Editor's Note: Steve Chapman is on vacation. The following column was originally published in February 2005.

In September 1985, Dennis Brown heard the words that sent him to prison for rape. The victim took the stand and had no doubt who had attacked her. "I had his face this close for at least 20 minutes," she said, holding her hand inches from her face, "and he's the man." Brown was convicted of aggravated rape and sentenced to life without parole.

But in October 2004, the 36-year-old Brown walked out of the Louisiana State Prison in Angola, having been exonerated by DNA evidence. Prosecutors dropped the charges. After being locked up for 19 years, more than half his life, an innocent man was free.

Like many people in his ill-starred position, Brown was snared by a mistaken identification. The victim picked him out of a police lineup, and her testimony provided the bulk of the evidence against him.

His case illustrates the dangers of relying on what used to be seen as the best kind of evidence — a person who was present at the scene of the crime who can attest, "I saw him do it." Time and again, thanks to DNA evidence, we've seen that a victim can be absolutely sure in identifying her attacker — and be absolutely wrong. Amy Klobuchar, prosecutor for Hennepin County, Minn., which includes Minneapolis, says faulty identifications are "the single most common error" generating bad convictions.

What is most disturbing is that the mistakes we know about represent only a tiny share of the total. Most of the exonerations involve rapes — where DNA can definitively establish the perpetrator. But police rarely find bodily fluids in robberies, muggings, burglaries and other far more common crimes. So if someone tabs an innocent person, the innocent person probably won't ever be cleared.

Many of the mistakes occur during police lineups, where witnesses try to pick out the perpetrator from a group of people or pictures. Experiments have shown that when confronted with several possible suspects at once (a "simultaneous" lineup), the witness is prone to choose whoever most resembles the actual criminal — even if the actual criminal is absent.

Presenting the choices one at a time (a "sequential" lineup) is more likely to yield a correct identification.

In 2001, the New Jersey attorney general required all police departments to change the way they handle lineups to prevent errors. The change that got the most attention was the adoption of sequential lineups. But Gary Wells, a psychology professor at Iowa State University who has been the chief pioneer in studying eyewitness identification, says that was not the most important reform. Even more critical was the use of "double-blind" testing — where the police officer conducting the lineup doesn't know which of the people is thought to be the guilty party.

Why does it matter? Because an officer can consciously or unconsciously steer a witness. In 2005, Wells, speaking at a Minneapolis conference on wrongful convictions, said that when a witness chooses the "wrong" suspect, the lineup administrator may say, "Are you sure?" or "Take another look at No. 3." But when the witness chooses the right suspect, the response may be, "Tell me about him."

Sometimes the police buttress the witness's memory by saying something like, "You got the right one." In an armed robbery in Iowa, Wells recalls, the victim was asked in the trial how detectives responded when she chose the defendant's photo. "They clapped," she said.

Normally, the cops aren't trying to manipulate the witness; they're just being human. But occasionally, the tilt is intentional. In 2002, the city of Chicago was ordered to pay $15 million to James Newsome because two detectives had rigged the lineup in which he was identified, leading to his mistaken conviction for murder. One witness said he was repeatedly told to look at Newsome. Newsome said he saw one detective point him out to another witness.

Problems like these can be avoided — by turning the lineup over to someone who knows nothing about the case, or by presenting the photos so the administrator can't see them, perhaps on a computer. Such changes have worked well in New Jersey, where 91 percent of respondents to a statewide survey of law enforcement agencies said the new methods created no major problems.

Eyewitness testimony can be extremely useful in catching criminals, but it needs safeguards to make sure it doesn't nab the innocent. After all the wrongful convictions in recent years, no one should have trouble seeing that.

To find out more about Steve Chapman, and read features by other Creators Syndicate writers and cartoonists, visit the Creators Syndicate web page at www.creators.com.

COPYRIGHT 2008 CREATORS SYNDICATE, INC.