It’s a trademark scene in any crime drama: a nervous eyewitness is brought into the station by police officers to finger a perpetrator from a lineup of known criminals and familiar trouble-makers. Flashbacks to the crime scene hint that the eyewitness isn’t completely sure of what he or she saw — it was too dark or the witness was too scared or there were too many people present — adding to the drama.
That tension has its roots in reality. While we like to think that our eyes won’t mislead us, mistaken eyewitness identification accounts for more than 70% of wrongful convictions that are ultimately overturned by what law enforcement officers are increasingly considering more objective DNA evidence. Police lineup techniques have long been part of the problem, but a new study suggests there are ways to improve them and increase — by up to 66% — the accuracy of identifying suspects.
The key is fingering the factors that contribute to accurate recall. Prior studies have found that quick recall suggests a strong (and therefore more accurate) memory, as does your own sense of confidence in its reliability. Memories are also heavily influenced by context — so people facing the task of identifying a suspect in a lineup may unwittingly come to false conclusions based on their own feelings or priorities during the identification procedure.
For instance, if you are more concerned about wrongly convicting the innocent, you may increase your skeptical view of your memory too much, leading to a fuzzy recall. Concern about letting a guilty suspect go free can wreak equal havoc on memory, by compelling you to tamp down your doubts about what you saw or heard too readily.
(MORE: Does Marijuana Use by Teens Really Cause a Drop in IQ?)
To address these issues, researchers led by Neil Brewer of Flinders University in Australia experimented with new lineup techniques. In the study, which was published in Psychological Science, more than 900 participants viewed short films of everyday scenes, and were told beforehand that they would be questioned about them later. After watching the films, the participants were told that crimes had been committed in the area they had just viewed, and the researchers provided the volunteers with descriptions of a specific suspect, such as a “male 20 to 30 years old with darkish hair and a pale complexion.” All of the suspects appeared in the films.
In one experiment, participants saw two films along with a final “distractor” film; five minutes later, they were asked to identify the suspect from a series of photographs. Half of the participants were allowed to take as much time as they wanted to make their identification and simply report whether or not the suspect appeared among the photos they were shown; in some cases the suspect was among the photographs, in other cases he was not. The other half of study volunteers had only a few seconds to make their decision, but rather than give a yes or no answer, they also had to rate on a 100-point scale how confident they were about each photo, with 0 meaning they were absolutely certain the person was not the suspect and 100 meaning 100% confidence they had their culprit.
To test various ways of modifying lineup procedures, the scientists also varied their questioning; in another version, the researchers used four films rather than two, and in the final experiment, the participants made their identifications a week after four short films were viewed, rather than minutes later.
Giving the “eyewitnesses” a shorter time period to decide and asking them to rate the confidence of their identification made a huge difference. In the first experiment, accuracy improved by 27%; in the second, it was increased 24%. But in the third condition, which was more similar to what happens in real crimes, in which the suspect isn’t identified within minutes but often days or weeks later, accuracy was improved by a whopping 66%. Researchers also found that when people had large differences between their maximum confidence rating and their next-highest one (say, 70% compared to 100%), they were very accurate. But differences of 20% or less signaled low accuracy, or, by extrapolation, more confusion on the part of the witness.
If these results are replicated and studied in more realistic contexts, say the researchers, “there may exist a superior procedure to the age-old and problematic identification test.” And perhaps eyewitness testimony would become a more reliable tool for law enforcement.
MORE: Brain Food: Berries Can Slow Cognitive Decline
Maia Szalavitz is a health writer at TIME.com. Find her on Twitter at @maiasz. You can also continue the discussion on TIME Healthland’s Facebook page and on Twitter at @TIMEHealthland.