Smart Language is not Smart Politics: A Computational Analysis of the 2012 Presidential and Vice Presidential Debates
November 1, 2012
By Nia Dowell, John Myers, and Arthur Graesser, Department of Psychology and the Institute for Intelligent Systems, University of Memphis
It’s early November in a presidential election year, and every time candidates open their mouths, they are trying to persuade potential voters. Clearly, language is a powerful tool used by politicians to deliver persuasive messages and gain support.
With this in mind, we examined the extent to which the presidential and vice presidential candidates utilized the theoretically grounded strategies of persuasion put forth by Petty & Cacioppo (1984). Their research discriminates between two distinct paths to persuasion, namely the central route and the peripheral route.
The central route is characterized by coherent, logical, expository language, resulting in a formal complicated message. The upside of this central route is that once you are persuaded, you usually stay persuaded. That is, the central route results in a rather stable opinion and attitude change. In the political context, however, the downside of the central route is that the majority of people in the general public audience may not be motivated or educated enough to digest a message that might appear a little dry. From a political candidate’s point of view, this obviously has many detrimental consequences, maybe millions if you consider the number of “tuned out” Americans out there.
The peripheral route offers more advantages to political candidates striving to gain the support and retain the attention of the masses. The peripheral route is characterized by entertaining, narrative discourse that is more informal and easier to comprehend. As you might have guessed, successful persuasion via the peripheral route is quite temporary and unstable. However, considering that the election is less than one week away, this may be all the candidates really need.
So, now you no doubt want to know, who took what route in the recent 2012 election debates: The Democrats or Republicans?
Using a computational linguistics facility, Coh-Metrix (Graesser, McNamara, & Kulikowich, 2011), we assessed the extent to which the presidential and vice presidential candidates pursued the central and peripheral routes to persuasion.
Coh-Metrix is an automated linguistic analysis tool that analyzes text by measures of genre, cohesion, syntax, words, and other characteristics of language and discourse. Coh-Metrix is sensitive to a wide range of linguistic/discourse features that reflect multiple levels of complexity. Five orthogonal Coh-Metrix dimensions of text characteristics were identified as accounting for most of the variability in text complexity among 37,521 texts in a principal components analysis:
- Narrativity. The extent to which a text features story-like elements, such as characters and events, which are closely affiliated with everyday oral conversation.
- Deep cohesion. The extent to which the ideas in a text are connected with causal, intentional, or temporal connectives at a deeper level of semantic understanding.
- Referential cohesion. The extent to which discourse contains explicit words and ideas that overlap across sentences and the entire text.
- Syntactic ease. Sentences with fewer words and simple, less embedded syntactic structures are easier to process and understand.
- Word concreteness. The extent to which a text features concrete words that tend to evoke mental images and are more meaningful to the reader than abstract words.
Using these measures we were able to compute a composite measure of formality. Formal text is informational (rather than narrative) with high cohesion, complex syntax, and abstract words. Informal text is more narrative, with lower cohesion, simple syntax, and concrete words.
Our analysis of the presidential and vice presidential debates revealed some interesting, and potentially alarming results for the future of the democratic candidates (Obama and Biden) with regard to their persuasive impact.
Debate I: For the first presidential debate, we observe a clear difference in language formality. This difference in formality may tip the ballot in favor of Romney, with regard to persuasion. Romney came out swinging, using the language of the “people” and capitalizing on the peripheral route to persuasion. Obama had more complex syntax and deeper cohesion, but those features of language are not aligned with the peripheral route to persuasion.
Debate II: For the second presidential debate, we see a slight increase in formality for Romney compared with the first debate, but we still observe this same pattern in language formality between the candidates.
Debate III: So who’s advising the president?! In the third debate, we watched him take more than a serving of the formal discourse pie and continue to use complex and stilted language. Did he lose the attention of the masses this time around?
Vice Presidential Debate
Similarly, we witness a formal and complicated message for the democratic vice president, Biden, as he follows suite with his running partner by using the central route to persuasion. Ryan, on the other hand, is keeping it more informal and simpler, steadily keeping the attention of the masses via the peripheral route to persuasion.
The Democrats may have more formal language and appeal to the smart voters who are attracted to the central route to persuasion. However, the peripheral route to persuasion is more effective for most voters. On November 6 we shall see which route reigns supreme.
Graesser, A.C., McNamara, D.S., & Kulikowich, J. (2011). Coh-Metrix: Providing multilevel analyses of text characteristics. Educational Researcher, 40, 223-234.
Petty, R., & Cacioppo, J. (1984). The effects of involvement on responses to argument quantity over quality: Central and peripheral routes to persuasion. Journal of Personality and Social Psychology, 46, 69-81.
