A police lineup is often the moment of truth in a criminal investigation. It's also, say many experts, highly fallible.
Of the 175 convictions overturned by DNA evidence, 75 percent were convicted largely because of eyewitness testimony that turned out to be mistaken.
Those exonerations have energized efforts to reform the way police conduct lineups and get eyewitness identifications. A growing number of counties and states are adopting measures to improve accuracy and limit influences on witness memory.
Now, though, a first-of-its-kind study from Illinois is casting doubt on a reform called "sequential double-blind." That method shows witnesses photos of potential suspects one at a time, rather than all at once, and even the administrator doesn't know who the suspect is.
The study's results - which suggest the old method was both more accurate and more likely to produce an identification - are a boost to police departments that have resisted lineup changes. Others say the study was flawed, and they worry that it will be used as an excuse to halt all eyewitness-identification reforms. For now, supporters say more study - and more action - is needed, and they hope that a single study won't derail years of effort to improve what they say is a highly flawed system.
"My fear is that the debate over sequential blind will obscure everything, and you'll have police departments who are reluctant to change at all, or not adopt anything," says Barry Scheck, a professor at Yeshiva University's Cardozo School of Law in New York and co-director of the Innocence Project.
Eyewitness reliability is often a hot- button issue, especially in sexual assault cases. Just last week, it arose in the Duke University case in which a stripper has said she was raped by several lacrosse players. She picked two out of a photo lineup, but critics faulted the lineup for containing no fillers, only lacrosse players, likening it to a multiple choice test with no wrong answers.
The Illinois study focused only on the question whether to do sequential blind lineups, a switch that just a handful of jurisdictions have mandated so far. Commissions in North Carolina, Wisconsin, Virginia, and California have recommended that approach, and other jurisdictions are considering it. Many are reviewing the Illinois study closely.
The study took place in three districts: Chicago, Evanston, and Joliet. During the course of a year, police compared the number of times a witness picked out the suspect using the traditional method - in which photos were shown simultaneously, and the administrator might know which is the suspect - with the new one.
Until now, research has shown that the sequential method sets a higher bar for accuracy: the witness compares the photo or person to his memory, rather than to the others in the lineup. Using administrators who are "blind" minimizes the risk that they will convey conscious or unconscious approval once the witness makes his pick - an action that could solidify a formerly hazy memory.
It's a more conservative approach that results in fewer overall identifications, and has raised debate about whether it's better to get more guilty people off the street or avoid a false conviction. "It's a policy decision on how cautious we want our witnesses to be," says Gary Wells, a psychology professor at Iowa State University who has conducted more than 100 experiments on witness memory.
So researchers weren't surprised that sequential lineups in the Illinois pilot showed a lower rate of overall identifications: Witnesses made IDs in 53 percent of lineups, versus 62 percent for simultaneous ones.
But the sequential lineups also had a lower rate of accuracy. Witnesses picked out an innocent person 9 percent of the time, compared to a 3 percent rate for simultaneous ones.
At a Chicago symposium Friday, academics, police, prosecutors, and defense attorneys from around the country heatedly discussed the study and the future of identification practices. Some reform advocates faulted the study for using a blind administrator with the sequential lineups, but not with the simultaneous ones.
"The resistance to the study [from reform advocates] is greater than any resistance from police to trying the sequential double-blind," says Sheri Mecklenburg, director of the pilot program and a general counsel in the Chicago Police Department. "It seems to be a results-driven criticism."
Logistical reasons have kept many cities from trying the new method. Finding a separate administrator isn't always feasible, police say, and dedicating people to administering lineups uses resources that could be better spent getting more police on the street. Some also worry about the fewer number of IDs they get with sequential lineups.
But reform advocates stress that many other, less controversial changes can also make a big difference: finding "filler" candidates who closely resemble the witness's description, for instance, or telling the witness that the suspect is not necessarily in the lineup. They want witnesses to provide "confidence statements," stating how sure they are of the ID. And they want juries informed about the fallibility of eyewitness testimony, especially across racial lines.
Police departments in Wisconsin have been more receptive to such reforms because the state has made them voluntary.
"Any time you mandate something, you're going to have more resistance to it," says Ken Hammond, the state's law enforcement education director.
Boston switched to the sequential method in 2004, after a series of wrongful convictions that involved mistaken eyewitness identification made headlines. Changing to the new model was time-consuming, says David Procopio, press secretary for the Suffolk County district attorney's office, but worth it - an opinion the Illinois study hasn't changed. "We don't believe one part of one study in one state is reason enough to roll back what we consider to be very progressive reforms based on very wide-ranging scientific studies that occurred over a period of many years," Mr. Procopio says.
Experts cite many reasons for faulty witness identification. There's often a desire on the part of a victim to see the person responsible put behind bars. They want to be able to make an identification, and memories can alter accordingly.
Even critics of the sequential lineup suggest that other measures might help improve accuracy. Conducting multiple lineups, some of which have no suspect, might weed out those witnesses who are too willing to pick anybody, says Ebbe Ebbesen, a psychology professor at the University of San Diego in California who says that districts were premature in switching to sequential blind lineups.
Even if most eyewitness identifications are reliable, "we know it is accounting for more wrongful convictions than all the other causes put together," including false confessions, jailhouse "snitches," and outright fraud, says Professor Wells. "When a witness takes the stand and says, 'that's the guy I saw,' that is so persuasive. We need to find ways to prevent mistaken IDs from happening in the first place."