Julius Earl Ruffin knows all too well how inaccurate eyewitness identification can be.
On May 3, 1982, in a Norfolk, Va. circuit court, the 29-year-old was convicted of a rape that he did not commit and was sentenced to five life sentences.
The case rested solely on the testimony of the victim, Ann Meng, a young mother of three who confidently pointed to Ruffin as her assailant.
Twenty-one years would pass before Ruffin was able to prove his innocence with DNA evidence. He details his experiences in the book "Why Me? When It Could've Been You!"
Watch the story on "Primetime" tonight at 10 p.m. ET
Ruffin's story is not unusual. In fact, according to the Innocence Project, a non-profit organization that assists prisoners who can be proven innocent with DNA testing, mistaken eyewitness identifications contributed to more than 75 percent of the more than 200 wrongful convictions in the United States that have been overturned on the basis of DNA evidence.
"Primetime" wanted to see what would happen if we set up scenarios where we asked onlookers to identify the person involved in a purse snatching.
We hired actors to play the role of the victim and the thief. Because scientific research has found that race plays a significant role in the accuracy of a witness' ability to identify a perpetrator, we chose three different actors for the experiment: a Caucasian, an African-American, and a Hispanic.
Along the edge of Atlanta's Piedmont Park we positioned seven hidden cameras at Willy's Mexicana to capture the scene from every possible angle. The customers were unaware of what was about to interrupt a leisurely weekend lunch at the outdoor cafe.
In the first scenario Justin, the white actor playing the "thief," walked up to Elizabeth, the "victim," and asked if she had a pen that he could borrow, explaining that he needed to write down a phone number. And then, suddenly his whole demeanor changed.
"Give me your purse! Sit down! Don't move!" he barked at her. Justin quickly grabbed the handbag and ran away.
Elizabeth was horrified and asked the other customers if they saw anything, knowing she'd need witnesses to help pick out the thief.
At that point ABC News correspondent John Quinones walked over to let people know that what just happened was not an actual robbery but instead an experiment about eyewitness identification. John asked the people to describe the thief.
Joe Donnelly recalled that "he had blond hair, had a hat on, about 35, 6 feet 1 inch."
But William Stark described him as "late 20's, white male … I would guess 5 feet 10 inches maybe."
Both were certain they could pick him out of a lineup.
But when we showed the videotape of the experiment and the bystanders' reactions to Professor Jennifer Dysart of New York's John Jay College of Criminal Justice, a psychologist who consults on criminal cases, she was doubtful.
"Eyewitness memory … does have its downfalls though, in that it is particularly susceptible to influences and biases," she said.
To test this idea we put together a line-up with six headshots of different Caucasian men, including Justin's, and showed it to the customers from Willy's Mexicana who represented many different races and ages.
We asked why they chose the photos they did.
William Stark noticed Justin's facial features and his hat, but added, "It's not so much that I recognize him exactly, but I tried to eliminate the ones that I strongly felt were not the guy."
It's exactly that kind of strategy that can lead to faulty identifications, according to Dysart.
"The lineup is supposed to be a memory test for the witness," she said. "And so therefore if you look at this particular lineup, this six-pack, you see it's almost like a multiple choice question … this tends to be the type of procedure that leads to inaccurate choices."
It's one reason this type of line-up is widely criticized.
Dwayne Winrow, a customer who came very close to the robber, was so sure about his choice that he claimed he "would probably go as far as to press charges."
But even though Winrow seemed very certain he chose the right person, it turned out that he was 100 percent wrong. Dysart cautions that "eyewitness certainty or confidence doesn't necessarily predict how accurate a witness is going to be."
ABC's results show just how faulty eyewitness accounts can be: 25 percent of the customers got it wrong; they identified someone other than the actor Justin as the thief. Those are not good odds when one's freedom is at stake.
The next set-up involved Aemon, a black actor. It was a perfect example of what happens during cross-racial identification — when someone of one race is asked to pick out someone of another race.
Amazingly, during the line-up every single one of the black eyewitnesses correctly identified the African-American "thief." One customer, Landon Williams recalled a key characteristic: his facial hair.
"I remember the hair that he had, the lower lip, it had connected to his beard," he said.
But most of the whites got it wrong. Seventy percent chose someone other than Aemon.
There's an interesting explanation for why the whites did so poorly and why the blacks did so well. It has to do with the particular details we notice and remember to describe when looking at people.
Dysart explained "that we tend to look maybe at the wrong cues. And so, for example, a white person would probably look at someone's hair and eye color. Unfortunately, that's not very helpful if they're being asked to distinguish amongst black people or Asians, in which hair color and eye color really doesn't vary too much. "
Scientists at Stanford University wanted to look at what actually happens in the brain during cross-racial identification and what they found is fascinating.
They took brain scans of 19 subjects while they were shown more than 100 pictures of Caucasian and African-American faces. Later they were shown some individual faces that were part of the previous group and asked if they had seen them before.
The images produced by the brain scans showed that when a white person looked at another white face, the area of the brain responsible for facial recognition lit up, indicating that it is active, the Stanford scientists said. CLICK HERE to view the image.
But when the white subject was shown a black individual's face, a much smaller area of the brain was highlighted and identification by the white person was less likely to occur, they said.
Perhaps this partly explains why eyewitness identification tends to be wrong.
In experiments no one suffers the consequences of a misidentification. But in Julius Earl Ruffin's case he paid dearly.
Still, he sees the bigger picture. He forgave his accuser but he cannot forgive the system that relied on her testimony.
"I'd been in prison 21 years, you know? For a crime I didn't commit, and … what had happened had happened, and I still felt like that it was a mistake, you know? It had to be a mistake," Ruffin said. "And, she didn't make it by herself. The judicial system had something to do with it also, you know."
CLICK HERE to visit The Innocence Project, an organization dedicated to exonerating wrongfully convicted people.