A composite picture of a robber who hit a New Castle jewelry store in 1988 caused a local police officer to suspect a well-known criminal named William Brothers.

	Alyssa Cwanger, Post-Gazette In a photo provided by his family, William Brothers stands outside the Lawrence County Courthouse on Aug. 19, 1996. Brothers has been incarcerated for 16 years and claims he is a victim of false identification by an eyewitness to a jewelry store robbery in New Castle. Click photo for larger image.

There was only one problem. Even though the composite -- a likeness made by matching facial parts to a witness's description -- evoked Brothers in the officer's mind, a witness to the robbery couldn't pick Brothers' photo out of a lineup afterward.

But police weren't done. Convinced they had the right man, they cobbled together another photo lineup using a different picture of Brothers, and this time, the witness identified him.

That process led to Brothers' conviction and earned him more than two decades in prison, despite the fact he now has a signed confession from a similar-looking man who said he committed the crime.

In North Philadelphia in 1981, an eyewitness to the robbery and killing of a newsstand worker quickly identified two of the three perpetrators, but couldn't say who the third person was.

Then police convinced the witness that a man called "Junie," the nickname of George Booker, was the third killer.

Twenty-three years later, that same witness says police coerced him into naming Booker, and three other people have emerged to say Booker is innocent, yet he sits in prison with only a distant hope of a successful appeal.

Those are just two of the hundreds of cases in Pennsylvania examined by the Innocence Institute of Point Park University in which the convictions were based on witnesses identifying the suspect through photos.

The institute is a joint venture with the Pittsburgh Post-Gazette that probes allegations of wrongful convictions while helping student journalists learn investigative reporting.

In its survey of eyewitness cases, the institute found that in nearly every one of them -- whether there was evidence of the person's innocence or not -- there was a widespread failure to follow procedures that prevent false identifications.

That not only can convict innocent people, but can allow guilty people to go free, says Barry Scheck, founder of the Innocence Project at the Cardozo School of Law at Yeshiva University in New York.

Scheck, best known as one member of O.J. Simpson's legal defense team, held a conference at the law school last fall as part of his effort to get law enforcement officials to embrace policies recommended by the U.S. Department of Justice and social scientists to reduce false eyewitness identifications.

In the same way that he has pioneered the use of DNA evidence to exonerate the innocent, Scheck argues it is good police policy to adopt science-based reforms to prevent wrongful convictions.

"There's no reason why we can't make the whole criminal justice system more scientific," he said.

Cases riddled with problems

The flaws in the two cases cited above were numerous.

In both the Brothers and Booker cases, police showed witnesses photos arrayed on a single sheet, increasing the chances the witnesses would guess and pick an innocent person.

The Booker case shows what can happen when police are allowed to deliberately influence a witness to identify someone. The Brothers case is an example of two more errors: Police based their initial identification on unreliable composite pictures; and they then compounded the mistake by giving a witness more than one chance to see the suspect's photo in a lineup.

After Hughes Jewelry in New Castle was robbed on the evening of Dec. 14, 1988, one of the clerks helped police make the composite picture.

The Justice Department warns that composites shouldn't be used as stand-alone evidence. Social scientists like Gary Wells take the criticism a step further.

	Gary L. Wells, of Iowa State University, says changes in eyewitness policies are needed in the criminal justice system.

"Composites are creating more problems for the legal system than they can possibly ever solve," he said. Wells, a psychology professor at Iowa State University, is a leading expert in the science of memory and identification and helped construct the Justice Department's guidelines for handling eyewitness i.d.'s.

Composite programs allow eyewitnesses to select facial features and combine them into a face using a kit or a computer. Research shows that composites not only tend to produce poor likenesses of suspects, but an experiment by Wells showed composites harm a witness's memory by replacing the actual face he saw with a mental image of the composite, reducing the chances the witness can later identify the real suspect.

The New Castle officer thought he recognized Brothers from the composite because Brothers had a long list of run-ins with the law, including previous convictions.

Drawing a blank

But when the female clerk in the jewelry store failed to pick Brothers' picture out of a photo lineup, police claimed the first picture had been blurry and then constructed a second lineup with a new picture of him. That time, she identified him and he was arrested.

Giving a suspect two shots at making an identification flies in the face of basic science, according to experts.

"There's only one chance to get it right," Wells said. Eyewitness identifications are as sensitive as physical evidence such as blood and can be contaminated, he said.

Once a witness's memory is tainted by the initial viewing of a suspect's picture, it becomes increasingly likely the witness will pick that suspect in a second viewing.

At the trial, the owner of the store, who chased the robber and fired his handgun at him but could not identify anyone in the photo lineup, joined his clerk in providing a positive i.d. of Brothers. Brothers was convicted and sentenced to 22 years in prison.

Then, two years after the crime, a man who looks so similar to Brothers that the convicted man's own mother once confused the two admitted in a notarized statement he was the actual robber.

But that man refused to testify at Brothers' appeal for fear of incriminating himself and later claimed he had signed the confession under pressure.

Thanks to additional drug and escape charges tacked on since his original conviction, Brothers will not be released until 2025, when his maximum sentence is over.