October 22, 2012
by James W. Pennebaker and Cindy K. Chung
Pundits are reporting that the third debate on foreign policy was a toss-up, perhaps with a very slight edge in Obama’s favor. The stubbornly objective computerized text analysis program, LIWC, refused to name a winner. But in counting different types of words by Obama and Romney, LIWC revealed some word patterns that tell us a little more about the two candidates.
The candidates as communicators: businessman versus professor. Romney and Obama break down ideas and express them in very different ways. They simply think differently.
Across the debates, Romney uses shorter words and shorter sentences. He tends to be more concrete, breaking problems into more precise categories. Compared to Obama, he uses a high rate of articles (a, an, the) which reflects his reference to concrete nouns. These nouns are well organized; by his high use of prepositions and specific references to spatial words (above, around, below, outside), Romney provides a roadmap of his thoughts.
Obama tends to be a more theoretical communicator. His sentences are much longer, his words bigger, and he is much less personal. Unlike Romney, Obama often sets up a context for his thoughts. Look at the way he often starts a response to a question: “Well, my first job as commander in chief, Bob, is to keep the American people safe. “; “You’ve got to be clear, both to our allies and our enemies, about where you stand and what you mean.“; “When I came into office, we were still bogged down in Iraq and Afghanistan had been drifting for a decade.“
Compare those first sentences to some of Romney’s: ”Well, my strategy is pretty straightforward, which is to go after the bad guys…“; “I don’t want to have our military involved in Syria. I don’t think there is a necessity to put our military in Syria at this stage.” ”Well I believe we should use any and all means necessary to take out people who pose a threat to us and our friends around the world.“
Obama thinks about a problem within a broader context; Romney tends to answer questions directly and then brings up contextual matters after the fact. These two types of thinking are likely to lead to very different strategies in decision making. A contextual thinker will likely spend more time looking at a problem from multiple perspectives. A more categorical thinker will likely have a more set world view and simply make decisions from pre-existing values or beliefs.
Ultimately, the two candidates are revealing their backgrounds. Romney is a businessman with a clear sense of economic realities, and a bottom line. Obama is, at heart, an academic — a world where there are multiple ways to get at truth, and so competing views must be acknowledged.
Candidates as people: social relationships versus achievement. Despite the nastiness of the 2012 presidential election, both candidates have come across as honest, decent people. They both appear to have warm and stable families. One thing that a text analysis method can help uncover are some of the basic values that are central to candidates. Across the debates, the two men have made references to money, religion, death, and family at similar rates. Two language dimensions that have emerged are references to other people and references to achievement and work.
One dimension of the LIWC program counts references to anything social — people, conversations, friendship, or anything that suggests an interest in or awareness of other people. Within each of the three debates, Obama has consistently used social words about 10% more frequently than Romney. In his mind, the current economic difficulties are a problem because they hurt people.
Romney, on the other hand, has consistently used words that suggest achievement (e.g., earn, success, win). Not surprisingly, Romney has also made greater references to the related dimension of work. Work words include a variety of more economically-relevant words such as job, paycheck, and company. Whereas Obama views economic difficulties as people-problems, Romney frames the recession as something that undermines achievement, jobs, and progress.
Obviously, a recession is bad for companies and for people. But the framing of the problem reveals differences in core values about people and human nature. Ultimately, it likely drives what each believes to be the most effective solution.
The debate language of Obama and Romney. It should be noted that debates are a very peculiar interaction in which to study language use and personality. Although many labs and companies have examined the men’s language use, body language, and voice characteristics, we are most impressed by how similar the two really are. They are both smart, eloquent, formal, and trustworthy.
The analysis of pronouns and other function words is helpful in revealing how people are speaking or thinking. LIWC can also give insight into personalities and even central values. The degree to which Americans agree with what the candidates are saying remains to be determined in the upcoming elections.
by James W. Pennebaker and Cindy K. Chung
The October 16, 2012 debate was interesting for several reasons. First, there was a shift in the social dynamics between Barack Obama and Mitt Romney. As you can see in an earlier post, the first debate revealed Romney to be more authentic and slightly less optimistic and warm than Obama. Both were equally evasive in answering questions.
Most striking has been the immediate reaction from pundits — who generally claim that Obama was far stronger, more assertive, and more engaged than he was in the first debate. Others note that Romney was more defensive. Our text analysis procedure showed that both candidates were surprisingly similar in their use of words across the two debates. We are pleased to officially announce that peoples’ personalities tend to remain constant over time and context.
But first, the award winners.
