As Rep. Mary Miller, R-Ill., embarked on her first congressional campaign, she described herself in salt-of-the-earth, all-American terms: a mother, grandmother and farmer who embodied the “Midwestern values of faith, family and freedom.”

“Hard work, using God-given talents, and loving each other well,” a voice declared over video clips of Miller, 63, embracing her family, praying and walking on her farm in an ad in early 2020.

“In the world today,” the ad continued, “we could use a lot more of this.”

But there is another side to Miller’s wholesome image. Since entering Congress, she has routinely vilified Democrats and liberals, calling them “evil” communists beholden to China who want to “destroy” America and its culture. And President Joe Biden’s plan, she seethed on Twitter last spring, is to “flood our country with terrorists, fentanyl, child traffickers, and MS-13 gang members.”

Miller’s inflammatory words underscore the extent to which polarizing rhetoric is now entrenched among Republicans in the House of Representatives, especially among those like Miller who voted against certifying the Biden victory, according to an examination by The New York Times of partisan language over the past 10 years.

The analysis of tweets, Facebook ads, newsletters and congressional speeches — more than 3.7 million items in all — relied largely on natural language processing, a technique that uses software to extract information from large amounts of text. The Times tallied words that were linked in academic research to divisive political content, as well as those identified by linguists and computer scientists to be used in polarizing ways — “fascist” and “socialist,” for example, “far-right” and “far-left.”


Republican representatives have ratcheted up such rhetoric since former President Donald Trump took office, the analysis found. In the year and a half after the Jan. 6, 2021, Capitol riot, Republicans on average used divisive words and phrases more than twice as often as Democrats in tweets and six times as often in emails to constituents.

At the forefront of this polarization are Republicans who voted to reject the Electoral College results that cemented Trump’s defeat last year. A recent Times investigation revealed how those lawmakers helped engrave the myth of a stolen election in party orthodoxy. Now a Times analysis shows that the language of the 139 objecting members is markedly more hostile than that of other Republicans and Democrats. In their telling, those who oppose them not only are wrong about certain policies but also hate their country.

The Times found that in the current Congress, representatives who fought certifying the election used polarizing language on Twitter about 55% more often than other Republicans and nearly triple the rate of Democrats. Objectors referred to their opponents as “socialist” in more than 1,800 tweets, more than twice as often as other Republicans. Democrats called the other side “fascist” about 80 times.

While provocateurs such as Reps. Marjorie Taylor Greene, R-Ga., and Lauren Boebert, R-Colo., have etched out combative personas, the analysis found that dozens of rank-and-file members, such as Miller, also joined the drumbeat of polarization.

The Times examined Democrats as well as Republicans. In the first years of the Trump presidency, Democrats on average spoke in a more outraged way than Republicans on Twitter and in constituent emails.

But Republicans have otherwise eclipsed Democrats, the analysis revealed. Republicans have more than quadrupled their use of divisive rhetoric since the early 2010s in the Congressional Record, which is dominated by stately and tedious speeches from the House floor.


Among other things, the algorithmic techniques used by the Times compared statements with one another and with examples of known incendiary language. Similar tools are used in spam filters and by companies tracking discussions about their products on social media. (One complexity is that this technique cannot always distinguish between incendiary rhetoric and factual descriptions of antidemocratic behavior.)

Political scientists at New York University reviewed and corroborated the Times’ findings and said the results underscored a broad shift in Republican rhetoric. “We are clearly living in a time in politics where this kind of aversion is increasing and also where we are seeing the emergence of an extreme faction in the Republican Party,” said Joshua Tucker, a politics professor and co-director at the university’s Center for Social Media and Politics.

Polarization on the rise

Five days after the Capitol riot, Rep. Kevin McCarthy, R-Calif., the GOP leader, implored members of his party to tone down their speech.

“We all must acknowledge how our words have contributed to the discord in America,” he said, according to a recording included in the audio version of “This Will Not Pass,” a book by Jonathan Martin and Alexander Burns, who covered the 2020 election for the Times.

McCarthy added, “No more name-calling, us versus them.”

But this spirit of reconciliation did not last. In dozens of tweets since then, McCarthy has referred to Democrats as “radical” leftists, said they prefer China to the United States and accused them of ruining America.

McCarthy’s rhetoric is in line with that of other Republicans who objected to the election.


“They are using what are called ‘devil terms’ — things that are so unquestionably bad that you can’t have a debate about them,” said Jennifer Mercieca, a professor at Texas A&M University who studies the history of political rhetoric.

Several political scientists say the factionalism is alarming because it makes compromise harder and normalizes such rhetoric throughout the population. Messages are particularly pernicious if they claim that political parties hate the United States, are in league with its enemies or cheat to win elections, the experts said.

‘Negative partisanship’

In February, Rep. Mike Rogers, R-Ala., joined the Republican National Committee chair on Twitter accusing the newly elected Democratic district attorney in New York City’s borough of Manhattan of having “radical pro-criminal” policies.

