5
Methodological Considerations in Eyewitness Identification Experiments
Adele Quigley-McBride and Gary L. Wells
Eyewitness identification experiments have been invaluable to our understanding of the conditions under which eyewitnesses can be relied upon to provide accurate information. After decades of research, we have a reasonable understanding of what makes eyewitnesses reliable and what processes can produce erroneous or correct identification decisions. In this time, researchers have also established, modified, and perfected various methodological features of the eyewitness experiment.
One reason for the success of eyewitness research is that policy makers and police can readily grasp the importance and implications for criminal investigations in the real world, because the experiments closely mirror the situations that they represent in the field. In fact, a police investigator conducting a lineup is actually conducting a procedure that is very similar to an experiment. The police investigator, like the experimental psychologist, is testing a hypothesis with their lineup procedure—that their suspect is the person who committed the crime. By asking an eyewitness to view a lineup containing the suspect tests this hypothesis by testing the reaction of the eyewitness, similar to how an experimenter might submit participants to an experimental protocol to observe behavior (see Wells & Luus, 1990 for a detailed treatment of this analogy). Ecological validity and the ability to explain the relevant concepts are only part of what makes an experiment useful—the methods need to be carefully considered to ensure they produce robust, reliable, and meaningful results.
Although eyewitness identification experiments can seem simple, there are many factors to consider behind the scenes, which can change the results and their implications in important ways. Thus, one goal of this chapter is to help researchers who are new to eyewitness identification research. As with planning any type of research, researchers planning an eyewitness identification experiment must make a large number of decisions about the design, materials, and procedure, which is a product of a complex series of decisions (researcher “degrees of freedom”; Gelman & Loken, 2013). We focus on the potential consequences of choices made during the research process. In general, we do not advocate for hard and fast rules for any decisions, with only a few exceptions.
Though this chapter is designed to aid the novice eyewitness researcher, we also present some more contemporary controversies and considerations that should be considered by even seasoned researchers. The chapter is organized according to the decision points that are unique to eyewitness identification research (see Brewer, Weber, & Semmler, 2005 for how these map onto the experience of a real eyewitness). The chapter begins with how to create and present a “witnessed event” in the experiment and important considerations therein, such as the importance of stimulus sampling, avoiding nested samples, and ensuring that the advertisement for the experiment and the task instructions do not signal to participants what the purpose of the study is. Next, we describe the decisions that researchers must make when building their lineup materials, such as how to mimic the concept of an innocent suspect in your study and how to select such a person, as well as how to choose fillers (known-innocent people that look like the culprit or suspect, but who will not be prosecuted). After discussing considerations in constructing the materials for the study, we discuss how best to report eyewitness identification data so that other researchers can adequately assess the relevant findings.
This chapter presents approaches to the questions that arise when conducting eyewitness research, and we present our best advice on each matter. However, as mentioned in previous chapters of this kind (e.g., Wells & Penrod, 2011), it is important to note that what is “best” will vary based on the research question. Thus, the information contained in this chapter should be considered in the context of the particular problem that each experiment is designed to address.
Creating the Witnessed Event
The first issue to consider is the creation of the to-be-witnessed event. Lineups occur frequently in the field and can be studied directly (Horry, Memon, Wright, & Milne, 2012), but researchers conducting laboratory-based eyewitness identification experiments know “ground truth,” i.e., the person who is the actual culprit. This is a crucial difference between field studies and lab studies—researchers create the witnessed events used in lab experiments and, hence, know with certainty everything that the eyewitness saw as well as whether the culprit is in a lineup presented later or not. Moreover, unlike in actual cases, researchers can systematically manipulate witnessing and testing conditions in experiments, control for any potential influences that are not of interest, and isolate cause and effect relations among variables. Hence, it is extremely important to carefully consider how to construct the witnessed event in light of the research purpose.
Decision: Should It Be a Live Event or a Video-Recorded Event?
The witnessed events have taken many forms. In early eyewitness identification experiments, most researchers used live staged crimes, such as thefts (e.g., Wells, Lindsay, & Ferguson, 1979). In these live-event experiments, participants were not expecting to be witnesses and believed the event to be real at the time (as would be the case with a crime in the real world). Sometimes, participants were led to believe that the crime was real until they had finished viewing the lineup and making their identification decision (e.g., Murray & Wells, 1982). Although live-staged crimes benefit from ecological validity, there has been a systematic shift in favor of using videos of simulated crime events instead of live events. Some researchers are still using live-staged crimes in which the participants believe that crime was real when making their identification decisions (e.g., Eisen, Cedré, Williams, & Jones, 2018), but this is now quite rare.
The shift to simulated-crime videos has almost certainly been driven by practical and methodological considerations. The amount of time and money required to run these experiments is vastly reduced when a video is used rather than a live event. A live event will also not occur in exactly the same way each time, so each group exposed to the live event may receive slightly different information, whereas a video event can be repeated in exactly the same way for all participants. Moreover, finding a setting where a crime event can be performed repeatedly for witnesses can be very difficult and sometimes is fraught with ethical concerns. Video events remedy some of these shortcomings in addition to offering more experimental control and easier data collection, but suffer a significant loss in experimental realism. In addition, people remember more details from a video-recorded event than they do a staged event, and thus this decision is not without consequences (Ihlebæk, Løve, Eilertsen, & Magnussen, 2003). Though this tradeoff seems worthy of debate, the shift away from live events to video events occurred without formal discussion in the literature. Thus, eyewitness scientists have largely accepted the use of video events, resulting in little motivation to use live events as the easier alternative is well-regarded in the field (Wells & Penrod, 2011).
Decision: Should People Participate in Person or Online?
Another recent shift involves the use of online data collection methods. That is, some eyewitness identification researchers have begun to use a method in which the simulated crime and lineup are viewed online rather than in a controlled lab setting. Commercial services such as Amazon’s Mechanical Turk (MTurk) have made this a particularly easy and inexpensive alternative to community-based or undergraduate samples. An advantage of online methods is that a more diverse sample of participant-witnesses can be obtained (Buhrmester, Kwang, & Gosling, 2011). In addition, ...