The eyes of Troy Anthony Davis as he entered a courtroom on January 16, 1991. (AP Photo/Savannah Morning News)
 
“We see what we want to see,” my grandmother used to say. This insight visited me recently after I ran across the mall chasing a woman I thought was my cousin. It wasn’t, as it turned out, but I didn’t realize that until after I had puffed up behind her, bopped her amiably on the shoulder and cried out, “Boo!”

How was it possible, I thought in retrospective embarrassment, to so wrongly misidentify someone I know so well? Empirically my experience was all too common. I’d been thinking about my cousin a few moments before and saw the woman through the lens of those thoughts. We often project our life’s associations onto the faces of strangers. Constantly—if mostly unconsciously—we familiarize them with learned stereotypes. If we are wise, we learn to take caution with our assumptions. We recognize this innate fallibility, and most of the time it doesn’t matter very much.

Oddly enough, however, we reverse that supposition in the one context where fallibility matters most: in criminal cases, eyewitness testimony is viewed as the ne plus ultra for the prosecution, despite a century’s worth of psychological and sociological studies revealing that, from Sacco and Vanzetti to Troy Davis, witnesses misperceive a startling percentage of the time. “Human beings are not very good at identifying people they saw only once for a relatively short period of time,” writes Cornell law professor Michael Dorf. “The studies reveal error rates of as high as fifty percent—a frightening statistic given that many convictions may be based largely or solely on such testimony. These studies show further that the ability to identify a stranger is diminished by stress (and what crime situation is not intensely stressful?), that cross-racial identifications are especially unreliable, and that contrary to what one might think, those witnesses who claim to be ‘certain’ of their identifications are no better at it than everyone else, just more confident.”

The costs of this phenomenon are perhaps best revealed in data compiled by the Innocence Project, which has concluded that out of 281 postconviction exonerations secured through DNA in the United States, eyewitness misidentification “was a factor in 75 percent…making it the leading cause of these wrongful convictions.” Luckily, there are substantiated ways to guard against such error. Experts have cited two main types of variables that can adversely affect eyewitness identification: “estimator variables,” the hardest to control for, which include things like the degree of lighting, distance or speed within a given crime scene, as well as the level of trauma to the witness; and “system variables,” defined as “those that the criminal justice system can and should control,” which include law enforcement tools like lineups and photo arrays. A number of reforms involving the latter have the proven capacity to boost the accuracy of witness IDs. These include “blind administration,” where an officer conducting a lineup is not aware of who the suspect is (and thus not capable of revealing his or her identity via gestures, vocal inflections or body language); “non-suggestive” lineups, made up of people who generally resemble a witness’s description, so that the suspect does not stand out; allowing witnesses to sign a statement indicating their level of confidence in their choice; and presenting members of a lineup sequentially rather than simultaneously (to mitigate the pressure to choose any kind of close-looking one when we are presented with a bunch of faces at once). Such remedial safeguards have so reduced the error rate—and so indisputably—that a number of local jurisdictions and eleven states thus far have adopted some or all of them as standard operating procedure.

It would seem logical, then, to implement these reforms universally, and for courts to screen eyewitness evidence for those basics of procedural reliability before such testimony is heard by a jury. But on January 11, in Perry v. New Hampshire, the Supreme Court rejected that notion, ruling that such a pretrial inquiry is not a requirement of due process “when the identification was not procured under unnecessarily suggestive circumstances arranged by law enforcement.” This is subtle language: it’s not the same as what we think of as police corruption, as in overt suppression of evidence. Rather, it relates to the kinds of situations at stake in Perry: Was the suspect the only black man in a lineup? Was he handcuffed and flanked by police? Was his image shown in photo array after photo array until he began to look familiar? If the chief investigator was the one administering a lineup, was his belief in the suspect’s guilt communicated to the witness via subtle coaching? All such factors may be highly suggestive, triggering the irrelevant associations and false memories that can lead to inaccurate results.

Perry does two unfortunate things. It undercuts pretrial examination of virtually all “estimator variables,” no matter how problematic, since those are less likely to directly involve police. And by drawing the line at “unnecessarily suggestive” actions by state actors, the ruling sets a very high bar for challenging eyewitness evidence, ignoring the hefty empirical proof that misidentification is a pervasive fact of life. Justice Sonia Sotomayor, the lone dissenter in Perry, wrote that this ruling invites arbitrary results by making “police arrangement” the “inflexible step zero.” The concerns of due process ought to be based on the actual likelihood of misidentification, said Sotomayor, “not predicated on the source of suggestiveness.” Reiterating that any preventable misidentification is a miscarriage of justice—not merely where the police are setting the stage—she underscored the Innocence Project’s concern that inaccurate eyewitness testimony is the leading cause of wrongful convictions in US courts. DNA has exonerated eight misidentified inmates on death row. If we have at our disposal simple reforms that have been proven to guard against such tragic mistakes, why on earth should we not implement them universally?