To lie or not to lie

It turns out to be difficult to tell when other people are lying. There are lots of cues that we believe will tip us off to whether someone is telling the truth. We expect people telling the truth to be more confident, to look us in the eye when they talk, and to speak more fluently. But, these cues aren't really reliable indicators of truth telling. Someone might be uncomfortable talking about a topic and look away from you, yet still be telling you the absolute truth.

A nice set of studies by Tom Gilovich, Kenneth Savitsky, and Victoria Medvec in the Journal of Personality and Social Psychology in 1998 made this point. They had people answer questions about themselves to a group. Some people were asked to tell the truth, while other people were told to lie. Afterward, the speakers who lied were asked to rate how many people in the room would think they were lying, while those in the audience rated each speaker for whether they were telling the truth or lying. Speakers who were told to lie strongly overestimated how many people would know they were lying. They felt as though the evidence for the lie was leaking out of them, even though the audience actually had a hard time determining who was lying and who was telling the truth.

So, do liars leave any trace of their lies?

My colleague at the University of Texas (and a fellow faculty member in our program in the Human Dimensions of Organizations) Jamie Pennebaker took up this issue in his highly readable book The Secret Life of Pronouns (Bloomsbury Press).

Pennebaker is interested in the way that people use language. You might think that liars and people telling the truth talk about different things. In his analysis, though, liars and truth tellers use language in subtly different ways that are hard for people to detect, but can be pulled out of detailed analyses of the words people use.

In particular, Pennebaker and his colleagues take samples of text and count different types of words that people use. For example, they count the types of pronouns people use. Are people using first-person pronouns like I and me, or are they using third person pronouns like he and she? Are they using cognitive words like "I think that I was told about that," or do they just say "I was told about that?"

If people are engaged in a conversation, they tend not to remember these kinds of words when people are using them, because they are focused on what the speaker is trying to say. However, analyses of the words people use suggest that you can separate the lies from the truth.

In his chapter on lies, Pennebaker compared writing of stories that were either truthful or false from a number of different sources ranging from stories by the reporter Stephen Glass who was eventually fired from his job at The New Republic when it was discovered that he had faked a number of his stories to the writing of people who were asked to write either about a traumatic experience that happened to them or one that they imagined happened to someone else. They even analyzed court transcripts from witnesses who were thought to be credible versus those who were convicted of perjury (for whom there was strong evidence that they were lying).

An analysis of the words used in these stories revealed a few reliable differences. Stories that were true had more words in them and more details than those that were fake. The true stories had fewer emotion terms in them than the fakes. The true stories had fewer verbs than the fakes. Finally, the true stories had more first-person pronouns in them than the fakes.

Let's go through this a bit more carefully.

Perhaps the least surprising aspect of lies is that they are shorter and have fewer details. It is just harder for a liar to come up with specific details of circumstances that did not happen to them. Consequently, stories that are false tend to have fewer words overall and fewer descriptive words than those that are true.

You might think that true stories would have more emotion terms in them than would the lies. But when people are telling the truth, they know how they feel, and so they often don't feel the need to express it. People who are lying need to make the point that they (should be) experiencing a particular emotion, and so they are more prone to talk about it.

The difference in the number of verbs comes primarily from the use of language devices that create some distance between the speaker and the situation when someone is lying. For example, someone telling the truth might say, "I knew better than to do that," but someone lying might say, "I ought to have known better than to do that."

Finally, people telling the truth use more I pronouns than people who are lying. The analysis of court transcripts found that defendants who were found guilty at trial but were later exonerated based on DNA evidence often used first-person pronouns. The expressed their innocence by talking about themselves. Defendants who were guilty of perjury used third-person pronouns (like he and she) as they tried to shift blame to other people.

In a number of studies, Pennebaker and his colleagues took sets of texts and used these principles to classify the truthful ones from the lies. They were typically able to do so correctly about 75 percent of the time (where 50 percent is chance). That is not perfect, but it is far more accurate than even expert judges can be.

What does this mean for us? It is hard (or perhaps impossible) to train yourself to become sensitive to these differences in language use. When you are talking with someone, you have to focus on what they are trying to tell you. And your language system is not designed to keep track of all the little words that help people express what they are trying to say. So, chances are you'll still have to hope that most people are telling you the truth most of the time.

However, there are implications for the legal system. Pennebaker points out that it is exceedingly difficult to figure out who is telling the truth based just on our memory of what they are saying. Word counting programs like the ones that he uses increase our ability to separate truth-tellers from liars. Physiological measurements like those used in lie-detector test also help to separate truth-tellers from liars. These techniques are not perfect, but no forensic techniques are perfect. Perhaps it is time to revisit the role of devices that can indicate the truth of a witness to help juries reach decisions.