That was the man who tried to rape her; she was sure of it. She had seen him before, around the apartment complex where they both lived. But her most vivid memory was of William Gregory breaking into her apartment and attacking her.
Now Gregory was sitting right across the courtroom. A lawyer was asking if she could identify the man who had assaulted her. She could. So could another victim, a 71-year-old neighbor who was raped a month later.
With both women swearing that he had attacked them, the case against Gregory was overwhelming. He was convicted and served eight years of a 70-year sentence before DNA evidence proved he couldn't have committed either crime.
"My life was snatched away from me," Gregory said. "I had a house, I had a car, I had a business. I had everything."
A decade ago, experts thought that cases like Gregory's were extremely rare. But since 1987, at least 76 men have been released from jail in the United States because DNA evidence proved their innocence. Eyewitness testimony played a major role in almost every one of their convictions.
It is difficult to determine how many people are doing time for crimes they did not commit, though estimates by professors Brian Cutler and Steven Penrod in their 1995 book "Mistaken Identification" put the error rate at 0.5%, or 5,000 of the 1 million convictions a year.
"It's the major cause of wrongful convictions," said Elizabeth Loftus, a psychologist at the University of Washington in Seattle. "I'm pretty sure of that."
An Associated Press review covering more than two decades of psychology research indicates that eyewitness testimony could be sending innocent people to jail with distressing frequency. Countless experiments show that human memory is fragile and malleable, but our justice system often treats it as an indelible record of past events.
"The principle is that the memory is like a videotape," Boston defense lawyer James Doyle said.
Malleable and fallible
We often imagine our memories faithfully storing everything we see and do. But there is no mechanism in our heads that stores sensory perceptions as a permanent, unchangeable form. Instead, our minds use a complex system to convert a small fragment of what we experience into nothing more than a pattern of connections between nerve cells.
Researchers have learned in the past few decades that this system can be fooled. Other people can convince us that we saw things that were never actually there. Things that never even happened can be remembered just as vividly as actual events. And the most confident eyewitness can simply be wrong when identifying the perpetrator of a crime.
In the 1970s, Loftus and her colleagues performed experiments showing that incorrect information provided after an event could distort a person's memory of it. In one experiment, for example, subjects watched a videotape depicting a traffic accident at a stop sign.
If Loftus then mentioned a yield sign being in the videotape, many subjects would pick up the wrong information. When asked later they would swear they had seen "yield" on the sign rather than "stop."
Further experiments showed that people's memories could be swayed even more subtly. If subjects watched a collision between two vehicles and then were asked, "How fast were the cars going when they made contact?" they gave a much lower figure than if they were asked, "How fast were the cars going when they smashed into each other?"
Eventually, psychologists found that they could actually plant false memories simply through the power of suggestion. Subjects could be made to remember barns in pictures of empty fields and guns in the hands of robbers armed with screwdrivers.
"In some of my studies where people get misinformation they can sometimes be more confident about their wrong answers than they are about the right ones," Loftus said.
Other experiments have shown that the effect translates to eyewitness identifications of perpetrators. In one, Gary Wells and Amy Bradfield of Iowa State University showed people a grainy surveillance video of a man and told them that he had later shot a security guard. Then they presented their subjects with five mug shots and asked which one was the perpetrator. It was an impossible task, because none of the five mug shots really matched the man in the video. Nevertheless, every one of the 352 subjects identified one of the mug shots as the man they had seen.
"People are pretty good at picking out the perpetrator if he's there. They're awfully bad at being able to detect that he's not there," Wells said. That's not all.
After the subjects had chosen, they were told either that they had made the correct choice, or that they had not picked out the actual gunman. Then they were asked to fill out questionnaires designed to gauge how much confidence they had in their decisions.
Subjects who had been told they made the right choice showed a great deal of confidence, saying they had gotten a good look at the perpetrator and estimating that they had been able to pick out his mug faster than they actually did. Those who had been given negative reinforcement after making their choice were much less sure, saying they hadn't gotten such a good look and over-estimating the time it had taken them to choose.
