"How Civil Wars Start," a new book by political scientist Barbara F. Walter, was cited all over the place in the days around the anniversary of last winter's riot at the Capitol. The New Yorker's David Remnick, Vox's Zack Beauchamp and my New York Times colleague Michelle Goldberg all invoked Walter's work in essays discussing the possibility that the United States stands on the edge of an abyss, with years of civil strife ahead.

The book begins with a story from the fall of 2020: the kidnapping plot against Gov. Gretchen Whitmer of Michigan, hatched by a group of right-wing militiamen who opposed Whitmer's pandemic restrictions. Fortunately "the FBI was on to them" and foiled the plot — but the alleged kidnapping conspiracy, Walter argues, is a harbinger of worse to come. Periods of civil war often "start with vigilantes just like these — armed militants who take violence directly to the people."

Here's a skeptical question, though: When we say the FBI was "on to" to the plotters, what exactly does that mean? Because at the moment the government's case against them is a remarkable tangle. Fourteen men have been charged with crimes, based in part on evidence reportedly supplied by at least 12 confidential informants — meaning that the FBI had almost one informant involved for every defendant.

And according to reporting from BuzzFeed's Jessica Garrison and Ken Bensinger, one of these informants, an extremely colorful convicted felon named Stephen Robeson, appears to have been a crucial instigator of the plot. He is alleged to have used government funds to pay for meals and hotel rooms, to have encouraged people "to vent their anger about governors who enacted COVID-19 restrictions" and "to plan violent actions against elected officials and to acquire weapons and bomb-making materials," and to have followed up aggressively, calling potential plotters "nearly every day."

Robeson's role has become enough of a headache for the prosecution, in fact, that they recently disowned him, declaring that he was actually a "double agent" (meaning triple agent, I think) who betrayed his obligations as an informant by trying to destroy evidence and seeking to warn one of the accused conspirators ahead of an arrest. Prosecutors had already ruled out testimony from an agent who ran one of their key informants, probably because he spent much of 2019 trying to drum up business for his private security firm by touting his FBI casework.

Presumably we'll find out more about all this when the case comes to trial, but for now it's reasonable to wonder whether Whitmer's would-be kidnappers would have been prepared to go all the way with their vigilante fantasies, absent some prodding from the feds.

And those doubts, in turn, might be reasonably extended to the entire theory of looming American civil war, which assumes something not yet entirely in evidence — a large number of Americans willing to actually put their lives, not just their Twitter rhetoric, on the line for the causes that currently divide our country.

Overall, the academic and journalistic literature on America's divisions offers a reasonably accurate description of increasing American division. The country is definitely more ideologically polarized than it was 20 or 40 years ago; indeed, with organized Christianity's decline, you could say that it's more metaphysically polarized as well. We are more likely to hate and fear members of the rival party, more likely to sort ourselves into ideologically homogeneous communities, more likely to be deeply skeptical about public institutions and more likely to hold conspiratorial beliefs — like the belief that Joe Biden and the Democrats stole the 2020 election — that undercut the basic legitimacy of the opposition party's governance.

At the same time, the literature suffers from a serious liberal-bias problem, a consistent naiveté about the left and center's roles in deepening polarization. For instance, in the Bush and Obama eras there were a lot of takes on the dangers of "asymmetric polarization" — the supposed ideological radicalization of the Republicans relative to the Democrats. Across most of the 2010s, though, it was clearly liberals who moved leftward much more rapidly, while Republicans basically stayed put — and yet somehow the perils of that kind of asymmetry get much less expert attention.

Likewise the drama of protest politics in 2020 is often analyzed in a way that minimizes the revolutionary symbolism of the left's protests — the iconoclasm and the toppled statues, the mayhem around federal buildings and the White House, the zeal to rename and rewrite — and focuses intensely on the right's response, treating conservative backlash as though it emerges from the reactionary ether rather than as a cyclical response.

