Crime witnesses frequently misidentify suspects in criminal lineups due to the human tendency to make comparative judgments, according to a scientist whose recent field study found that showing a witness suspects one-by-one reduces that risk.

Gary Wells, a professor at Iowa State University, discussed his findings Wednesday with a state task force charged with studying the use of sequential criminal lineups rather than simultaneous lineups, where the witness sees all the suspects at once. The task force was created as a result of legislation that requires the group to make recommendations to lawmakers about whether the state should mandate the practice.

Wells’ study, released last month, represents the first real field test of the lab-tested theory that one-at-a-time lineups restrict the ability of a witness to identify which suspect or photograph looks most like the person who committed the crime, rather than actually identifying that person.

Relative judgments are natural and difficult to discourage, he said. The problem is, in criminal cases, they can lead to false convictions.

“Because it comes so naturally we couldn’t really get people to stop doing it per se from things like instructions. But one of the things you can do when something comes naturally for people is instead just telling them not to do that, which doesn’t work, you can change the task,” he said. “That’s the idea behind the sequential [lineup]. Let’s try to create a task where they can’t do that.”

Criminal lineups typically feature one suspect with a number of other “fillers,” or people who are known to be innocent. In lab tests with over 13,000 eye witnesses conducted all over the globe, sequential lineups have been shown to reduce the number of times witnesses identified a filler as the suspect by 22 percent, Wells said.

However, critics of sequential lineups are reluctant to mandate that police use them because the lab tests also found that they reduced the number of times witnesses identified the culprit by 8 percent, he said.

“There then you get into value issues like if we reduce mistaken identification by 22 percent, does that compensate for an 8 percent reduction in the identification of the culprit and nobody can answer that question, I mean it’s a value question,” Wells said.

But when he and other researchers conducted a field study using actual crime witnesses at four police departments across the country, the rate at which people identified the suspect did not drop.

The lineups were conducted “double-blind,” so the detective in the room with the witness was not the case detective and could not inadvertently influence their choices.

According to the study, simultaneous lineups yielded an identification of a suspect in 25.5 percent of the lineups. Whereas the sequential lineups yielded identification in 27.3 percent of the cases. Researchers determined that the increase in the percentage was within the margin of error and not statistically significant.

There was, however, a significant drop in the number of times victims misidentified someone. The percentage of times a victim identified someone who was not a suspect as the perpetrator dropped from 18.1, under the simultaneous method, to 12.2 using the sequential method.

He noted that there were some differences between the way the lab and field tests were conducted. For instance, in the lab settings, the experiment ended the moment the subject identified a suspect. For practical reasons, that wasn’t the case when sequential lineups were used in actual police investigations, he said. Witnesses would view all the photographs no matter what, he said.

Another change was that actual crime witnesses would sometimes ask to look at the photographs again. During the lab experiments subjects would only look at them once, he said. Allowing witnesses to take “a second lap” with the lineup proved significant, he said.

When Wells removed the data where a witness looked at the photos a second time, the results looked more like the lab experiments and showed a drop in the number times a suspect was identified. He noted that even though the second viewing produced better results, it made the process a more comparative one. He recommended that police not tell witnesses they can look again unless they ask.

“The goal here is to get them to, as much as possible, treat this as an absolute recognition task not a relative recognition task,” Wells said.

Chief State’s Attorney Kevin T. Kane, a member of the task force, said it was important to not altogether discount relative judgments by witnesses.

“I think when we talk about relative judgment that can be important and it might be legitimate evidence for a jury to hear in a trial,” he said. “To have a witness describe the robber was taller than Senator Coleman or shorter than Judge Borden, that’s certainly relevant and proper for the jury to hear.”

That kind of relative observations also help police narrow their investigations or discount suspects, he said. It’s important that when the task force makes its recommendations, it drafts a policy that produces the best results possible but does not create a barrier that would prohibit juries from considering circumstantial facts, Kane said.

Wells agreed. After the meeting he said he generally prefers to try to get police departments to make reforms on their own, rather than a statewide mandate.

“The reason for that is because you get better buy-in. In other words, by getting them to make the switches, make the change to these kinds of reforms on their own, they’re more likely to accept them, follow them properly, and so on,” he said. “If it’s coming from the state it’s like they’re telling us how to do our business.”