Threats of political violence are all around us. House Speaker Nancy Pelosi’s husband was viciously attacked in their San Francisco home by an assailant looking to harm his wife. This comes after the House of Representatives had already beefed up its own security in July in response to a growing number of violent threats and attacks. And we’re just a few months away from the second anniversary of the worst violent attack on the Capitol since the War of 1812.

As a nation, we appear to be more comfortable with political violence than ever before. A report from the Carnegie Endowment for International Peace found that a fifth of Republicans and 13 percent of Democrats thought that political violence was justified in some cases. And as we move into the midterms, election deniers are on the ballot throughout the country. Many on the far right, meanwhile, are preparing to challenge the results of the election.

This trend paints a picture of American democracy under duress. For the past decade experts have been looking at ways to reverse these divisions. Much of that research has focused on reducing political polarization. If you empathize with those in the opposing party, or so the thinking goes, the less likely you will be to engage in political violence or other antidemocratic attitudes against rivals.

But focusing on polarization and political animus may be misguided, according to a body of research from Stanford University political scientist Robb Willer and colleagues. As part of a large-scale research project called the Strengthening Democracy Challenge at Stanford University, Willer and his team have been investigating “interventions” for preserving American democracy.

The team’s findings are documented in a new study released October 31 in Nature Human Behavior. In it, researchers following a large sample of 8,385 people, found that it’s relatively easy to reduce political polarization using simple online exercises. For example, by thinking of a friend from an opposing political party or correcting misperceptions of a rival’s extreme views.

The surprise in the research team’s study was the finding that reduced polarization doesn’t necessarily translate into a safeguard against antidemocratic attitudes, encompassing voting for undemocratic candidates, a willingness to sacrifice democratic principles and support for political violence. Those who participated in the “friendship” and “misperception” interventions as part of the study showed no reduction in their appetite for political violence or antidemocratic attitudes. One important exception, however, was that those who began with less animosity for the opposing party were less likely to support undemocratic candidates. The researchers surmised that they didn’t bear so much enmity toward their rivals that they were unable to cross party lines when voting. The study’s conclusions about reducing political animosity were not universally endorsed by other researchers in the field.

Social science researchers have long focused on the implications of partisan animosity rather than violence as a means of studying political differences. But times have changed. “Until recently political violence was pretty rare and it was unheard of that elites would tolerate or encourage it in any way,” Willer says. It was also thought that reducing animus would have a direct link on reducing violence. But this study found that if you want to take on political violence, it needs to be done through direct measures—taking time, for instance, to understand the other party’s perceptions of violence and initiating steps to change those views. And according to the research, any initiatives to reverse this trend should begin at the highest levels of U.S. political discourse. The public, Willer says, takes cues about nonviolent responses to pressing national issues from trusted political elites. Unfortunately, as illustrated by the events of January 6, 2021, the opposite messaging can also hold sway. When leaders foment violent rhetoric, some followers may be quick to get behind it.

Still, the most powerful tool for reducing political violence has to do with building a clearer picture of how rival parties view any call to arms. In another of Willer’s studies, published last April in the Proceedings of the National Academy of Sciences (PNAS), researchers found that correcting misperceptions about political violence across the red-blue divide reduced support for violence by 44 percent. “We have drastically inflated views of how the other party views violence,” Willer says.

Similar approaches can also reduce support for undemocratic candidates, says study co-author and Stanford University doctoral student Jan Gerrit Voelkel. We tend to be exposed to only the least palatable political rivals, which means even if our party’s candidate holds undemocratic values, we’re slow to cross party lines. A voter tends to filter out all but the most extreme views from the opposite political party and ignore any trace of undemocratic values from their own party. Crossing party lines, as a result, tends to be a rare occurrence. “Being exposed to more relatable candidates from the other party reduces the need to vote for undemocratic candidates,” Voelkel says.

And news about partisan divisions isn’t all negative. University of Pennsylvania political scientist Matthew Levendusky observes that just because parties can’t seem to get along does not signal any readiness to destroy opponents or topple democracy. Dislike for the other party may fester but far fewer embrace partisan violence, he says. Willer’s research also shows that there’s no easy link tying partisan animus, undemocratic politicians and violence.

Levendusky points to a PNAS study published in March, which found that current research overstates the American taste for violence. In a forthcoming paper, Levendusky and his team found that support for the January 6 attack on the Capitol was limited. “Animus is only weakly related to partisan violence, suggesting that violence—while incredibly troubling, and worthy of careful study—is not the same thing as dislike and distrust of the opposition,” he says.

Jennifer Wolak, a political scientist at Michigan State University, agrees. “There is a small minority of people who are willing to support undemocratic behaviors,” she says. “But they are the exception rather than the rule, and overall make up a very modest share of the electorate.” Wolak says that 70–85 percent of Americans reject violence. So while the assault on the Capitol and other incidents may achieve high visibility in news headlines, fears of a wider contagion of violence could also be inflated.

Additionally, both Levendusky and Columbia University social psychologist Peter Coleman question methods used in the Nature Human Behavior study. They challenge the premise that solid conclusions can be achieved using brief online surveys. “Eight-minute interventions aren’t going to move the needle,” Coleman says. “If you’re deep into partisan violence, thinking about someone from the other side who’s nice isn’t going to change that,” he says. Longer-term exchanges between voters from different parties might at times achieve more enduring results.

In his book The Way Out: How to Overcome Toxic Polarization, Coleman contends that while these issues are complex, reengaging with the other side can make a difference. He points to Watertown, a small city in upstate New York, where Trump voters abound. A 2019 article in the Atlantic, described Watertown as among the most politically tolerant places in the country, largely because around a quarter of couples are in marriages in which partners belong to opposite political parties. Coleman contends that such couples have different conversations surrounding politics and more politically tolerant kids. “When red and blue Americans mix together in mundane ways like bake sales, Little League and choir, it’s a good thing,” Coleman says. “You can’t,” he adds, wait until you’re in a civil war to engage.”

With the midterms upon us and with constant daily reminders of the prevailing political toxicity on news and social media, no abundant quick fixes are to be found. Both Willer and Coleman agree that changing attitudes will require a slow process of national engagement. Willer’s next steps are to apply these interventions in real-world settings by appealing to politicians, their parties and bipartisan organizations as well as social media platforms to deploy these strategies on the broadest possible scale.