The authenticity award: Mitt Romney.
Authenticity reflects markers of self-reflection (I-words), cognitive complexity, and relative low rates of negative emotion. Compared to the first debate, both Romney and Obama are both becoming more authentic. However, Romney maintains a small lead (4.6 vs 4.2; compared to the first debate where Romney beat Obama by 4.4 vs 3.5).
The warmth and optimism award: Barack Obama
In the first debate, both men were quite similar, with Obama leading slightly. In the second debate, Romney made strides in sounding more optimistic. Ultimately, however, both were consistently emotionally distant. Contrary to common wisdom, the rate of anger words was virtually identical to the first debate. Most pundits claim that this was a more hostile debate but in anger word usage, it wasn’t.
The evasiveness award: Barack Obama
Using Language Style Matching, or LSM, we can calculate the degree to which each presidential candidate uses language congruent with the moderator Candy Crowley or the TV-studio audience question-askers. We assume that the more similar the language style, the more the candidate is answering the questions and reflecting the thinking of the questioner. In the first debate, both men were equally evasive with Lehrer.
In the second debate, both men answered all the questions more directly. However, Romney was reliably more on target with the questions than was Obama (LSM coefficients for Romney were .94 vs .92 for Obama with Crowley; .90 for Romney vs .86 for Obama with audience questions).
Interestingly, the two candidates matched in their language styles slightly more in the second debate (LSM coefficient of .95) relative to the first debate (.92), indicating greater engagement with one another. So, although the candidates were not firing more angry words, the more subtle language style metric indicates that the candidates are increasingly vigilant about what the other is saying; things are indeed heating up.
The Big Picture
Standing back, what does all of this mean about the basic personality or character of both men? It’s important to emphasize that both men use language the same ways they did four years ago. The debates are simply revealing that they are the same people we have always known.
Obama. Barack Obama is psychologically cool. He does not use personal or emotional language at high rates. Although he is a narrative thinker, he is not close to people. If you listen to the way he answers questions, he tends to provide a brief history and context. He is highly confident in his abilities. Above all, he is consistent. He has probably changed less in the last four years than any other president since Franklin Roosevelt changed during their first term.
Romney. Mitt Romney is surprisingly similar to Obama. He is psychologically distant although he uses I-words at rates slightly higher than Obama. He is not particularly emotional nor is he socially connected to others. The one difference from Obama is that Romney’s thinking is very male-like. He uses a very high rate of articles and nouns — a sign that he tends to categorize and label problems in a traditional way. The way he approaches problem is not unlike that of George W. Bush, Jimmy Carter, and Richard Nixon.
Tune into Sosolimited’s web app, reconstitution2012.com, during the third debate (10/22/12) for a live language analysis. Following the third debate, we will explore the psychological dimensions of both candidates in greater detail.
Also, check out a brief video discussion: http://youtu.be/yWE9V-P_q-Q
October 12, 2012
by Cindy K. Chung and James W. Pennebaker
In last night’s (10/11/12) vice-presidential debates, Vice President Joseph Biden and Congressman Paul Ryan presented their views on domestic and foreign policies. Martha Raddatz of ABC News moderated the debate. Using our word watching tools, we noticed some interesting throwbacks to previous blunders, flubs, and gaffes by each party.
At one point in the evening, in defense of several unfortunate comments that Romney has made in recent weeks, Ryan made an indirect reference to Biden’s previous blunders: “I think the Vice President knows very well that sometimes the words don’t come out of your mouth the right way.” Biden seemed to take the comment in stride (chuckling, “But I always say what I mean.”), and took the opportunity to again emphasize Romney’s disparaging 47% comment.
Ryan was the most emotionally expressive in his word use (5.45% of his overall word use were emotional words) relative to Biden last night (4.81%), and relative to Obama (4.73%) and Romney’s (4.87%) in the first Presidential debate last week (see our 10/3/12 post below). This effect was largely due to his greater use of negative emotion words (2.39% vs. 1.96% by Biden, 1.93% by Romney, and 1.15% by Obama). Both vice-presidential candidates were more hostile than the presidential candidates, using anger words at twice the rate of the presidential candidates. Based on his function word use, Ryan was by far the least authentic of the four candidates.
Interestingly, Ryan used first person plural pronouns (e.g. we, us, our, etc.) at high rates (4.56% vs. 3.04% by Biden, 2.29% by Romney, and 3.51% by Obama), perhaps having taken Romney’s erroneous introduction of his running mate as “the next President” too seriously, or more likely, perhaps subtly reaffirming Ryan’s strategic role in warming more right wing supporters to the Republican ticket.
See what else we’ve been working on….
