questioning, circular and inferential responses did not differ in frequency, but circular responses significantly exceeded both contextual and referential answers, and inferential responses were more frequent than contextual ones (p < .05). This complex pattern of useless and potentially confounding responses was suppressed by feature-intensive questioning, to an average of less than a single “useless” response of each type per respondent.
A final analysis addressed the effect of the assailant’s gender on the separate responses of male and female participants. This analysis simply repeated the results of the work reported above; there was no identifiable or statistical effect of assailant’s gender on the responses or response patterns of either male or female respondents.
Table 1 reveals that very few responses of any type were generated in regard to these questions. Within the scope of this research, eyewitnesses were shown to provide few details of weapons observed, whether accurate, useless, or plain wrong. However, any consultant, investigator, or legal expert with courtroom experience should be aware of the crucial effect that a single detail may have in any given proceeding; the second author has observed this phenomenon in several real-world cases. Low in numbers as they may be, these types of responses can have a devastating impact in court.
The voluminous literature on the Cognitive Interview demonstrates the importance of following relatively gestalt questions with more feature-intensive inquiries in order to generate more accurate responses under feature-intensive conditions. However, as noted above, these studies have addressed memory for persons and scenes, as opposed to weapons. The present results stand as a cautionary exception to this general rule. In this study, feature-intensive inquiry did not enhance veridical memory of detail. Rather, it merely suppressed the tendency, on the part of witnesses, to give useless responses to the questions. The present results should not be extrapolated inappropriately beyond their scope. However, within that scope, it appears that additional veridical details of weapons are unlikely to be evoked from witnesses by means of a specific, feature-intensive request for weapon type and the details supporting that identification. More feature-intensive questioning did suppress the flow of useless information, which might lead an investigator or attorney to the subjective impression that memory was “improved” by the course of questioning gradually becoming more narrow and specific, such as that of the Cognitive Interview. However, it must be emphasized that this subjective impression, within the scope of present results, would be illusory. Additional accurate details were not forthcoming.
A more encouraging finding lies in the low numbers of inaccurate responses generated either under gestalt or feature-intensive question conditions. Although Bartlett (1932) reconfiguration is a fact of eyewitness memory (Sharps et al., 2009), most witnesses probably have a relatively limited prior framework of firearms knowledge (Bransford & Johnson, 1972) from which to base extensive confabulations of a given weapon. Thus, and again within the scope of the present results, it appears that witness confabulation, so important in memory for persons and scenes, is unlikely to be a major factor in weapon recall and identification.
Finally, and perhaps surprisingly, the issue of gender turned out, in the present study, to be a non-issue. We know that weapon focus, the tendency to attend to a weapon at the expense of surrounding details, can reduce eyewitness memory for persons and surroundings (Steblay, 1992). We also know that there are differences in eyewitness memory for persons between armed male and female assailants (Sharps et al., 2009), yet these effects appear to be one-way: a violent situation involving a gun may alter one’s memory of a male or female assailant, but again, within the scope of the present results, the gender of the assailant had no identifiable effect on one’s memory of the gun. Thus, at this point, it appears that memory of the weapons wielded by male and female perpetrators may be interpreted equally and without bias.
The present results demonstrate the following with regard to eyewitness memory for weapons, within the scope of the present methods:
Future research should expand the scope of this work to additional contexts, and should address the degree to which new interview techniques may elicit more accurate details in eyewitness memory for weapons. Ultimately, and hopefully, this line of inquiry will generate a corpus of information on eyewitness memory for weapons that will be congruent with the sophisticated body of information available on other aspects of eyewitness cognition.
Ahlberg, S.W. & Sharps, M.J. (2002). Bartlett revisited: Reconfiguration of long-term memory in young and older adults. Journal of Genetic Psychology, 163(2), 211-218.
Bartlett, F.C. (1932). Remembering: A study in experimental and social psychology. London, UK: Cambridge University Press.
Bergman, E.T. & Roediger, H.L. (1999). Can Bartlett’s repeated reproduction experiments be replicated? Memory and Cognition, 27(6), 937-947.
Bransford, J.D., & Johnson, M.K. (1972). Contextual prerequisites for understanding: Some investigations of comprehension and recall. Journal of Verbal Learning and Verbal Behavior, 11, 717-726.
Brewer, N., & Burke, A. (2002). Effects of testimonial inconsistencies and eyewitness confidence on mock-juror judgments. Law and Human Behavior, 26(3), 353-364.
Felson, R.B. (2002). Violence and Gender Reexamined. Washington, DC: American Psychological Association.
Fisher, R.P., & Schreiber, N. (2007). Interview protocols to improve eyewitness memory. In Toglia, M.P., Read, J.D., Ross, D.F., & Lindsay, R.C.L. (Eds.), Handbook of Eyewitness Psychology (Vol. 1, 53-80). Mahwah, NJ: Erlbaum.
