It may be true that seeing is believing, but we can be dead wrong about what we think we remember seeing-sometimes with tragic consequences. Just ask one of the more-than 100 people who have been freed from prison after the minutest trace of DNA evidence totally refuted the "iron-clad" eyewitness evidence that led to the prison sentence in the first place.

Eyewitness identification, in which witnesses to a crime pick potential perpetrators out of a lineup, is a widely accepted form of evidence. In police work, eyewitness ID's are critical to solving many crimes. But psychologists have amassed powerful evidence that eyewitness identifications are far less reliable than many people may think. And they believe that changes can and should be made in police lineup procedures to reduce the number of false identifications-especially those leading to long prison sentences for rape and murder.

The cost to society of inaccurate eyewitness identifications is twofold, emphasizes psychologist Rod Lindsay, Ph.D., of Queens University in Kingston, Ontario. It's a double error," Lindsay says. "Not only are you convicting the innocent-or at least putting them through the process of having to get out of the situation-but the guilty are still out there doing the crimes."

Eyewitness Identifications: Why Things Go Wrong

Since the 1970s, a series of studies has revealed factors that increase the chance that an eyewitness identification could be inaccurate. Here are some of the best-documented factors, according to a report by Iowa State University researchers Gary L. Wells and colleague Elizabeth A. Olsen in a recent issue of the Annual Review of Psychology:

RACE: People are better at recognizing the faces of members of their own racial or ethnic group than of another.

AGE: Very young children and elderly adults make more mistaken identities than young adults when a lineup does not contain an actual culprit-just suspects and innocent "filler" individuals. When the lineup does contain the culprit, age plays less of a role.

ATTRACTIVENESS: People form clearer memories of faces they consider to be attractive or unattractive, compared to those that they see as plain or average.

DISGUISES: Simple disguises such as covering the hair or wearing sunglasses can make it significantly harder to recognize a face.

ABSTRACT JUDGMENTS: If you see a face and think, "He looks like a mean person," which is an abstract judgment, you are more likely to later recall that face than if you thought, "He has a big nose," which is simply a matter of recognizing a pattern. Abstract thought and pattern recognition use different parts of the brain, which may explain the difference in the strength of memories.

WEAPONS: Research has identified something dubbed the "weapons-focus effect" that tends to reduce the chance that an eyewitness makes an accurate ID. Tracking of eye movements has shown that the presence of a weapon distracts attention from culprit's faces, which could weaken the memory for the face. Seeing a weapon may also trigger fear and stress that can weaken recall for an event.

Retroactive Rightness

Researchers who study the psychology of eyewitness identifications advocate for double-blind lineups, in which neither the witness nor the officers conducting the lineup know who the suspect is. This cuts back on comments or behaviors by police that may bias a witness—often unintentionally.

But bias can also be introduced after a witness identifies someone in a lineup. In a study published recently in the Journal of Experimental Psychology, Gary Wells and colleagues Elizabeth A. Olsen and Steve D. Charman documented the effects of “confirming feedback” from police to witnesses at lineups. If police investigators tell witnesses after a lineup that they chose the prime suspect, the witness becomes more confident in the accuracy of his or her identification. That in itself seems obvious: if told you were right, you feel more confident.

But the researchers also showed that receiving confirming feedback also tends to distort the person’s actual memory of the “witnessing event.” Witnesses may feel they had a better view of the culprit's face than they really did; they may think they were paying more attention than they actually were.

Confirming feedback is dangerous because it may make witnesses more confident in their testimony, and a more confident witness will provide testimony that -- though inaccurate -- will seem more believable to a judge or jury. That could put an innocent person in jail. “Eyewitnesses who are confident, say they had a good view, and say they paid close attention to the culprit at the time of witnessing might be accurate witnesses, or they may be inaccurate witnesses who were given confirming feedback,” Wells and his colleagues noted in their report.

The opposite is also true, the researchers found: When police lead witnesses to believe they picked the wrong person out of a lineup, their confidence plummets. This may lead them to give shaky, unconvincing testimony in court. Ironically, if they did happen to identify the culprit, their lack of confidence in their own testimony might undermine the case and allow the guilty person to go free.

