Unfair

by Adam Benforado

  This seems backwards—we would expect that fighting off an attacker would make us better at remembering that person, but that’s just not the case. One implication is that when police officers and other witnesses make errors in recalling particular details after a strenuous encounter with a criminal, we need to resist our gut instinct that they are lying or trying to protect the perpetrator. It’s likely that they just don’t remember.

  —

  As influential as it is, the legal system usually has no control over the conditions in which we encode our memories. If a witness is far away or standing in the dark at the time of the shooting, there is nothing a police officer, prosecutor, or judge can do. Yet these officials have significant command over the period when a memory is retained and later retrieved. And the processes and practices we use to handle eyewitnesses are often just as critical to identifying the true perpetrator and learning what really happened.

  It’s strange, then, to discover how haphazard and variable the approaches to eyewitnesses are across the twenty thousand or so law enforcement departments in the United States. A large majority of officers have not received any formal training on how to properly interview a witness or conduct an identification, and many police departments lack guidelines altogether. Those that do often employ faulty procedures that seriously imperil the likelihood of a just and effective investigation.

  The flawed handling of the victim in White’s case was evident from the start.

  It is routine stuff in any TV crime drama. Frequently, when the police don’t have leads, they first bring the victim in to help them come up with a composite image of the criminal. The picture is then put up around town and circulated to officers, and eventually the perpetrator is recognized (just in time for the credits to roll).

  The real-life rape case followed the script. When a Georgia Bureau of Investigations agent saw the sketch of the rapist, he thought it looked a bit like someone he was investigating on an unrelated burglary: John Jerome White. Great story, except that White wasn’t the perpetrator.

  And, in fact, the overall track record of composites is weak. Those who view a composite image are only slightly better at picking a suspect out of a lineup than someone guessing randomly. That wouldn’t be so bad if the creation of the composite didn’t have adverse consequences in itself. But the very process of working with an artist or computer technician to construct an image may actually alter the witness’s initial memory. In certain instances, the composite may even come to replace the initial memory.

  In White’s case, any distortion of the victim’s memory was almost certainly exacerbated when the police brought her back in to look at a photographic array.

  On the positive side, photo arrays (or photographic lineups, as they are sometimes called) are more accurate than the most common identification procedure—the “show-up,” in which the police ask the witness to observe a single suspect, often at the crime scene. Cops like show-ups because of their speed and convenience, but they are highly suggestive and often lead to erroneous identifications: when you see someone sitting on the curb handcuffed, the clear implication is that the police must be fairly certain that he is the culprit. The array, which provides simple headshots of multiple people, is an obvious improvement, but it brings its own serious problems.

  For one thing, when witnesses are shown a photo array, they naturally assume that the perpetrator is included. After being told that they chose an innocent filler, over half of witnesses will go ahead and pick out someone else. The good news is that if you expressly tell a witness that the perpetrator may or may not be present, this stems the tendency, but police officers routinely fail to offer any caution at all.

  Furthermore, showing multiple photographs on a single page just encourages witnesses to choose the person who looks most like the perpetrator. There is an easy solution here, too, but it’s often ignored: presenting images sequentially forces the witness to forgo a relative judgment and to determine instead whether the particular person in front of him corresponds with his memory.

  Perhaps the biggest problem with the photo array employed in White’s case was simply that it occurred so long after the attack—some six weeks later. Time is the enemy of eyewitness accuracy, so if you go into the police station the night you witness a crime, you are likely to be far more accurate than if you wait a month or more. Indeed, the most precipitous drop in accuracy appears to occur after about seven days. In one study, witnesses who identified a perpetrator a month after an incident were only half as accurate as those who did so after just a week.

  Of course, the photo array was not the only identification procedure the police used in White’s case; a week later, the victim was brought back to do a live lineup. Although it’s less common than some of the alternatives, the live lineup tends to be the most accurate eyewitness identification tool. A lot depends on how the lineup is constructed and administered, however, and in White’s case there were major problems.

  One of the basic principles of creating a valid lineup is ensuring that the people in it all match the initial description provided by the witness so that the suspect does not stand out. Lineups in the United States tend to have five fillers, and when only some of the fillers match the initial description, those who do not match become irrelevant. So a lineup may really present only two or three potential perpetrators—or sometimes just one.

  In White’s case, the victim initially described her attacker as a stocky, well-built man with a round face. She said he had short hair and was clean shaven. Flip back to the photograph and you immediately notice a problem with the lineup: White has a skinny body and narrow face and could never be called well built. Moreover, his hair is fairly long, and he has a mustache. The two men shown to the left of White and the first man to the right are also slender, with longer faces. Indeed, the only man who seems to meet the victim’s initial description is the man on the far right—the true perpetrator, James Edward Parham. In a very unusual twist, however, this deeply flawed lineup did not focus attention on Parham because a completely different police error produced an even more powerful distortion: the victim had seen and picked out an image of White just one week earlier.