Innocence Project (2013). Innocence project updates. Retrieved April 15, 2013, from www.innocenceproject.org.
Kleck, G., & McElrath, K. (1991). The effects of weaponry on human violence. Social Forces, 69, 669-692.
Loftus, E. (1974). Reconstructing memory: The incredible eyewitness. Psychology Today, 116-119.
Loftus, G. (2010). What can a perception-memory expert tell a jury? Psychonomic Bulletin & Review, 17(2), 143-148.
McCabe, D.P., Presmanes, A.G., Robertson, C.L., & Smith, A.D. (2004). Item-specific processing reduces false memories. Psychonomic Bulletin and Review, 11, 1074-1079.
Meissner, C.A., & Brigham, J.C. (2001). Thirty years of investigating the own-race bias in memory for faces: A meta-analytic review. Psychology, Public Policy, & Law, 7, 3-35.
Memon, A., Hope, L., & Bull, R. (2003). Exposure duration: Effects on eyewitness accuracy and confidence. British Journal of Psychology, 94, 339-354.
Montejano, D. (2004, November). Fresno Police Department Conference on stress in police work. Fresno, CA.
Moore, L. (2006, November). Conference on the use of force in law enforcement. Office of the United States Marshall, Fresno, CA.
Narby, D.J., Cutler, B.L., & Penrod, S.D. (1996). The effects of witness, target, and situational factors on eyewitness identifications. In: Sporer, S.L., Malpass, R.S., & Koehnken, G. (1996). Psychological issues in eyewitness identification. Mahwah, New Jersey: Lawrence Erlbaum Associates, Inc.
Penrod, S., & Cutler, B. (1995). Witness confidence and witness accuracy: Assessing their forensic relation. Psychology, Public Policy, & Law, 1, 817-845.
Semmler, C., Brewer, N., & Douglas, A.B. (2012). Jurors believe eyewitnesses. In: Cutler, B.L. (ed.). Conviction of the innocent: Lessons from psychological research. Washington, DC: American Psychological Association, 185-209.
Sharps, M.J. (2003). Aging, representation, and thought: Gestalt and feature-intensive processing. New Brunswick, New Jersey: Transaction Publishers.
Sharps, M.J. (2010). Processing under pressure: Stress, memory, and decision-making in law enforcement. Flushing, New York: Looseleaf Law Publications, Inc.
Sharps, M.J., Barber, T., Stahl, H., & Villegas, A.B. (2003). Eyewitness memory for weapons. The Forensic Examiner, 12, 34-37.
Sharps, M.J., Herrera, M., Dunn, L., & Alcala,E. (2012). Repetition and reconfiguration: Demand-based confabulation in initial eyewitness memory. Journal of Investigative Psychology and Offender Profiling, 9, 149-160.
Sharps, M.J. & Hess, A.B. (2008). To shoot or not to shoot: Response and interpretation of response to armed assailants. The Forensic Examiner, 54-64.
Sharps, M.J., Hess, A.B., Casner, H., Ranes, B., & Jones, J. (2007). Eyewitness memory in context: Toward a systematic understanding of eyewitness evidence. The Forensic Examiner, 16(3), 20-27.
Sharps, M.J., Janigan, J., Hess, A.B., & Hayward, B. (2009). Eyewitness memory in context: Toward a taxonomy of eyewitness error. Journal of Police and Criminal Psychology, 24, 36-44.
Sharps, M.J. & Nunes, M.A. (2002). Gestalt and feature-intensive processing: toward a unified model of human processing. Current Psychology: Developmental, Learning, personality, Social, 21(1), 68-84.
Sporer, S.L., Malpass, R.S., & Koehnken, G. (1996). Psychological issues in eyewitness identification. Mahwah, New Jersey: Lawrence Erlbaum Associates, Inc.
Steblay, N.M. (1992). A meta-analytic review of the weapon focus effect. Law and Human Behavior, 16, 413-424.
Villegas, A.B., Sharps, M.J., Satterwaite, B., & Chisholm, S. (2005). Eyewitness memory for vehicles. The Forensic Examiner, 14, 24-28.
Wells, G.L. (1993). What do we know about eyewitness identification? American Psychologist, 48(5), 553-571.
Wells, G.L., Lindsay, R.C.L., & Ferguson, T.J. (1979). Accuracy, confidence, and juror perceptions in eyewitness identification. Journal of Applied Psychology, 64(4), 440-448.
Wells, G.L., Memon, A., & Penrod, S.D. (2006). Eyewitness evidence: Improving its probative value. Psychological Science in the Public Interest, 7(2), 45-75.
Correct, Useless, and Incorrect Responses for Gestalt and Feature-Intensive Questions
About the Authors
Publisher, Dr. Robert O'Block, American College of Forensic Examiners.