Trump scores low for grammar, but not lowest
This year's fierce battle for the presidency is mostly being waged with words a middle schooler could understand.
This year's fierce battle for the presidency is mostly being waged with words a middle schooler could understand.
At least one candidate - guess which one? - lands even lower when it comes to sentence structure, but he still fares better than one of our former presidents, according to a new analysis of candidate speeches done at Carnegie Mellon University's Language Technologies Institute.
Maxine Eskenazi, LTI principal scientist, and Elliot Schumacher, a graduate student, performed a "readability analysis" of multiple speeches by five candidates - Ted Cruz, Hillary Clinton, Marco Rubio, Bernie Sanders and Donald Trump - as well as single speeches by four former presidents and President Obama.
It was based on the idea that some words and types of grammar are more common at certain ages than others. For instance, the word "win" is common in documents aimed at 3rd graders. "Successful" shows up more often in later grades. Dependent clauses are another mark of higher education levels. The study did not look at whether a candidate's grammar was correct or not.
One caveat is that most of us use shorter sentences and simpler grammar when we talk than when we write. While tests of this sort are meant for the written word, the researchers chose the readability model they thought was the best fit for speeches.
What they found was that most of the candidates used grammar typical of students in grades 6 to 8. Trump's grammar fell just below 6th grade, but not as low as former president George W. Bush's, at around 5th grade. Rubio, who has suspended his campaign since the study began, was the highest of this year's candidates, about a half-grade above Cruz, Clinton and Sanders. (John Kasich was not included in the analysis.)
The champ was Abraham Lincoln, whose Gettysburg address scored around 11th grade.
When it came to word use, Bernie Sanders was doing the most to stretch his listeners' vocabularies. He came in above 10th grade. Rubio was a notch lower. Cruz and Hillary Clinton were both above 9th grade. Trump again trailed, at a little above 7th grade.
Eskenazi said the candidates have trended toward simpler communication in recent weeks.
Hillary Clinton's speeches showed the most variation in vocabulary, a possible sign of tailoring her words to specific audiences. Her speech at a town hall meeting on immigration in Nevada last May was at the 6th grade level. When she spoke at the Schomburg Center for Research in Black Culture in Harlem last month, she used 11th grade words.
Cruz's vocabulary varied the least.
Eskenazi said she is particularly interested in how candidates change the way they talk for different groups. Ideally, she'd like to compare speeches given to the same group. "I think it would be interesting to see, I guess, what their opinion of their audience is," she said.
Sanders hit the group high - 12th grade - for vocabulary in his election night speech in New Hampshire. Trump scored the low - 5th grade - in his victory speech in Nevada last month. That was also the group's low for grammar: 4th grade.
In the vocabulary comparison, George W. Bush was well above Trump. His speech was at the 10th grade level, the same as President Obama's. Former President Ronald Reagan was the highest, at 11th grade.
An earlier analysis by the Boston Globe was based on sentence length and the number of syllables in words. It found Trump speaking at the 4th grade level, four to five grades below most of his current competitors and six below Sanders.
Arguably, an analysis of speeches may say more about a candidate's speech writers or target audience than it does about the candidate. The Carnegie Mellon team decided not to analyze how candidates communicated during debates when they were - mostly - choosing their own words.
Eskenazi said candidates choose their writers and have ultimate control over what their speeches say. The debates this year have been "so crazy" that they would be difficult to analyze, particularly because there are no reliable transcriptions. She said crowd sourcing might make that possible.
The debates - or interviews - would also require a different kind of analysis, she said, because written and spoken communication are so different.
215-854-4944
@StaceyABurling