On its face, the attack had nothing to do with the House of Representatives or its speaker, Rep. Nancy Pelosi, D-Calif. But Rogers made it about her.

“Socialism on the march,” he wrote. “We must fire Pelosi.”

That same week, Republicans tweeted more than 180 times about Pelosi; 165 of those posts were from objectors like Rogers. The legislators were addressing subjects as varied as a technology bill, the Beijing Olympics and the investigation of Jan. 6.

Even when Republicans have power, vilifying their opponents has worked well for them, said Alan Abramowitz, a political scientist at Emory University. It is part of a broader trend in American politics toward “negative partisanship,” he said. “On both sides, there’s a very intense mistrust or dislike of the other party.”


A media bubble

The articles and videos that lawmakers share on social media serve as a guide to their most trusted information sources, as well as evidence of a widening gap in the news partisans consume.

Again, the objectors stand out, turning to one source above all others: Fox News. In the most recent Congress, they tweeted links to the network more than 5,300 times, at nearly twice the rate of other Republicans.

Democrats almost never link to Fox, preferring CNN, the Times and The Washington Post. They also link much more frequently to government websites such as those for finding vaccines or health coverage — sites Republicans have almost never promoted.

Objectors differ not only from Democrats but from their fellow partisans in their willingness to direct followers to more stridently right-wing sources, including The Daily Caller, co-founded by Fox News host Tucker Carlson, and Breitbart News, once led by the erstwhile Trump strategist Steve Bannon. They linked to Breitbart in nearly 900 tweets, compared with about 70 for other Republicans.

The language of faith

On social media, Miller regularly quotes the Bible and writes “Happy Sunday” messages to her followers. She posted one such tweet while taking respite from the campaign trail in June, sharing a photo of herself on the sofa with seven of her grandchildren. She wrote, “I am so blessed!”

Five days later, Miller’s Twitter took a different tone. “The Left tells our children a hopeless message that they do not come from God, they are not born for any purpose, and they cannot obtain salvation,” she wrote, before pledging to defend the right to bear arms.


In December, Miller tweeted a picture of a cloven-hoofed sculpture that the Satanic Temple, a self-described nontheistic religious group, had installed near a Christmas tree and Nativity scene inside the Illinois state Capitol. A sign said the state, which is led by Democrats, could not “legally censor” such controversial installations under the First Amendment.

Miller turned it into a line of attack — tweeting that “the left cheers this” because they “are not only an anti-American party, they are an anti-Christian party.”

“We’re at war for the heart & soul of our country,” she added, concluding, “Christ is on our side and we will prevail!”


Methodology: The Times studied lawmakers’ rhetoric by evaluating nearly 3 million tweets, more than 100,000 email newsletters, 300,000 Facebook ads and 350,000 statements from the Congressional Record from 2010 through this past June.

The analysis employed language software, Receptiviti, to tally how often the legislators used words that academic researchers had linked to antagonistic speech online.

Times reporters checked thousands of randomly selected tweets and newsletters to see whether the words were indeed being used in a hostile way. This resulted in some words being removed from the evaluation list. Reporters also performed a computer analysis on another large sample to determine which words were much more likely to appear in messages that expressed anger, disdain or distrust toward the opposing party. When negative terms were strongly identified with one party — such as “left-wing” — reporters included a corresponding term from the other whenever possible.


The material went through further analysis. Tweets were counted as polarizing if they included any words from the revised list, but in longer texts such as newsletters, multiple words were required.

The Times worked with Patrick Wu, a postdoctoral fellow at the Center for Social Media and Politics at NYU. Using a technique called trained machine learning, he found the same pattern as the Times in recent tweets.

Ashique KhudaBukhsh, a professor at the Rochester Institute of Technology who researches political polarization, also reviewed the methodology and findings.

To learn about different subjects lawmakers discussed, the Times relied on topic modeling, a machine learning technique that calculates which words are likely to cluster together to form topics. And the Times evaluated religious content using the Linguistic Inquiry and Word Count, a tool that checks texts for related words.

All of these methods are widely used by academic researchers and can evaluate enormous amounts of text, but they are not precise enough to rank individual lawmakers. This is because any given message can produce a false positive or negative. Someone may use an angry word in criticizing such language, for example, or speak in terms too subtle to register as polarizing.

Over hundreds of thousands of examples, these types of errors tend to even out to produce reliable indications of patterns. The Times’ findings about the current congressional term were statistically significant.

The Times obtained data from Twitter; the Congressional Record; Facebook’s Ad Library; the Congressional tracking company LegiStorm; and DCinbox, an academic project that collects email newsletters. Tweets and ads from government accounts as well as campaign accounts were included. The Times collected the data in conjunction with the Tow Center at the Columbia Journalism School.