Slow but steady progress toward improving eyewitness identification

In the wake of a landmark decision from the New Jersey Supreme Court mandating blind, sequential police lineups in that state ("blind" meaning the lineup administrator doesn't know who is the suspect, "sequential" meaning photos are shown one at a time rather than in a group), a new study adds to the growing body of evidence that police eyewitness identification procedures need reform. Via AP:

A new study says those lineups you see on television crime dramas and often used in real-life police departments are going about it all wrong.

The study released Monday by the American Judicature Society is part of a growing body of research during the past 35 years that questions the reliability of eyewitness identifications under certain circumstances. That research has been taken more seriously in recent years with the evolution of DNA evidence clearing innocents of crimes they were convicted of committing, often based on eyewitness testimony.

The new study finds witnesses should not look at a group of people at once to pick a perpetrator. Instead, they should look at individuals one-by-one with a detective who doesn't know which is the real suspect — known as a double-blind lineup to avoid giving witnesses unintentional cues — preferably on a computer to ensure appropriate random procedures are used and to record the data.

The study found witnesses using the sequential method were less likely to pick the innocents brought in to fill out the lineup. The theory is that witnesses using the sequential lineup will compare each person to the perpetrator in their memory, instead of comparing them to one another side-by-side to see which most resembles the criminal.

"What we want the witness to do is don't decide who looks most like the perpetrator, but decide whether the perpetrator is there or not," said Gary Wells, an eyewitness identification expert at Iowa State University and the project's lead researcher.

You can view a copy of the study here (pdf) and listen to audio of a press conference releasing the study results here (5mb wav file). The sequential method in particular significantly improved accuracy. In the study, "fillers" (i.e., non-suspects) were wrongly identified 12% of the time in sequential lineups compared to 18% in photo arrays presented as a group, according to press accounts. Austin PD was one of four departments participating in the inquiry.

Texas law enforcement agencies will soon have an opportunity to implement this advice. (Dallas PD was an early adopter of the improved methods.) This year the legislature finally passed a statute which will eventually lead to all Lone Star law enforcement agencies developing policies for eyewitness identifications. (A whopping 88% have no written policy, found a 2008 study.) The new law ordered the Law Enforcement Management Institute of Texas at Sam Houston State University to develop a model policy that departments could (but don't have to) follow, and LEMIT has convened a working group this Friday in Huntsville to give input to the academics charged with developing it. That model policy must be completed by the end of the year, then departments have until Sept. 1, 2012 to adopt their own.

(In the interest of full disclsure, I lobbied for the new statute on behalf of the Innocence Project of Texas as part of my day job, and will be participating in the LEMIT working group on Friday.)

In the New Jersey Supreme Court ruling, reported the New York Times:

the court strongly endorsed decades of research demonstrating that traditional eyewitness identification procedures are flawed and can send innocent people to prison. By making it easier for defendants to challenge witness evidence in criminal cases, the court for the first time attached consequences for investigators who fail to take steps to reduce the subtle pressures and influences on witnesses that can result in mistaken identifications.

“No court has ever taken this topic this seriously or put in this kind of effort,” said Gary L. Wells, a professor of psychology at Iowa State University who is an expert on witness identification and has written extensively on the topic. 

The notable flaw in Texas' statute is precisely that the Legislature failed to enact "consequences for investigators who fail to take steps to reduce the subtle pressures and influences on witnesses that can result in mistaken identifications." If departments don't follow their own, written policies, defense lawyers can raise the issue in court but there's no pretrial hearing required, tainted evidence would not be excluded (Governor Perry's folks had said the bill would be vetoed if the exclusionary rule applied), and thanks to (IMO bizarre) opposition by the criminal defense bar, there isn't even a jury instruction if the cops don't follow their own rules.

Even so, Texas' new statute should improve the process in most cases once local departments have adopted written policies. Police won't want to be embarrassed in court or the media for not following their own procedures, and simply training on improved practices and putting the policies in writing should cause most officers to (eventually) acquiesce and embrace the new approach. If departments do routinely flout lineup policies, there's always the option of going back to the Lege down the line to create a remedy for violations.