Jennifer Thompson wanted to be the perfect eyewitness.

Just two days after the 22-year-old college student was raped in her North Carolina home in 1984, she positively identified Ronald Cotton as her attacker when she picked his picture out of a police photo lineup.

SIGHT UNSEEN FIRST OF A SERIES

Chuck Burton, Associated Press Jennifer Thompson talks with Ronald Cotton in Greensboro, N.C., in 2000. Thompson identified Cotton as the man who raped her in 1984. After serving 11 years of a life sentence, Cotton was released from prison when DNA testing revealed another man had raped Thompson. They have become friends since Cotton's release. Click photo for larger image. The series Day Two: Most local police unaware of witness guidelines Day Three: How shaky eyewitness testimony can lead to jail Related coverage Kirkwood robbery case brings witnesses' memories into question

He insisted he was innocent. But after she identified him again at two trials, he was sent to prison for life.

More than a decade later, DNA testing proved that Cotton wasn't guilty. Someone else had raped Jennifer Thompson. She was dumbfounded. "It's like being told that the children I had were not my children," she said.

Now Thompson is speaking out around the nation about how the unreliability of witnesses' memories can lead to false identifications. She says she's proof that an honest, well-meaning person can send an innocent one to prison.

Her case is a high-visibility example of what an emerging field of science is proving more and more often: Crime victims are often quite poor at accurately remembering what a criminal looked like.

Making matters worse, experts say, is that most police and prosecutors are using techniques that raise the odds that witnesses will identify the wrong person.

According to the Innocence Project at the Cardozo School of Law at Yeshiva University in New York, of the 157 people nationwide it has helped exonerate through DNA evidence, nearly 80 percent were convicted based on eyewitness identifications.

Seven of the 157 exonerations involved Pennsylvania prisoners, including at least four convicted because of eyewitness identifications.

The Innocence Institute of Point Park University, a partnership with the Pittsburgh Post-Gazette where journalism students examine allegations of wrongful convictions, has analyzed hundreds of additional cases in the state in which the convictions were based on witnesses identifying suspects from photos.

In most of them, the analysis found, police failed to follow the steps that can help prevent false identifications.

In other regions, wrongful convictions have helped spur reforms.

In New Jersey, North Carolina and Virginia and a few counties in California, Illinois, Massachusetts and Minnesota, police are implementing federal guidelines on the best way to conduct photo lineups and using scientifically tested techniques to prevent unjust convictions.

In Illinois, Gary Dotson became the first of a series of convicts whose exonerations based on DNA evidence led to changes in law enforcement procedures.

Dotson was convicted of rape in 1979 based on a woman falsely identifying him.

A fabricated crime

	Seth Perlman, Associated Press Gary Dotson spent eight years in prison for a rape he didn't commit and became one of the first Americans later exonerated based on DNA evidence. He was identified from a photo lineup after a woman fabricated the story of a rape that never occurred. Click photo for larger image.

The woman told police she'd been raped, but she'd actually had sex with her boyfriend and made up the rape story after worrying she might be pregnant. She then invented a description of an attacker and helped a police sketch artist create a drawing of him, which police used to prod her to pick Dotson out of a book of police photos because he resembled the picture.

A decade later, he became just the second American exonerated on DNA evidence.

Now police in the Chicago region are throwing out their old practices and testing a new way of arranging and conducting photo lineups.

Police used to show witnesses a sheet of photos and ask them to pick out the criminal, but detectives are now handing the photos to witnesses one at a time. Research shows this approach is more accurate. It encourages the witness to decide whether he saw the person in each picture, rather than "comparison shopping" and picking the picture that most closely resembles the person he saw.

Chicago police also are putting a detective who doesn't know the identity of the suspect in charge of conducting the lineup, borrowing a well-known technique from scientific experiments.

In most experiments that test new drugs, for example, neither the researchers conducting an experiment nor the people participating know who's getting the real drug and who's getting a placebo.

Using this "double blind" technique with police photo lineups ensures that the person running the show won't influence the outcome. If an officer involved in the investigation handles the identification process, he can sometimes prompt a witness to identify that person, if only through body language.

Experts also advise against the use of composite sketches -- drawings assembled from different facial parts -- because tests have shown they often don't resemble the criminals.

A composite drawing led to the conviction of Kurt Bloodsworth, the first death-row inmate exonerated by DNA evidence.

The Baltimore crab fisherman, who had only a minor criminal record, was convicted in the 1984 rape and murder of a 9-year-old girl.

Real killer ignored

One of Bloodsworth's neighbors had identified him as resembling a composite picture made of the attacker. Police were so convinced they had the right man that they ignored another man who had shown up at a mental health clinic with scratches on his face on the day of the rape and told a therapist he was "in trouble with a little girl." That man turned out to be the real killer.

AP file photo Ronald Cotton, right, is shown in this 1984 police photo. Cotton was convicted of the rape of Jennifer Thompson, but DNA testing eventually showed the real rapist was Bobby Poole, shown at left.

Bloodsworth said his experience spending nearly a decade in prison shows the fallibility of relying on eyewitness memories to convict people.

"Your eyes will tell you a lie, and we've got to start telling the truth," he said.

Experts say there are other simple changes in identification procedures that police should make.

Those conducting photo identifications should tell witnesses that the criminal may or may not be in the group of photos they're about to see.

One study found this cautionary instruction resulted in 25 percent fewer cases of false identifications with no loss in correct identifications, according to Nancy Steblay, a psychology professor at Augsburg College.

And if the witness says one of the photos is the criminal, he should be asked how confident he is in that choice, and his answer should be recorded. Police should not give the witness any feedback until this confidence report has been taken.

The number and types of photos used in a lineup also are important.

The Justice Department recommends using at least six photos, and stresses the need to match the photos to descriptions made by witnesses and avoid any photos that stand out too much from the others.

Andy Taslitz, a Howard University law professor who has helped the American Bar Association craft recommendations on eyewitness procedures, said lineups should contain as many photos as possible.

"There is no magic right number, [but] bigger is better," he said.

David Angel, deputy district attorney of Santa Clara County, Calif., said his office instituted many of these eyewitness reforms in 2001. "Basically, it was not difficult to make the switchover," Angel said, and there was almost no additional cost.

And since defense attorneys recently have started to seize on faulty eyewitness identification procedures at trials, Lee Solomon, a deputy U.S. attorney in New Jersey, said using the guidelines would prevent acquittals of guilty people on technicalities.

When his state adopted the reforms, "I really believed it would help us win more cases."

But there's an even bigger reason to make sure eyewitness identifications snare the right people, said Amy Klobuchar, attorney for Minnesota's Hennepin County.

If the wrong people are identified, she said, "not only are we going to be convicting the wrong people, we're going to have rapists and murderers out on our streets."

Thompson knows that very well. She identified Cotton in two trials before DNA evidence proved another man had raped her.

That man, Bobby Poole, was free for nine months after he raped her, until he was charged with other rapes and sent to prison for life, where he bragged about having attacked Thompson, too.

Although she now knows Cotton is innocent, Thompson said her experience showed how picking the wrong photo from a lineup can forever alter the victim's perception of the crime.

"I know Ronald's an innocent man, but I still see his face in my nightmares," she said.