More analyses on Word Watchers to follow.
For real-time language-based data visualizations of the second (10/16/12) and third (10/22/12) Presidential debates, tune in live to http://reconstitution2012.com, a collaborative project by the renowned art and technology studio, Sosolimited.
October 3, 2012
by James W. Pennebaker and Cindy K. Chung
It was the beginning of the World Series for word counters. After a four year break, the WordWatchers team came out of hibernation to analyze the first presidential debate of the 2012 election. Using LIWC (Linguistic Inquiry and Word Count, LIWC) and other text analysis tools, we analyzed the ways that Romney, Obama, and the moderator Jim Lehrer used words.
As a brief refresher, much of our research focuses on the ways people use function words, such as pronouns (I, you, she), articles (a, an, the), prepositions (to, of), auxiliary verbs (am, was), and a handful of other word categories that most people overlook. Function words are the most forgettable words in our vocabulary. Nevertheless, they usually account for over 55% of the words we read, say, and hear. Their magic is that they often reveal honesty, emotional state, status, and the quality of social relationships. Function words, then, are the hidden words that tell us a great deal about people. And that would include the presidential contenders.
The analysis of the language of the participants in the October 3rd debate yielded some interesting if not surprising results. Tonight, we have four awards to present. All are based on scientifically validated research methods. Drum roll please.
Mr. Authenticity: Mitt Romney
This is a bit of a surprise. Linguistic authenticity is calculated based on the rates that people are self-referential (use of the word “I”) plus markers of cognitive complexity (exclusive words such as without, but exclude), while at the same time avoiding negative emotion words. Overall, Romney rated a 4.35 to Obama’s 3.46. Authenticity is something that Romney has been criticized for in the past but tonight he connected.
Mr. Optimism and Warmth: Barack Obama
Across previous analyses, Obama has consistently come across as emotionally distant. He rarely uses I-words and his rates of both positive and negative emotion word use is low. Tonight’s debate was a subtle departure from his usual emotional tone in that he used slightly more positive emotion words than Romney (3.6 versus 3.0). In reality, the battle between Obama and Romney for the Mr. Optimism and Warmth title is a fairly sad race. Although Obama won, his competition barely made it to the plate (warmth score for Obama was 2.5 and Romney’s was 1.1).
The Evasiveness Award: A dead tie
Much of our research focuses on language style matching, or LSM. LSM can be measured by calculating the similarity of function word use of two people — in their speeches, letters, or emails. Using an algorithm we have written (click here for a demonstration), we can determine the degree two people are directly communicating with each other. For the debate, we wanted to know the degree to which each of the candidates matched the language style of the moderator, Jim Lehrer. In theory, the more similar they used language in response to his questions, the more they should have been answering his questions. The less they matched Lehrer’s language, the more evasive they were likely to be. The numbers were compelling. Romney matched Lehrer at 0.86 compared to Obama’s match of 0.87. Both of these numbers are relatively low for natural conversations. Interestingly, Romney and Obama’s LSM scores for each other was 0.93. In other words, they were paying attention to each other and not the question asker.
Stay tuned to this website for the remainder of the debates. Some more analyses follow.
August 18, 2011
On Tuesday, August 16, Rick Perry was surprised to learn that other politicians did not approve of his statement claiming that it would be “treasonous” if the Federal Reserve Chair printed more money. Oddly, throwing around the word treason is not something that most presidential candidates have historically done. According to the Austin American Statesman:
Perry didn’t back down. “Look, I’m just passionate about the issue,” he said in Dubuque, Iowa. “And we stand by what we said.”
Did you catch that? “And we stand by what we said.” Hmmmm. We? Who exactly is we? This is a classic way that people psychologically distance themselves from what they are saying.
Watch the pronouns.
James W. Pennebaker
Author of the forthcoming book, The Secret Life of Pronouns (NY: Bloomsbury)
By James W. Pennebaker and Raj Persaud
With the final debate on Thursday night, people have heard each of the three candidates spew over 17,000 words – that’s more than the average human says in a full day. Using computerized text analysis methods, we now have a fairly good picture of how each of the candidates uses language within the debate setting.
Recall that the words people use in everyday speech reflects who they are. Of particular relevance are a group of words called function or junk words. These almost-invisible words include pronouns (such as you, me, they), articles (a, an, the), prepositions (for, with, to), etc. The ways people use these words can tell us about speakers’ emotional states, formality, honesty, thinking styles, and other dimensions of how they approach their worlds.
Gordon Brown, David Cameron, and Nick Clegg all have distinctive speaking styles that highlight different parts of their personality. Some of these differences are obvious; others are not.