The other bias in the civil-war literature is toward two related forms of exaggeration. First, an exaggerated emphasis on what Americans say they believe, rather than what (so far, at least) they actually do. It's absolutely true that if you just look at polling data, you see a lot of beliefs that would seem to license not just occasional protest but some sort of continuing insurrection. This includes not only the Trumpist stolen-election theories but also popular beliefs about recent Republican presidents — that George W. Bush had foreknowledge and allowed Sept. 11 to happen or that the Russians manipulated vote tallies in order to place Donald Trump, their cat's-paw, in the White House.

However, an overwhelming majority of people who hold those kinds of beliefs show no signs of being radicalized into actual violence. For all the talk of liberal "resistance" under Trump, the characteristic left-wing response to the Trump administration was not to join Antifa but to mobilize to elect Democrats; it took the weird conditions of the pandemic and the lockdowns, and the spark of the George Floyd killing, to transmute anti-Trumpism into national protests that actually turned violent.

Likewise, despite fears that Jan. 6 was going to birth a "Hezbollah wing" of the Republican Party, there has been no major far-right follow-up to the event, no dramatic surge in Proud Boys or Oath Keepers visibility, no campaign of anti-Biden terrorism. Instead, Republicans who believe in the stolen-election thesis seem mostly excited by the prospect of thumping Democrats in the midterms, and the truest believers are doing the extremely characteristic American thing of running for local office.

This has prompted a different liberal fear — that these new officeholders could help precipitate a constitutional crisis by refusing to do their duty in a close election in 2024. But that fear is an example of the other problem of exaggeration in the imminent-civil-war literature, the way the goal posts seem to shift when you question the evocations of Fort Sumter or 1930s Europe.

Thus we are told that some kind of major democratic breakdown is likely "absent some radical development" (as Beauchamp puts it); that we are already "suspended between democracy and autocracy" (as Remnick writes); that "the United States is coming to an end" and the only question "is how," to quote the beginning of Stephen Marche's new book, "The Next Civil War." But then it turns out that the most obvious danger is an extremely contingent one, involving a cascade of events in 2024 — a very specific sort of election outcome, followed by a series of very high-risk, unusual radical choices by state legislators and Republican senators and the Supreme Court — that are worth worrying about but not at all the likeliest scenario, let alone one that's somehow structurally inevitable.

Similarly, we are first told that "civil war" is coming, but then it turns out that the term is being used to mean something other than an actual war, that the relevant analogies are periods of political violence like the Irish Troubles or Italy's "Years of Lead." And then if you question whether we're destined to reach even that point, you may be informed that actually the civil war is practically here already — because, Marche writes, "the definition of civil strife starts at twenty-five deaths within a year," and acts of anti-government violence killed more people than that annually in the later 2010s.

That kind of claim strikes me as a ridiculous abuse of language. The United States is a vast empire of more than 330 million people in which at any given time some handful of unhinged people will be committing deadly crimes. And we are also a country with a long history of sporadic armed conflict — mob violence, labor violence, terrorism and riots — interwoven with the normal operation of our politics. If your definition of civil war implies that we are always just a few mass shootings or violent protests away from the brink, then you don't have a definition at all: You just have a license for perpetual alarmism.

I am very aware that I'm always the columnist making some version of this calm-down argument, sometimes to a fault. So I want to stress that the problems that undergird the civil-war hypothesis are serious problems, the divisions in our country are considerable and dangerous, the specific perils associated with a Trump resurgence in 2024 entirely real.

But there are also lots of countervailing and complicating forces, and the overall picture is genuinely complex — at least as complex, let's say, as the informant-riddled plot against Whitmer. And as with that conspiracy, it's worth asking whether the people who see potential insurrection lurking everywhere are seeing a danger rising entirely on its own — or in their alarm are helping to invent it.

Ross Douthat joined the New York Times as an Opinion columnist in 2009. Previously, he was a senior editor at the Atlantic and a blogger on its website. He is the author of "The Deep Places: A Memoir of Illness and Discovery," which was published in October 2021.