Are You Being Lied To?

Research proves our assumptions about liars wrong, and finds something better.

Posted Nov 24, 2015

The author of this post is Steven Jackson.

He does not answer questions, or gives evasive answers; he speaks nonsense, rubs the great toe along the ground; and shivers; his face is discolored; he rubs the roots of his hair with his fingers. — Description of a liar, 900 B.C.E.

Source: VILevi/Shutterstock

Through the ages, we’ve tried hard to ferret out the liars among us to protect ourselves from deception. Most of us think we’re pretty good at it; anyone who fits our mental image of a sweaty-palmed foot-shuffler, we believe, is probably not to be trusted.

There’s good evidence that at least one liar stereotype is common across cultures: In 2006, the Global Deception Research Team catalogued beliefs about liars around the world. They surveyed thousands of people in 75 countries and found that 72% of participants agreed that liars avoid eye contact.

The researchers theorized that all such liar stereotypes are based on social learning. As children navigate enculturation, they learn that lying is bad and something to be ashamed of. Since gaze aversion is an expression of shame across cultures, children typically learn to recognize this signaling of shame at a young age. Perhaps, then, this pan-cultural shifty-eyed stereotype comes from commonalities in enculturation.

But there’s just one problem with the gaze-aversion stereotype—and all the other stereotypes about lying, too: They don’t actually help us spot liars.

A meta-analysis of 206 studies found that, when people try to determine if someone is lying or not, they are correct about 54% of the time. In other words, human judgment is barely more reliable than flipping a coin.

Several years ago, researcher Charles Morgan and colleagues conducted a series of lie-detection experiments around the world, from Lebanon to Russia to Singapore. They spoke to more than 1,000 people using the principles of cognitive interviewing, a technique frequently used with eyewitnesses. The idea behind the approach is that even difficult-to-access memories can be retrieved if the interviewee is prompted in particular ways.

Participants in Morgan’s studies were asked to give a narrative account of something that had happened to them—a family outing, a concert, a first date. But there’s a catch: One group was instructed to tell a true story, while another was told to make something up. In both groups, after hearing a story once, the interviewer asked the participant to retell it several more times, focusing on a different sensory detail each time.

“What’s been found over the years in many studies of cognitive interviewing is that using those...sensory prompts—what you would’ve seen, heard, smelled, thought, touched, or tasted—they trigger more memory recall,” Morgan said in an episode of the criminal-justice podcast Criminal.

In the final phase of the studies, the interviewer prompted the subject to tell the story backward, step-by-step. Although it was a lengthy process, most people were able to relate their memories in reverse order—at least those who had been sharing true accounts. The participants who had made up their stories struggled because the cognitive resources needed to tell a fabricated story backwards are substantial.

“The overall result is that they have very little to say,” Morgan said. “It’s sort of like comparing a digital photograph of your house with the tree in the front yard and a child’s picture of it [where] there’s not a lot of detail.”

Even with this obvious difference in detail, though, interviewers (who did not know in advance which participants had been instructed to tell false stories) struggled to figure out when they were being lied to. Their accuracy? Little better than chance. So Morgan decided to remove the human element. He entered interview transcripts into a computer, calculating the total word count and the number of unique words in the interview—basically a tally of how repetitive the participant’s language was.  

“All the computer is doing is counting those two variables,” Morgan said. “And when we sort people based on those two variables—response length and unique word count—the computer is right typically 80% to 85% of the time.” When the computer was programmed to categorize short, repetitive transcripts as untrue, and long transcripts with many unique words as true, it was correct in identifying liars 8 times in 10, which is much better than the performance of human interviewers.

No matter the technique used, most of us just can’t spot liars on a consistent basis. One potential solution to the problem is to remove “people” from the equation. Just let a computer do it.


  • Bond, C. F., & DePaulo, B. M. (2006). Accuracy of deception judgments. Personality and social psychology Review, 10(3), 214-234.
  • Global Deception Research Team. (2006). A world of lies. Journal of Cross-Cultural Psychology, 37(1), 60-74.
  • Kassin, S., & Wrightsman, L. (1988). The American jury on trial: Psychological perspectives. New York: Hemisphere Pub.