  There is substantial evidence that people regularly confuse individuals in their memories, in part because they lose track of the source of a recollection. The process of retrieving memories, remember, is a constructive process: sometimes we cut out someone’s face or body and paste it into a completely different setting. This unconscious transference can arise without any police intervention at all; a witness might point the finger at someone who was merely a bystander at the scene of the crime, or someone who worked at the McDonald’s where the witness ate that morning. In fact, there is some evidence that merely seeing a person’s image previously on a social media site can corrupt an identification procedure.

  More often the problem arises when police departments ask a witness to view a suspect multiple times. Experimental evidence confirms that having come across a person’s face in a book of mug shots significantly increases the likelihood that a witness will select that person at a later lineup, even when the actual perpetrator—someone else—is present.

  The use of multiple identification procedures is generally born of good intentions: the police really want to make sure they have the right person. In White’s case, the second procedure was organized because the victim was not completely sure of her initial identification with the photo array, and the police were worried that perhaps the photo did not actually look like White. What they did not realize is that by making White the only person who appeared in both the photo array and the lineup, they were actually increasing the risk of misidentification. The victim may have felt considerably more certain when seeing White in the flesh—in fact, she stated as much—but exactly what was she remembering? The face of the man who brutally attacked her in her home almost two months before, or the face of the man in the photograph she had picked out a week earlier?

  As a cas
e like this proceeds, a witness commonly becomes more and more certain of her identification, but that certainty often fails to correspond to a rise in accuracy. That is because a witness’s level of confidence may be inflated by factors that are completely unrelated to the memory itself. When a police officer offers a supportive statement like “Nice work—you chose the guy we picked up,” witnesses tend to express less doubt in their identification and greater confidence that they saw the perpetrator clearly.

  In fact, of all the factors we’ve discussed, the comments made by other people may have the biggest impact on a witness’s memory. After a witness observes a crime, she is often exposed to additional information about what happened. Some of this information may come from news reports or chatting with other witnesses; other details may be picked up from interactions with the police. There is, of course, the potential that this additional information may make a person’s memory more accurate, but often it appears to be the other way around.

  In one recent study, participants were shown slides of a man taking a woman’s wallet and then concealing it in his jacket pocket. After looking at the photos, participants listened to a description of the slides that contained misleading information like “Then the man hid the wallet in his pants pocket.” When participants were asked to recall details from the pictures they had seen, a significant number made errors based on the misinformation, including claiming that they had observed the man hide the wallet in his pants.

  Even a single word choice by a police officer can alter an eyewitness’s memory. In another study, two groups watched a film of a car accident and were then asked to estimate how fast the vehicles were traveling when they either “smashed” or “contacted.” When the word smashed was used, the average estimate was nine miles an hour faster. Participants in the “smashed” group were also more inclined to recall shattered glass at the scene, although there wasn’t any in the actual footage.

  Young children and the elderly appear to be particularly susceptible to misinformation effects, as are those with low IQs, those who are sleep-deprived, and those strongly driven to please others.

  Perhaps most shocking is that suggestion can create false autobiographical memories. A number of studies have shown how easy it is to plant false memories—everything from receiving a painful enema to meeting Bugs Bunny at Disneyland (which is impossible, because Bugs is a Warner Brothers character). One group of researchers found that they could get 50 percent of participants to remember all or part of a childhood hot-air balloon trip that never happened after showing the person a doctored photo of the event along with other real photos from the person’s youth.

  The research is sobering. And it makes the manner in which officers interact with witnesses critical. Both verbal and nonverbal cues can jeopardize the accuracy of witness statements and identifications, regardless of the good intentions of the police or the awareness of the witness.

  Imagine you’re a cop administering a photo array. The witness begins to say something like “I’m pretty sure it’s this guy,” pointing to the photograph of someone who is unquestionably innocent. You know how much harder it will be to clear the case if she picks the wrong person, so you respond, “We are in no rush here. Take your time and just be sure you’re certain.” You haven’t told her whom to pick or not pick, so there’s no problem, right?

  It seems like good advice and harmless, but this is just the type of shepherding that can encourage a mistaken identification. And a simple cough, sigh, or gesture may be just as powerful. In one study, when an experimenter merely stroked his chin, people were three times as likely to report that a man in a video they had watched had a beard, when in fact he was clean shaven. In most cases there is going to be no record of a suggestion this subtle: the witness is nudged, and no one will ever know.