This sort of thing can and does happen in the real world, Wells said. Crime witnesses who do not immediately recognize the perpetrator in a police lineup compare the five or six faces to their memory. Then they choose the closest one. If the police investigator gives positive feedback after that choice, saying something like "Yep, that's our suspect," the witness becomes more confident. They may even generate a false memory of the selected person committing the crime.
Those false memories can wield incredible power. Jennifer Thompson, a North Carolina woman whose testimony sent an innocent man to prison for 11 years, reported dreaming of being raped by Ronald Cotton even long after she learned that he didn't do it. Both of the women who named Gregory in the Breckinridge Apartments rapes still insisted that he was their attacker, even after DNA evidence proved he couldn't have been. "They don't want to accept that they put the wrong man in prison," Gregory said.
New guidelines for lineups The legal system is woefully unprepared to deal with such situations. The U.S. Supreme Court has spelled out five standards for judging eyewitness identifications: accuracy, certainty, the amount of time since the incident, the witness' level of attention during the crime and how good a look he or she got. But experiments by Loftus, Wells and others show that false memories can meet those criteria just as well as reliable ones.
For years, prosecutors and other law enforcement officials denied that the experimental results would ever apply in the police station, much less the courtroom. But in 1996 the U.S. Department of Justice released a report called "Convicted by Juries, Exonerated by Science," analyzing the mistakes that had led to the first 28 convictions overturned by DNA evidence. In almost every overturned case, eyewitness or victim testimony had been the most compelling evidence.
After she read the report, Attorney General Janet Reno appointed a commission of psychologists, police officers and lawyers to recommend ways to decrease the number of mistakes. The panel's report, released last year, provided a set of guidelines for lineups and other procedures used to identify perpetrators. "I certainly learned a lot from what the researchers had to share with us," said Northampton, Mass., police officer and panel member Ken Patenaude.
A police lineup is a remarkably sensitive process, prone to error at just about every stage. Officers should carefully select "fillers" who resemble the suspect as closely as possible, the report recommends. They should tell witnesses that the culprit might not be in the lineup at all.
Psychologists also recommend that the lineup should be administered by an officer who does not know who the suspect is, to avoid any subtle influence on the witness. And if possible, lineups should be presented in sequence rather than simultaneously, to make it more difficult to compare them all and then select the one closest to the real culprit.
In a sequential lineup, the witness is shown one photo or individual at a time. Rather than asking one time, "Which is the perpetrator?" the administrator of the lineup asks each time, "Is this the person?" Under those circumstances, people make an absolute judgment, comparing the face in front of them with the face they remember from the crime.
Such techniques can reduce the number of mistaken identifications to about 2%, Wells said. But that will happen only if the panel's recommendations are followed.
So the Department of Justice is writing a training manual for police departments spelling out the best ways to do lineups and other identifications. "There's always room for improvement," Patenaude said. `Prosecutorial zeal' blamed Some psychologists remain skeptical of claims that mistaken eyewitnesses are the primary cause of wrongful convictions, even though they accept the research of their colleagues. Psychologist Howard Egeth of Johns Hopkins University in Baltimore said he does not dispute that eyewitnesses sometimes make mistakes; but he wonders if more of the blame for wrongful convictions should fall on legal, rather than psychological, flaws.
"What's not clear to me is whether it's really eyewitness identification that's the problem, or prosecutorial zeal," Egeth said. Calvin Johnson's story illustrates the complicated interaction between eyewitness testimony and the legal process. Johnson was arrested in 1983 at his parents' home in College Park, Ga., and charged with rape.
The victim had picked him out of a photo lineup, but then she failed to identify him in person. Later, she couldn't remember if the photographs she had seen were color or black-and-white. Hair samples from the crime scene didn't match Johnson.
In retrospect, prosecutors should have encouraged police to keep looking for the perpetrator.
On the basis of a matching blood type and the victim's testimony, Johnson was convicted and sentenced to life in prison. He spent 16 years in the Georgia penitentiary before DNA evidence proved him innocent.