Three people, including the teacher raped in her Towson apartment, picked Bernard Webster from a police lineup in 1982, essentially ensuring the Baltimore man's conviction.
Two years later, two boys from Rosedale said they recognized Kirk Bloodsworth as the man who walked away from Becky's Pond with 9-year-old Dawn Hamilton, who was later found raped and murdered.
This fall, Montgomery County police received tips about a white truck or van speeding from the sniper shootings and searched scores of matching vehicles. The suspects charged in the killings were driving a blue Chevrolet sedan.
For decades, witness identification has played a key role in criminal investigations, directing police to suspects and swaying juries. But a wave of DNA-based exonerations, including those of Bloodsworth in 1993 and of Webster this month, have many questioning this once-basic piece of evidence.
Some attorneys, academics and even the Justice Department have recommended basic changes to the way officers collect witness information. Others have encouraged court systems to address the growing body of research showing that witnesses, despite best intentions, are often wrong.
"When people witness a crime, whether they're the victim or a bystander, they're taking in a lot less information, and a lot less detailed information, than they may think," said Gary L. Wells, a professor of psychology at Iowa State University who has conducted research on this topic since the early 1980s.
But change has come slowly. While some districts -- most notably the state of New Jersey -- have adopted these recommendations, most have not.
In Maryland, few if any police departments have changed the way they conduct live lineups and photo arrays to correspond to a policy backed by the Justice Department. The departments say they already take precautions to ensure witness credibility.
In addition, Maryland judges are not required to let experts talk to jurors about the fallibility of witness testimony. And the state's stock jury instructions about such testimony are outdated and actually contradict current science, defense attorneys say.
`God knows how many'
"You know, it's really scary," said defense attorney Carroll McCabe. "We see all the people now who have been proven innocent through DNA. God knows how many people have been falsely convicted of armed robbery who will spend many years in jail because there's no DNA to exonerate them."
The Innocence Project, a New York-based group that works to identify and free those wrongly convicted, estimates that incorrect eyewitness testimony helped convict 70 percent of the people subsequently exonerated by DNA evidence.
Extensive research has been conducted on why people so often misidentify suspects.
"People are very suggestible," said Yale Law School Professor Steven Duke. "We just don't remember what we see."
A classic law school exercise proves the point. After a staged surprise attack, students are told to write descriptions of the "suspect" they just saw.
"You get an incredible range, from age, height, weight to race," said University of Maryland Law School Professor Douglas L. Colbert. "I remember that there was a disproportionate number of witnesses who identified the perpetrator as a person of color."
Stress, psychologists say, makes memory worse. The idea that a victim will never forget the face of an attacker is simply a myth, they say.
"Jurors believe eyewitness accuracy is tied strongly to confidence," said Amy L. Bradfield, an assistant professor of psychology at Bates College in Maine who has written extensively on misidentification. "It turns out there is a small relationship between confidence and accuracy. You can have this really confident, compelling witness who can also be wrong."
The teacher who identified Webster as her rapist remains convinced he is guilty, prosecutors say, even after DNA tests exonerated him.
Small changes in policy
Researchers say police can limit misidentifications with small changes in policy. Witnesses should look at possible suspects or their pictures one at a time, the researchers say, rather than in a group. That way, witnesses are comparing images to their own memory, not picking the person who looks most like the attacker compared with other photographs or people.
Researchers also say that someone unfamiliar with the suspect should run the lineups rather than a detective working on the case. There are too many ways for an officer involved in a case to unconsciously push a witness to one choice or another, according to Wells and other psychologists.
The Justice Department supported this research in its 1999 "Eyewitness Evidence" guide.