If he would comply with the Pennsylvania Parole Board's release prerequisite of taking responsibility for the crime, he could be paroled next year. He refuses.

"No. 1, why would I admit to something I didn't do?" he said by phone from the State Correctional Institution at Mercer. "No. 2, why would I admit to something that someone else admitted to doing?"

While Justice Department guidelines permit side-by-side photo lineups like those used in the Brothers case, Iowa State's Wells said his studies show they encourage witnesses to make a guess, even if they are not sure the person they select is the right one.

Wells said showing photos one at a time is a better method, and shouldn't stop a good eyewitness from making an identification.

"That doesn't strike me as that difficult a task for a witness to pass," he said.

Police in New Jersey have adopted the sequential approach, and officials say officers have embraced the method as a simple change that required little additional effort.

The Boston Police Department also has begun using the sequential method and detectives in Chicago are now using it on a trial basis.

Ken Patenaude, detective bureau commander of the Northampton, Mass. police department, serving a town of 32,000, said detectives there initially resisted the change to sequential lineups. They were more comfortable with the old procedure, he said, and felt the change amounted to questioning their integrity.

"We hadn't changed the way we do eyewitness procedures in decades," he said, but "DNA [exonerations] obviously have shown us that we have to change."

Using an independent officer

Most departments that have accepted sequential photo lineups usually have paired them with a second, more important reform in the way photo lineups are assembled -- the practice of having an independent person who doesn't know the suspect present the photo arrays.

Throughout Pennsylvania and much of the rest of the nation, photo lineups are still administered by detectives running the investigations.

That means the person presenting the lineup usually knows the identity of the suspect. Wells said this can lead detectives -- whether they are unethical or not --to influence a witness to choose that person.

Sometimes the prompting is blatant, such as showing a darker-skinned suspect's picture with a set of lighter-skinned people's photos.

Sometimes the detective gives verbal clues, and sometimes he uses body language.

The 1981 Philadelphia case that ended up convicting George Booker of homicide involved those kinds of suggestions.

Paul Lehmen was beaten, robbed and shot by three people after he left his newsstand on the night of Jan. 20, 1981. An eyewitness identified two of the men, but couldn't identify the third.

Yet 30 hours later, the witness changed his story and identified Booker from a police photo array as the third man, which led to his conviction and a life sentence.

Booker had a juvenile burglary record, but had never been arrested as an adult.

Twenty years later, the eyewitness was involved in a serious car accident and decided he had been given a chance to make amends for his past mistakes. He came forward to say police threatened him and insisted he identify Booker, a man he did not know.

Change requires more work

To prevent police influencing a witness's selection from a photo lineup, experts such as Wells propose that photo lineups should mimic scientific experiments.

In a so-called double-blind drug test, for instance, neither the person conducting the experiment nor the test subject knows who is getting the real drug and who is getting a placebo.

In a similar way, the person conducting a photo lineup shouldn't know who the suspect is, he said. The investigating officers can put together the photo array that includes the suspect's picture, but the person administering the lineup wouldn't know which picture is the suspect's.

While the Justice Department doesn't take a stand on the issue, police in New Jersey were the first to adopt this technique, and police in the Minneapolis area are among those who have done so since then.

Many police departments resist this change for a simple reason: It involves more work.

New Jersey police had to train patrol officers and desk administrators to conduct identification procedures, said Lori Linskey, deputy attorney general of New Jersey. "It was the toughest hurdle to overcome," she said.

In Minnesota, Hennepin County has used both sequential and double-blind lineup procedures for more than a year. Al Harris, the county's senior assistant attorney, said the changes were "a pain in the butt" but proved to be a boon for law enforcement by taking a potential criticism away from defense attorneys seeking to pounce on improper eyewitness procedures.

"We want to be able to show people we have done everything we can to get it right," he said.

In Booker's case, the witness and at least three others insist police got it wrong.

Booker's aunt says she saw him sleeping at home around the time of the crime, but was never called to testify at his original trial. On top of that, two people -- including one of Booker's fellow inmates -- came forward and said in a sworn statement they also witnessed the crime and didn't see Booker there.

Cathie A. Abookire, a spokeswoman for the Philadelphia district attorney, said Booker's claim of a false identification and others like it are matters for the courts.

"If they have some new claim or some new evidence they should invoke their legal rights of review so that their allegations can be heard in a court of law," she said.

Booker has certainly tried. He has appealed unsuccessfully six times.

Despite that, he said he has faith justice will prevail in his latest attempt.

"Because of false eyewitnesses, I have lost the most productive years of my life. But God keeps me strong and hope gives me comfort," he said in a telephone message.

Officials say reforming identification procedures can prevent innocent people from being convicted, as well as helping ensure that juries and appeals courts don't reject good cases because of a badly done identification.

The reforms only strengthen criminal cases, said David Angel, deputy district attorney in Santa Clara County, Calif. "If you don't do this, you risk having good convictions and good identifications thrown out," he said.

Lee Solomon, a deputy U.S. attorney in New Jersey, said the guidelines simply amount to good law enforcement.

"I've never met a prosecutor who didn't want to get it right every time," he said.