Debates: Linguistic trick boosts poll numbers

October 18, 2016, University of Michigan

A study of U.S. presidential debates between 1976 and 2012 found that matching certain aspects of an opponent's language can lead to a bump in the polls.

"Linguistic style matching," says a University of Michigan professor who led the study, has nothing to do with tone, cadence or the number of times one candidate interrupts the other. Nor is it about content—the nouns and regular verbs that make up "what" a speaker says.

It's much more subtle. Linguistic style matching zeroes in on so-called function words that reflect how a speaker is making a point. It refers to conjunctions like "also," "but" and "unless;" quantifiers like "all," "remaining" and "somewhat;" and other supporting parts of speech.

"These function words are inherently social, and they require social knowledge to understand and use," said study author Daniel Romero, an assistant professor in the U-M School of Information, as well as in computer science and engineering. "We think that matching an opponent's shows greater perspective taking and also makes one's argument's easier to understand for third-party viewers."

Also collaborating on the study are Roderick Swaab, associate professor of organizational behavior at INSEAD in Singapore; Brian Uzzi, professor of leadership and organizational change at Northwestern University's Kellogg School of Management; and Adam Galinsky, professor of business at Columbia University.

The researchers obtained transcripts of 26 debates from 36 years of presidential campaign seasons, going back to Jimmy Carter and Gerald Ford. They looked at the degree to which each candidate linguistically matched the other in eight different style markers through rounds of questions (one of the reasons the team chose to analyze debates is because candidates take turns being the first to answer the moderator).

The eight style markers they evaluated amount to 444 words: quantifiers, conjunctions, adverbs, auxiliary verbs, prepositions, articles, personal pronouns and impersonal pronouns. Examples include "about," "especially," "perhaps," "must," "might," "these" and "our." Each candidate in each debate was given a score for how closely they matched their opponent's linguistic style according to these parameters, when their opponent had been the first to speak.

Then the team examined Gallup polls and meshed the data. They found that linguistic style matchers gained a median of one point. And those that didn't match lost a median of one percentage point in the polls.

"We already knew a lot about how linguistic matching can affect a relationship between two people. It can lead to better outcomes for negotiators, for example. In this case, we were interested in something different," Romero said. "And that's when a third person is watching the exchange and judging who is doing a better job. We didn't know a lot about that before."

The outcome didn't surprise Romero.

"We think linguistic style matching is linked to processing fluency and if that's the case, it helps the third person have an easier time understanding the candidate's response," he said.

No candidate over the years stood out as being a supreme style matcher. Some did well in particular debates, only to get low marks in others against the same opponent. And poll data didn't always correlate with election outcomes.

For example, Gerald Ford received a positive linguistic style matching score of .02 in the '76 election's first debate. His poll numbers spiked 6.5 percentage points. In contrast, Carter's linguistic matching score was -.53. He was not adept in that case at mirroring how Ford made his points. Carter's poll numbers dropped by 2 points. But it was Carter who prevailed in November.

In contrast, the first debate of 2000 turns out to have predicted the White House inhabitant. George W. Bush matched Al Gore well, for a score of 1.43. He rose two percentage points in the polls. Gore, on the other hand, got a negative matching score of -0.41. He fell in the polls by three points.

A paper on the work, "Mimicry is Presidential: Linguistic Style Matching in Presidential Debates and Improved Polling Numbers," has been published in Personality and Social Psychology Bulletin. The research was funded by the Northwestern Institute on Complex Systems, the Army Research Laboratory and the Defense Advanced Research Projects Agency.

Explore further: Want the most telling presidential polling data? Professor says turn to Twitter

Related Stories

The language of persuasion

April 27, 2016

It's not what you say, but how you say it. By analyzing online arguments, Cornell researchers have identified how language and interaction with the other party contribute to winning an argument.

From dating profiles to Brexit—how to spot an online lie

June 29, 2016

There are three things you can be sure of in life: death, taxes – and lying. The latter certainly appears to have been borne out by the UK's recent Brexit referendum, with a number of the Leave campaign's pledges looking ...

Recommended for you

Sensual fresco discovered in ancient Pompeii bedroom

November 19, 2018

Archaeologists have found a fresco in an ancient Pompeii bedroom that depicts a sensual scene of the Roman god Jupiter, disguised as a swan, and a legendary queen of Sparta from Greek mythology.

Excavators find tombs buried in Bolivia 500 years ago

November 17, 2018

Archaeologists say they found tombs at a Bolivian quarry containing remains from more than 500 years ago that give an insight into the interaction of various peoples with the expanding Inca empire.

Preventing chemical weapons as sciences converge

November 15, 2018

Alarming examples of the dangers from chemical weapons have been seen recently in the use of industrial chemicals and the nerve agent sarin against civilians in Syria, and in the targeted assassination operations using VX ...

0 comments

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.