Skip to main content

A Few Simple Techniques to Discern Dishonesty

A telephone survey of prospective voters finds long pauses and “speech fillers” can be signs that someone isn’t telling the truth.

By Tom Jacobs


Voters fill out their ballots in Wilmington, North Carolina. (Photo: Logan Mock-Bunting/Getty Images)

Given the polarized nature of our politics, this year’s presidential contest will likely be a “turnout election” — one where the outcome rides less on convincing uncommitted voters, and more on each side getting its partisans to the polls.

That can be a tricky task. The simplest way to determine if someone is going to vote — asking them — is notoriously ineffective. Historically, many people who pledge to participate never get around to casting a ballot.

So how do you separate the sure-things from those who need a nudge, or a ride to the polling place? New research suggests even a brief conversation can be highly revealing.

Researchers report that untrained callers can do quite a decent job of distinguishing likely voters from those who will probably flake. What’s more, a little bit of guidance could make them even better.

“Strangers can use nonverbal signals to improve predictions of follow-through on self-reported intentions,” a research team led by Todd Rogers of Harvard University writes in the Proceedings of the National Association of Sciences. This insight, it adds, could also be valuable in such non-political realms as medicine and education.

Rodgers and his colleagues describe two studies, the first of which was conducted in New Jersey one week before the 2009 general election. Workers at a non-profit organization called African Americans and Hispanics who had voted in 2008 and asked whether they would be returning to the polls for this lower-profile contest.

Only 47 percent of those who said they would vote actually did so.

The phone conversations were quite short; the callers read a 15-second script encouraging the person to vote, and then asked whether they planned to do so. When someone responded that they would indeed cast a ballot, the caller — using his or her intuition — estimated the likelihood they would follow through.

The vast majority — 4,487 citizens — indicated they planned to vote, while 1,696 said they would not, and another 644 were unsure. Comparing their responses to a publicly available voter file, the researchers found only 47 percent of those who said they would vote actually did so.

More importantly, they found the callers’ sense of their intentions was a better predictor of actual behavior than the potential voters’ assertions.

Specifically, when callers expressed strong doubts that a person pledging to vote would follow through, their predictions were “highly accurate,” the researchers report. “Citizens who self-predicted that they would vote, but were rated by callers as 0 percent likely to vote, flaked out in 74 percent of cases.”

The second study took place in Texas the weekend before that state’s 2010 gubernatorial election. As in the first study, citizens were asked their intention to vote, and callers estimated how likely they were to actually do so. Compared to those stated intentions, the callers’ hunches were, again, more predictive of the citizens’ actual behavior.

In addition, three research assistants listened to recordings of each conversation to get a sense of what cues the callers were picking up on. They found “sounding uncertain, sounding insecure, and having longer latencies before responding” were all accurate predictors of not actually voting.

The callers did miss one major signal: “Speech fillers” such as “ah” and “um” were a strong sign the person was not being totally honest. The researchers note that lying requires greater cognitive resources than telling the truth, and one way to deal with one’s less-than-optimal thought processing is to interject meaningless syllables.

In addition, the callers mistakenly felt sounding tense was an indicator of lying. The researchers’ analysis found this was not necessarily true.

Armed with this knowledge, campaigns could do even better in discerning actual voting intentions. Stronger predictions of voting intention “could increase the efficacy of the allocation of campaign resources,” Rogers and his colleagues write.

They add that this technique “is likely to be valuable in other domains as well. For example, it could be used to improve the targeting of costly interventions that increase patient compliance in medical care — a context in which millions of dollars are wasted due to patients’ lack of follow-through.”

Indeed, there are many arenas where a better sense of whether someone is lying would come in handy. This study suggests that goal is easily achievable.

It also implies that if you notice someone is speaking slowly, inserting long pauses, and interjecting lots of “errs” and “umms” into their sentences, you have every right to be suspicious.