SAVANNAH, Ga. (AP) -- When Georgia executed Troy Davis last week, it brushed aside international protests that too many witnesses had recanted trial testimony that he was the gunman who killed a police officer in 1989.
The issue raised in Davis's case, however, is getting harder to ignore. With scientific studies showing the human memory can be surprisingly faulty, the once-damning weight of eyewitness testimony has come under question in courts and state legislatures.
Last month, New Jersey's top court made it easier for criminal defendants to challenge the credibility of eyewitnesses, while the U.S. Supreme Court is set in November to hear its first case dealing with eyewitness evidence in 34 years. Such issues also played a role in the abolition of Illinois' death penalty earlier this year and a 2009 law narrowing when capital punishment can be sought in Maryland.
Davis' execution outraged hundreds of thousands of people who said they feared an innocent man was being put to death, based on his defense attorneys' assertion that witnesses who had identified Davis in court as a killer two decades ago had tried years later to take it all back. Dorothy Ferrell was one of those witnesses.
"Well, I'm real sure, positive sure, that that is him, and you know, it's not a mistaken identity," Ferrell told a Savannah jury in 1991. "I did see him and you know, on the fact of what happened and how it happened, you know, I'm pretty sure it's him."
Nine years later, Ferrell signed an affidavit saying she didn't actually see the 1989 shooting of off-duty police officer Mark MacPhail, but pointed at Davis to tell police what they wanted to hear.
Legal experts say Davis' case serves as an example in the debate over eyewitness reliability, particularly in death penalty cases, when scientific studies show the human memory can be surprisingly faulty.
"There's going to be some broader discussions about whether the death penalty is viable at all, but before that happens there's going to be efforts to reform and see what can be done in states that believe in it and regularly use it," said Richard Dieter, executive director of the Death Penalty Information Center, which opposes capital punishment.
Even before Davis' execution last Wednesday, several states had reduced reliance on eyewitnesses.
The Supreme Court of New Jersey, which abolished the death penalty in 2007, last month issued a ruling making it easier for criminal defendants in its state courts to get pre-trial hearings challenging eyewitness evidence. It also requires judges to give juries more detailed instructions about potential flaws in eyewitness identifications.
In 2009, Maryland lawmakers prohibited prosecutors from seeking death unless they have DNA evidence, a videotape of the crime or a videotaped confession from the suspect.
"Eyewitness testimony is so horribly inaccurate - even under the very best of circumstances," said Rob Warden, director of the Chicago-based Center on Wrongful Convictions at Northwestern University. "We should never depend on eyewitness testimony in death penalty cases."
The center says that nationally, out of 138 defendants sentenced to death for murder and then later exonerated since the mid-1970s, 32 had been convicted in whole or in part based on erroneous eyewitness testimony.
As Illinois moved to abolish its death penalty in March, state officials cited Anthony Porter, who was condemned for a 1982 double murder based on eyewitness testimony that authorities later determined was false. Porter got a reprieve just two days before his execution in 1998, and was released from prison the following year.
Meanwhile, a researcher who's been studying eyewitness issues for 30 years released a study this month that shows police can reduce chances that witnesses will mistakenly point to innocent people in lineups by adopting a few simple procedures.
Gary Wells, an Iowa State University psychology professor, studied 497 instances of witnesses to real crimes looking at lineups on police computers in four states. He found that when witnesses looked at a group of photos all at once, they were more likely to compare faces and pick the one that most resembled the suspect - whether it was correct or not.
The rate of wrong identifications declined, from 18 percent to 12 percent, when witnesses viewed the photos one at a time.
Willis says the one-by-one approach would also make in-person lineups more reliable. It also helps if the officer working with the witness doesn't know the suspect is, to avoid influencing the outcome. He says police should also tell witnesses it's OK if they can't pick a suspect out of a lineup.
"These kinds of events that people witness, whether a victim or a bystander, often happen very quickly, they're unexpected," Wells said. "It's not like the only thing to look at is the perpetrator's face. There are other things going on; people fear for their safety."
Prosecutors balk at the idea that people are sentenced to death based purely on eyewitness testimony. In Davis' case, for example, prosecutors used shell casings recovered from the scenes of two different shootings hours apart to link the crimes to Davis, who admitted being at both places when shots were fired. A firearms examiner testified it was likely, but not certain, the casings came from the same gun. Some witnesses who identified Davis as the killer have never backed off their stories.
Scott Burns, director of the National District Attorneys Association, said advances in crime scene investigating technology have made it tougher for prosecutors to lean too heavily on eyewitnesses. He said he prosecuted a car-theft case in Utah years ago in which jurors asked if he had any DNA evidence.
"It has raised the expectations of juries," Burns said. "People want all of their senses stimulated. They want to see pictures, they want to watch video."
But eyewitness testimony remains a cornerstone of prosecutions, with many cases yielding very little physical evidence, said Brandon Garrett, a law professor at the University of Virginia.
In his recent book "Convicting The Innocent: Where Criminal Prosecutions Go Wrong," Garrett looked at 190 criminal cases where eyewitnesses helped win convictions for a range of crimes that were later overturned by DNA evidence. He found that witnesses often seemed more confident in identifying suspects from the witness stand years later than they were when interviewed by police right after a crime.
"You had these eyewitnesses almost without exception come into the courtroom and say they were absolutely certain they saw the defendant do the crime," Garrett said. "But more than half remembered being unsure at the time they saw their first lineup."
Among the exonerations Garrett studied was that of John Jerome White, who spent nearly 30 years in a Georgia prison for rape until he was exonerated by DNA testing in 2007. The case came with a startling twist: after White's release, police arrested another man for the same crime - a man who had stood in the same police lineup with White in 1979.
Barry Scheck, co-founder of the Innocence Project, said the legal system is poised to change how it handles eyewitness evidence. The U.S. Supreme Court in November is slated to hear a New Hampshire case that asks whether courts should throw out eyewitness testimony that's been influenced by friends and neighbors in the same way they would reject witnesses tainted by police.
"The Troy Davis execution came at a time where we're at tipping point or there's critical mass concerning eyewitness reform," Scheck said, noting the Supreme Court hasn't ruled on the issue since 1977. "Thirty-four years later, the science dictates it has to change."
Associated Press writers Michael Tarm in Chicago and Eric Tucker in Washington contributed to this story.