They and We. I am but We are. According to a statistical analysis by Simon DeDeo of the first 2016 presidential debate on September 26, the strongest distinction between Clinton and Trump was not what, but who.

Clinton's use of we was 50 percent more prevalent than Trump's, and Trump used they 70 percent more often than Clinton.

Perhaps most surprising, however, was the strength of the differences (or, in this case, similarities) in word choice. "In the modern history of presidential debates so far, we've almost never seen two candidates share so many of the same patterns and vocabulary," DeDeo says. "Candidates usually distinguish themselves from each other much more strongly. If you look at Obama-McCain and Obama-Romney, you find their use of language both richer and more distinct."

DeDeo performed his overnight statistical textual analysis using transcripts from the first Clinton-Trump debate and those from previous presidential debates. He presented his results the following morning during a live online news conference with other scientists who analyzed the candidate’s facial expressions, mannerisms, and other performance indicators. The event was a partnership between SFI and Newswise, a science news distribution service.

"To see something like what we saw that night, you have to go back to Bush-Gore in 2000," he says. "In fact, both candidates share a great deal of the language not just from those debates, but from the party positions at the conventions themselves. In terms of their rhetoric, the candidates are in a race to the center."

"Words -- even single words, or pairs, taken in isolation -- do an enormous amount," says DeDeo. "A hot-button word like super-predator can be a topic itself. Beyond their choice of pronouns, one of the words that best distinguished Clinton from Trump was think; Trump, for his part, urged viewers to look."

Says DeDeo: "The difference between telling someone what you think versus telling someone to look encodes a great deal -- and along with the insider/outsider pronoun choices, it's one of the biggest splits between the candidate's choices of words we find."

But the really surprising finding is how few such splits there were. Comparing to the last three elections, splits should be far more common. "It's an utterly crucial moment in the country's history," DeDeo says. "Yet the candidates chose their words in ways at odds with the starkness of the choices they present."

DeDeo's analysis relies on information theory, a set of tools developed at the birth of the networking age. Information theory allows a scientist to quantify the strength of a signal -- whether it be a radio wave signaling a cell phone, or a choice of word signaling a political affiliation or an underlying meaning.

"A candidate at the podium is signal-generating machine," he says. "Everything from the pauses in their speech to the inflections and pitch of their voices carries information, and often unexpected patterns of similarity to what's come before, and what comes next," says DeDeo. "It's a symphony of signals. If the words themselves are first violin, it's fair to say that the messages are to be found somewhere else in the orchestra."

Simon DeDeo is an external professor at the Santa Fe Institute, and with Carnegie Mellon University's Department of Social and Decision Science. He and his collaborators have studied big arguments in everything from the British criminal justice system to the French Revolution to the post-war Serbian parliament and arguments over facts on Wikipedia.


• Pronoun choice, and look vs. think, are the strongest distinguishing signals between the candidates.
• Word patterns are very similar overall -- more similar than any debate in our records, except Bush-Gore in 2000.
• Race to the center: Of all the party platforms in our data, candidate word choices were closest to the 2000 Democratic platform -- for both Trump and Clinton.

Read the article on (September 29, 2016)

Read the article in the Pacific Standard (September 26, 2016)

Read the article in the Las Cruces Sun News (September 27, 2016)