Slippery memory

It’s often said that the camera doesn’t lie. But human memory does: The details of a memory can become distorted, and events that occurred at different times can become conflated with events that we imagined. As a result, a witness to a crime forms a memory that may differ markedly from the actual event. The likelihood of these so-called “retrieval errors” increases if a witness is prompted—sometimes deliberately, sometimes unintentionally—to fill in details of an event that the witness may not have noticed at the time.

For example, suppose a woman who observed a fatal traffic accident is rehearsing her testimony with a lawyer. The lawyer says, “How fast was the car going when it went through the red light?” At the time, she didn’t notice the color of the light, but the way the lawyer phrased the question plants the suggestion in mind that the car ran a red light. As a result, the woman may form an image of the traffic light in her mind’s eye—an image she didn’t really see at the actual event.

Later if she practices her testimony and retrieves the accident from her memory, the image of the red traffic light is replayed, too. Each time, it becomes more strongly embedded in memory. She may grow more and more confident that a car running a red light was part of the actual event.

In the courtroom, this woman may give inaccurate eyewitness testimony, although she didn’t intend to mislead the court. If she seems credible to a jury, her testimony might send an innocent person to jail. The bottom line: Memory does not always consistently distinguish between the parts of a story that are real and the elaborations or embellishments that were added later.

Memory and the courts

There remains a disconnect between scientific understanding of memory and the popular perception, Lindsay notes. “It’s old news to psychologists that memory is constructive, that there are inferences involved in memory, but the rest of the world doesn’t necessarily understand that,” he says. “People know there are errors, but I don’t think they really understand how easy it is to elicit those errors and how systematic the errors can be.”

Lindsay says that if you weigh in all the studies done to date on the accuracy of eyewitness identifications conducted with standard lineups, the accuracy rate is about 60 percent. That means four out of every 10 eyewitness ID’s are wrong.

Lindsay and other researchers have developed improved procedures that can significantly reduce these errors. These include sequential line-ups, where witnesses see faces one at a time rather than choosing one out of a small set. Another procedure is double-blind lineups, in which neither the witness nor the police officer handling the lineup knows who the prime suspect is. This prevents the officer from inadvertently tipping the witness off to make the "correct" choice, or from giving confirmatory feedback ("Good, you got him") that may affect the witness's memory—like the woman "recalling" the red light in the example above.

Making changes

Lindsay is certain of the need for double-blind lineups and other reforms in the justice system. "Double blind is just something that should be done automatically," he says. "It's amazing to think that scientists have said for years that they couldn't trust themselves not to bias the data. You do double-blind testing with placebos for drug studies because you can't trust yourself not to influence the results. We knew that 80 years ago, and yet nobody ever transported it to police investigations and the influencing of witnesses."

Lindsay and his colleagues are testing a new procedure that could reduce mistaken identifications even more so than a double-blind sequential lineup. In the technique, photographs of the suspect and five innocent "foils" are used to construct a sequential lineup. But instead of just showing faces, the witness sees only the head of each person in the lineup, then sees photos of just the bodies from the shoulders down with the faces electronically edited out, and finally hears a sample of each person's voice.

In preliminary testing, witnesses who can correctly identify the face, body, and voice of the person are considerably more accurate than witnesses who can only correctly identify one of these factors (typically the face). In fact, when witnesses chose the face, body, and voice, the identification is virtually always accurate. This kind of "multiple identification" technique is still in the research stage, but it does suggest that innovative, new procedures-if adopted by police-could significantly cut down on the number of inaccurate eyewitness identifications.

Research on eyewitness identifications highlights the fact that memory does not accurately record events-even extraordinary events like murder and rape. So next time you are in a heated argument with a friend or spouse over something you supposedly heard him say or saw her do, slow down for a second and think of the dozens of people who have sat in jail for years until new evidence revealed what psychologists and neuroscientists have known for years: that seeing is indeed believing, but what you "remember" isn't necessarily what actually happened.

Further Reading:

• "Eyewitness Testimony," by Gary L. Wells and Elizabeth A. Olson. (Annual Review of Psychology, 2003, Volume 54, pp. 277-295).
• "Distorted retrospective eyewitness reports as functions of feedback and delay," by Gary L. Wells, Elizabeth A. Olson, and Steve D. Charman. (Journal of Experimental Psychology: Applied, 2003, Volume 9, Number 1, pp. 42-52).