  Interviews are especially fraught: it’s difficult to conduct one without influencing the witness. Since most officers have very little training in the interview process, they often end up employing approaches that hinder the task of recovering complete and accurate memories. Among other things, they fail to gain a rapport with the witness, they interrupt too much, and they ask questions that are either highly suggestive or too narrow. When they reach a dead end, they often continue to probe, urging the witness to work a bit harder to uncover the details that must certainly be there. This may be effective on occasion, but in many cases it can lead the witness to remember things that never happened—either spontaneously, in an effort to fill in missing details, or because of specific information provided by an interviewer (“Do you remember a smell of smoke on his clothes?”). And the more that detectives repeat questions and conduct multiple interviews in hopes of cracking a case, the more confident the witness grows in her memory, regardless of its accuracy.

  —

  These distortionary factors might not be so harmful if the ultimate decision-makers in criminal cases were aware of them and could adjust their expectations accordingly. However, research suggests that most judges and potential jurors place too much faith in witnesses’ memories. And they are often oblivious to the things that can reduce accuracy and reliability, which makes them bad at determining how much weight to give to, say, a particular eyewitness identification.

  Compounding the problem is that not only do jurors fail to appreciate the significance of poor witness viewing conditions or biasing lineup procedures, but they also depend on dubious factors to decide whether an identification or account is accurate. For instance, jurors are two to three times as trusting of extremely confident witnesses as they are of less confident ones, and these impressions affect their behavior: the more confident the witness, the more likely it is that a jury will convict. When a woman like the victim in the White case has the courage to come forward to identify the man who raped and beat her and says she’s certain, we feel assured that we have the right man.

  Jurors appear to be particularly impressed by in-court identifications, in which they can actually observe the drama of the perpetrator being pointed out. Yet these identifications might be the most troubling of all: in many cases, the witness has already viewed the defendant multiple times and is under enormous pressure to be consistent. Had the victim in the White case stepped out of the witness box and failed to point out her attacker, not only would she have seemed incompetent, but she might also have derailed the whole case, undermining weeks of work by the police officers and lawyers trying to help her receive justice. Prosecutors commonly have eyewitnesses carry out this farce on the stand because they know how convincing such performances are to the jury. And judges almost never intervene because they operate with the same mistaken notions about memory. They believe that standard legal tools like the defense’s ability to cross-examine eyewitnesses is sufficient safeguard, without realizing that such a tool may be largely useless—most witnesses have no idea that their memories have been corrupted.

  —

  In his 1908 book, On the Witness Stand, the pioneering psychologist Hugo Münsterberg recounted a famous meeting of a German “scientific association, made up of jurists, psychologists, and physicians.” On the night of the convention, there happened to be a public festival out in the street. All of a sudden, in the midst of the proceedings, the doors to the hall were flung open and “a clown in highly coloured costume” burst in, followed by a man carrying a revolver. The men shouted at each other and began to wrestle, at which point the gun went off. A few moments later the intruders ran out of the room. Given that a criminal investigation was a near certainty, the president of the association asked the forty people in attendance to carefully write down, individually, exactly what they had seen. Unbeknownst to them, the entire scene was an experiment in eyewitness observation and memory, staged by the president.

  The results were disheartening: the statements were filled with false details and omitted numerous key facts. According to Münsterberg, when considered alongside his own memory experiments on Harvard undergraduates, the poor showing served as a “w
arning against the blind confidence in the observations of the average man.” Indeed, “in a thousand courts at a thousand places all over the world, witnesses every day affirm[ed] by oath…mixtures of truth and untruth, combinations of memory and of illusion, of knowledge and of suggestion, of experience and wrong conclusions.”

  More than a century later, that indictment still feels fresh. Look what happened after John Jerome White was finally exonerated in 2007. In the direct aftermath, there was a push in Georgia to seriously reform the eyewitness identification procedures across the state with best practices and mandatory officer training. But other than the creation of a voluntary training program for police officers handling eyewitness identification procedures, significant research-based reforms have not been adopted. In the vast majority of jurisdictions across the country, the story of entrenchment is no different, and the Supreme Court has failed to intervene meaningfully at the national level.

  Of course, some progress has been made in the decades since Münsterberg first applied psychological insights to the law. In the last thirty years, more than two thousand studies have been published on eyewitness identifications alone. With more rigorous scientific methods and a greatly expanded understanding of memory, we are now in a much better position to make lasting changes in the way our criminal justice system approaches eyewitnesses.

  The first step is to convince the majority of those inside and outside the legal system that there is a problem and that justice demands we address it. As Münsterberg put it, “If the time is ever to come when even the jurist is to show some concession to the spirit of modern psychology, public opinion will have to exert some pressure.”

  We need to start thinking about witness mistakes in the way that we think about other types of errors that individuals, groups, and institutions make. As I mentioned, studies of actual police lineups show that eyewitnesses select innocent people more than 30 percent of the time. Would we as a society tolerate the sale of a car whose brake lights malfunctioned on every third trip, or a hospital that handed out the wrong medicine to every third patient? Obviously not; we would demand immediate change. So why do we accept the claim that the legal system works just fine as it is?