Optimism. People who are upbeat and optimistic tend to use present and future tense verbs, first person plural pronouns (we, our, us), simple words, and words that denote positive feelings and, at the same time, tend to avoid words that express negative feelings. Although the first debate found Clegg and Cameron to be quite high in optimism, Clegg’s upbeat language has moderated whereas Cameron’s has increased ever since. For last night’s debate, David Cameron was by the most upbeat and optimistic followed by Clegg with Brown far behind.
Want an upbeat, optimistic PM? Vote David Cameron.
Language Markers of Optimism
|Present tense verbs||11.53||12.55||11.62||High|
Note that numbers refer to percentage of total words used by each of the candidates. So 11.53 percent of all of Brown’s words were present tense verbs. The Direction column indicates what numbers are associated with high optimism. That is, optimism is associated with high use of present tense verbs and low rates of negative emotion words.
Honesty. Over the last 10 years, more than a dozen studies of all kinds have analyzed the language of honesty and deception. At least five language dimensions are reliably linked with honesty and another 3-4 are associated with lying. People are more likely to be telling the truth if a) their sentences are longer and more complex, b) they use I-words more (e.g., I, me, my), c) they use bigger words, d) the make more references to time and motion, and e) they use more self-reflective words such realize, understand, and think. The best markers of deception are would-should-could verbs, positive emotion words, and you-words. Averaging across all these dimensions, Gordon Brown comes across as having the most honest language profile. Clegg comes in a distant second with Cameron close behind Clegg.
Want an honest PM? Vote for Gordon Brown.
Language Markers of Honesty
|Words per sentence||20.66||17.38||18.98||high|
Except for average number of words per sentence, all numbers refer to percentage of total words used by each of the candidates. The Direction column indicates what numbers are associated with high honesty. That is, honesty is associated with high rate of words per sentence and low rates of would-should-could words.
Thinking style. The various text analysis methods find that all three candidates are quite bright. They do, however, think differently. One interesting difference in thinking is how people break down a complex problem. For example, if confronted with a new challenge, one strategy is to reduce the problem into its component parts. To do this, people generally use concrete nouns (which are reflected in the use of articles) and, the more specific they become, they will need prepositions and other linguistic devices (such as relativity words) that reflect specific concepts, objects, and things. We refer to this as analytic thinking. Another approach is to trace the evolution of problems and project how they will change in the future. Looking at how events unfold over time requires more verbs, especially past and future tense. This is often called dynamic thinking.
Gordon Brown is quite analytic in his approach whereas David Cameron is strongly dynamic. Nick Clegg is midway between the other two on both of these dimensions. If you would like to see examples of these differences in thinking, read how the three responded to a question about what their party would do to help families pay for housing:
Gordon Brown: …there is a pent-up demand for housing in our country. There are one million more home owners than there were just over ten years ago, so more people are buying their homes. …Shared equity is something that might be considered because that’s a chance to buy up a part of your house, and it’s become a more popular way of doing things and we are able to help finance that and work with the building societies and banks.
David Cameron: I have every sympathy with you because, frankly, today in our country, people who try and work hard and save, and obey the rules, and do the right thing. All too often, they just find hurdle after hurdle put in their way, whereas people who actually don’t play by the rules, who don’t think about saving and don’t think about their behaviour often get rewarded and that’s not right.
Nick Clegg: … this is one of the things that I, along with immigration, actually, that I probably hear about more than anything else as I travel around the country, a lack of affordable housing as I travel round the country. The lack of affordable housing. The people in your situation, but then there are, I think, 1.8 million families, that’s five million people, who are still on the waiting list for an affordable home.
As you can see, Brown is coolly analytic about the problem, evaluating what aspects of the economy may be contributing to the problem. Cameron traces what people have done and are doing. He sees the problem more of the action of others in the past and the present. Clegg actually doesn’t think much about the problem at all but, instead, later talks about how he would fix it.
Want an analytic thinker? Vote Brown.
Want a dynamic thinker? Vote Cameron
Want someone who is somewhere in between? Vote Clegg.
Caveat: Linking natural language use with social, cognitive, and personality dimensions is a relatively new science. It’s important to think of it in probabilistic terms. Our approach is more accurate than flipping a coin but far from 100% accurate. Also, these analyses are based purely on how the candidates spoke in the debates. As we’ve seen, once the microphones are thought to be turned off, the candidates may actually talk and think differently from how they might appear on the international stage. Finally, optimism, honesty, and thinking styles are important qualities of leaders. Remember that there are dozens of other qualities that contribute to good leadership that we are not measuring here. Consider these traits just the tip of the iceburg.