Last Thursday, White House aide Kelly Sadler started a political firestorm when she reportedly joked about Sen. John McCain’s impending death in a staff meeting. The White House declined to apologize. McCain, suffering from brain cancer, vocally opposes Gina Haspel, President Trump’s nominee to run the CIA, who faces a tough confirmation vote because of her association with President George W. Bush’s torture program. Fellow Republicans have been attacking McCain’s stand.

This is hardly the first time that McCain has frustrated his own party. The longtime senator, former Republican presidential nominee and former prisoner of war has long been known as a “maverick.” He was strongly criticized in 2017 when he and two other Republicans voted against the Obamacare repeal, saving the law by one vote.

But given his diagnosis, the attacks now are striking, if not startling. Our research suggests that one way to view them is that they reveal in-group policing – designed to root out disloyalty at a time when party conflict is high and margins for victory are extremely slim. For people with an intense loyalty to their party, at a time when a single defection can change a vote’s result, some even wish harm on their own members who buck the party line.

Here’s how we did our research

Partisanship encourages intense animosity against partisan opponents, as a great deal of research shows. But partisan conflict can also increase anger at fellow party members who don’t toe the line.

We designed a national survey to measure extremely hostile partisan attitudes, both toward opponents and toward politicians in one’s own party who buck the party’s agenda — like McCain.

YouGov administered the online survey to a nationally representative sample of 1,000 U.S. adults in November 2017. The data was collected through the 2017 Cooperative Congressional Election Study led by Liz Salazar and Brian Schaffner, funded in part by the National Science Foundation. About three-quarters of respondents identified with (or felt close to) Democrats or Republicans. In this analysis, we exclude those who did not express any kind of party leaning.

A small proportion of partisans people feel extreme hostility toward opponents

We asked specifically if respondents wished physical harm against both kinds of political opponents.

For instance, we asked, “when a [your party] politician votes against the party on a key issue, have you ever wished they would get sick and die?” Five percent of partisans said yes, 95 percent said no. While that number is low, it isn’t zero. Keep in mind this is not just dislike or disfavor; it is wishing death on fellow partisans.

To compare extreme in- and out-party animosity, we asked, “Have you ever wished that someone would physically injure one or more politicians?” While 82 percent said no, 12 percent said they wished someone would injure politicians in the opposing party. Four percent of partisans said they wished someone would injure leaders in both parties, and 1 percent said they wished injury mostly for leaders in their own party.

Finally, we asked, “Who would you rather see bad things happen to: [opposing party] politicians, or [own party] politicians who vote against the party on a key issue?” Three-quarters chose “neither,” and 10 percent chose the other party. Nine percent wished harm equally against in-party defectors and partisan opponents. And 4 percent declared that defectors — apostates — deserved more harm.

In sum, only a minority of politically focused Americans express violent hostility against either partisan opponents or mavericks in their own party; of the those, the other party is the more common target.

Still, that means that a minority of partisan citizens — tens of millions — do sometimes wish for political violence toward those they oppose.

Who holds such extremely hostile views?

In general, the individuals most likely to endorse harm against political opponents (whether in their own or the other party) hold strong partisan loyalties; see the other party as an existential national threat; or have aggressive personality traits.

Strength of party attachment matters in interesting ways. Among the strongest half, 3 percent wished someone would injure apostate politicians in their own party; no weakly identified partisans said the same. But fully 7  percent of weak partisans wished injury on opponents in both parties equally, more than the 1 percent with strong loyalties.

Of the 20 percent of partisans who strongly agreed that the opposing party is “downright evil,” 10 percent often wished death on disloyal party leaders. That’s more than we found in the 25 percent of those who somewhat agreed, where 6 percent wished for in-party opponents’ deaths, and three times more than the majority who did not see the other party as evil.

Seeing the other party as “evil” increased the likelihood of wishing to see opposing partisans injured by 14 percentage points. This group was also 11 points more likely to wish for general harm against both opponent types equally.

But the people most likely to support political aggression were those who reported more aggression in their everyday lives. Among those in the top quarter of aggressive traits, 10 percent wished death on disloyal partisans, while only 4 percent of the rest did. Aggressive traits also increased support for out-party harm.

We also tested partisan differences but found no clear asymmetries between Democrats and Republicans.

Of course, we should be cautious interpreting such responses. Some respondents may hide their real, unpopular views. If so, then our results may underestimate the prevalence of extreme attitudes — although the fact that our survey took place impersonally, online, reduces this concern.

Conversely, some individuals may show team loyalty even if they don’t truly wish anyone physical harm. Even then, expressive hostility — which appears frequently online — is itself a troubling indicator of extreme political incivility.

What does all this tell us?

In an era of partisan polarization, we tend to focus, appropriately, on conflict between the two parties. But when polarization is high, even “disloyal” members of one’s own party may face intense vitriol from their party’s voters.

Lilliana Mason (@lilymasonPhD) is assistant professor of government and politics at the University of Maryland, College Park, and author of Uncivil Agreement: How Politics Became Our Identity (University of Chicago Press, 2018).

Nathan Kalmoe (@NathanKalmoe) is assistant professor of political communication at Louisiana State University’s Manship School of Mass Communication and department of political science, and co-author of Neither Liberal Nor Conservative: Ideological Innocence in the American Public (University of Chicago Press, 2017).

This study was supported by the Tom Jarreau Hardin Professorship at LSU’s Manship School.