“The ballots are a disaster,” President Trump said almost two weeks ago. During his debate with Democratic presidential nominee Joe Biden last week, Trump said, “This is going to be a fraud like you’ve never seen.” It’s a claim he has been repeating for months, in his Twitter feed, in interviews, as often as he can. He falsely asserts that large numbers of ballots are discarded, that mail-in voting invites fraud, and that his supporters need to “monitor” polling places in cities.

Some Republicans have dismissed Trump’s rhetoric as inconsequential. When Trump refused to commit to a peaceful transfer of power, Sen. Ben Sasse (Neb.) commented, “He says crazy stuff. We’ve always had a peaceful transition of power. It’s not going to change.”

But many observers, including political scientists, are alarmed. Political scientist Jacob Levy argues that any entreaty to ignore Trump’s language “is faux pragmatism, ignoring what is being communicated to other countries, to actors within the state, and to tens of millions of fellow citizens.”

Words have power. Words can be used to bolster democratic norms in the public, and words can be used to subvert these norms. Political scientists Russell Muirhead and Nancy Rosenblum argue that Trump’s “new conspiracism” — such as calling an election “rigged” — affects “citizens’ attitudes and emotions, insisting that the defining elements of political order are not supported.”

So does language like this undermine trust in the election? To find out, I worked with political scientist Kimberley Guiler to understand the effects of vote-rigging rhetoric. In our recently published research, we found that allegations of vote rigging do weaken support for democratic norms.

Here’s how we did our research

During the last presidential election, then-candidate Donald Trump alleged that Democrats were rigging the vote. The national security community had announced concerns about Russian election interference. We sought to understand whether these concerns would hurt American public confidence in the election’s legitimacy.

We did this through a survey experiment of 1,000 U.S. voting-age adults recruited via Amazon’s Mechanical Turk; the data and code are posted here. The study was administered the day before and on the morning of the 2016 presidential election. After participants took the survey, we told them what it was about so as not to diminish their enthusiasm for voting, though many of them had already voted early.

Participants first read a short mock news excerpt, in which an election expert claimed either that 1) the election would be secure, 2) Democrats were trying to rig the election for Hillary Clinton, or 3) Russians were trying to rig the election for Trump. Of course, the latter two scenarios aren’t precisely comparable. One story accuses the Democrats of rigging the election. The other story accuses a foreign government — Russia — of rigging the election. We chose these based on accusations that were widely circulated at the time. We could have matched the culprits superficially, with one story accusing the Democrats and the other accusing the Republicans, but we did not think the second would sound likely.

After reading the story, we asked participants whether they thought fraud was likely. Next, we asked about their emotions and about their confidence in the election, their willingness to accept the results, and their beliefs about how important it would be for the loser to concede. Finally, they provided basic demographic information.

Republicans and Democrats responded differently

The stories only changed participants’ beliefs about whether fraud was likely if the fraud would hurt the participant’s political party. In other words, Democrats were willing to believe only that Russia was trying to manipulate the election, and Republicans were willing to believe only that the Democrats were trying to rig the election.

Further, allegations that Russia was trying to rig the election in favor of Trump made Democrats less confident in the election and less likely to accept the results. Allegations that Democrats were trying to rig the election in favor of Clinton made Republican respondents less confident in the election, less likely to accept the results, and less likely to say it would be important for the loser to concede.

But the two groups responded differently in two key ways. Hearing about vote rigging — no matter who was responsible, Democrats or Russians — made Republicans less willing to accept the results. This may be because Republican respondents considered both the Russians and the Democrats to be outsider groups, easily mistrusted. Further, while allegations against the Democrats made Republicans less likely to think the loser should concede, Democrats thought it important for the loser to concede regardless of whether they were told that Russians were trying to rig the election.

The difference may grow from Trump’s language. In 2016, much as he’s doing now, he used his false allegations about vote fraud to explain that he might not concede the election.

What does this mean for November?

Our study doesn’t necessarily tell us how Americans will respond this fall. We relied on a newspaper clipping, not on a politician’s words. Trump’s own allegations of vote rigging might actually influence Republicans more than the report we offered. Or it might do the opposite: A politician’s allegation could be explained away as self-interested, and Americans might consider an “expert” source more credible. Further, our study offered a single dose of a news story. Both social media, and the news and opinion media, are amplifying Trump’s claims about fraud; repeated exposure might intensify the effect.

The way people talk about elections has consequences. Our democratic norms are sustained — or eroded — through our language and our actions. Allegations that the election is rigged reduce Americans’ support for the democratic norms that underlie the U.S. system of government. Senate Republicans recently expressed confidence that a Biden electoral victory would result in a peaceful transition of power, without directly mentioning Trump’s language. This might help, but the president has the loudest microphone and the largest audience for his claims.

Bethany Albertson (@AlbertsonB2) is an associate professor at University of Texas at Austin and co-author of “Anxious Politics: Democratic Citizenship in a Threatening World”(Cambridge University Press, 2015).