Decades of lie detection research has been unrealistic

But in a new paper, a team led by Tim Levine argues that nearly all previous research has been unrealistic. The field has been dominated by studies that place the "lie detector" in a passive role, tasked with spotting "tells" leaked by the liar. But this just isn't how deception detection works in real life, say Levine and his team. Rather, the interrogator interacts with the suspect and asks strategic questions to extract the truth. In this context, the researchers predicted that expert lie detection performance would be high.

To test this, they first invited 33 students to take part in a difficult quiz with cash prizes for correct answers. They played in pairs with what they thought was another student but was really an accomplice of the researchers. When the researcher left the room for a short spell, the student participants had a chance to cheat by looking at the answer sheet. Four of them cheated in this way.

After the quiz, all the students were interviewed about what had happened during the quiz. Unbeknown to the students, their questioner was an expert interrogator, qualified in the Reid Technique and a trainer of the police and military in interrogation techniques.

The interviews lasted about four minutes, during which time the questioner asked the students how they knew the answers they'd got right. They were also asked, if someone were to cheat, how they would have gone about it. They were also told that their partner would be interviewed next, and asked what they thought he/she would say. The questioner threatened an investigation if the student had cheated to sabotage the study, but told them that it wouldn't be so bad if they'd cheated for money.

All four of the cheaters confessed about their transgression before the end of the interview. There were no false confessions. Unsurprisingly perhaps, the questioner achieved 100 per cent accuracy after the interviews when asked to identify the cheaters from among the 33 interviewees.

The tapes of the interviews were then shown to 136 more student participants who were asked to identify the cheaters. They achieved 92.7 per cent accuracy. This is consistent with the idea that strategic questioning can provoke "diagnostic answers" that anyone can interpret because they reveal the truth about what happened.

A follow-up study was similar but this time there were five expert interrogators (one woman), all of them federal agents in the US. They each had different styles of questioning and the interviews varied from 3 minutes to 18 minutes. This time, 40 of the 89 students who played the quiz cheated, or their partner did (remember, they didn't know it but their partner was actually a research assistant).

During the ensuing interviews, confessions about cheating were obtained for 34 out of 40 of the cheating episodes. There were no false confessions. The interviewers' accuracy at correctly detecting whether cheating had occurred varied from 100 per cent (for three of them) to 94.7 per cent. The interviewers identified the specific true culprit (the student or their partner) in 95.5 per cent of interviews. When the video clips were played to 34 more students, these students achieved 93.6 per cent accuracy in judging whether cheating had occurred.

"These findings suggest that high levels of deception detection may be possible," the researchers said, "but require that the right questions are asked the right way in a situation where message content is useful and where the solicitation of honesty is a viable strategy."