The Backfire Effect Is Mostly a Myth

You have probably heard that facts do not change minds, and that showing someone evidence against a cherished belief only makes them cling to it harder. The claim has a name, the backfire effect, and for much of the 2010s it was repeated in newsrooms, classrooms, and countless articles about polarization. It feels true to anyone who has argued with a relative over dinner. It also had a convenient implication: if correcting people is worse than useless, nobody is obliged to try.

The best current evidence says that the backfire effect, in the sense of a correction increasing belief in the falsehood it corrects, is rare, fragile, and not a sound basis for deciding whether to correct misinformation. What follows is the story of how the idea took hold, how it was tested, and what the research recommends in its place. It is also a good example of science correcting itself, in which the original authors took part.

In 2010 the political scientists Brendan Nyhan and Jason Reifler published "When Corrections Fail." Participants read mock news articles in which a politician made a misleading claim; some versions included a correction. The corrections often failed to reduce misperceptions among the people ideologically inclined to believe the claim. In two cases they appeared to do worse. Conservative participants who read a correction stating that Iraq had not possessed weapons of mass destruction at the time of the 2003 invasion became more likely to say that it had, and a similar pattern appeared for the claim that the Bush tax cuts had increased government revenue. The authors also tested a misperception more common among liberals, that President Bush had banned all stem cell research. There the correction failed to reduce the belief among liberals but did not backfire.

The paper was careful. Its samples were undergraduates, the backfire appeared on some items and not others, and the authors presented it as a finding to be explored. The popular retelling dropped those qualifications, and "corrections backfire" became a supposed law of human nature.

Thomas Wood and Ethan Porter set out to map when backfire occurs. Across five experiments they enrolled more than 10,000 participants and tested corrections on 52 politically charged issues, using misleading statements from politicians of both major American parties. They expected to find backfire on the most divisive topics. They found it nowhere. Across issues and across ideological groups, people who saw a correction moved toward the accurate answer, including when the correction cut against their own side. The one hint of backfire came on the original Iraq item, and it vanished when the question was worded more simply. They titled the paper "The Elusive Backfire Effect" and concluded that citizens by and large heed factual information.

Other teams reached similar conclusions. In 2020 Briony Swire-Thompson, Joseph DeGutis, and David Lazer reviewed the literature on the two proposed varieties: the worldview backfire effect, in which a correction threatens identity, and the familiarity backfire effect, in which repeating a myth in order to debunk it makes it more familiar and therefore more believable. They found that backfire results had mostly come from studies with less reliable measures, such as a single question, where random noise and regression to the mean can pass for a real reversal. Studies using more reliable measures seldom found it. Their conclusion was that backfire is not robust and that fear of it should not deter fact-checkers.

The original authors agreed. Nyhan and Reifler joined Porter and Wood for a study of corrections to statements made by Donald Trump during the 2016 campaign, and found that corrections improved factual accuracy even among his supporters. In a 2021 article in the Proceedings of the National Academy of Sciences, Nyhan wrote that the backfire effect does not explain why political misperceptions are so durable. Researchers who publicly revise an influential finding of their own deserve credit for it.

"Backfire is rare" is not the same as "corrections solve the problem." Three limits are well established.

Beliefs move; attitudes often stay put. In the Trump study, supporters who accepted the correction did not rate the candidate any less favourably. The authors summarized this as taking fact-checks literally but not seriously. The pattern is not peculiar to one party. People of every persuasion can concede a point of fact and keep their overall judgment, because that judgment rests on values and on identity more than on any single claim.

Misinformation lingers after it has been retracted. In a classic experiment by Hollyn Johnson and Colleen Seifert, participants read a report of a warehouse fire that at first attributed it to carelessly stored paint and gas cylinders, and then retracted that detail. Many went on using the retracted cause to answer later questions, even though they remembered the retraction. This continued influence effect is much weaker when the correction supplies an alternative explanation that fills the gap. "That is false" leaves a hole in the story. "That is false, and here is what actually happened" does not.

The effect of a correction fades. A single correction is up against a false claim that may be encountered again and again, and repetition makes claims feel true. Corrections need repeating as well.

The Debunking Handbook 2020, written by Stephan Lewandowsky, John Cook, Ullrich Ecker, and some twenty other researchers, condenses the evidence into a simple structure:

• Lead with the fact. State what is true, clearly and concretely, so that the accurate version is what sticks.
• Warn before the myth. Say once that a misleading claim is coming, and then state it once. The familiarity risk from a single mention is small, but there is no reason to make the myth your headline.
• Explain the fallacy. Show why the claim is wrong and, where possible, how it misleads: a cherry-picked date range, a fake expert, a false comparison. That helps people recognize the same technique next time.
• End on the fact. Restate the accurate account, including the alternative explanation that fills the gap.

The handbook states directly that backfire effects are uncommon and should not stop anyone from debunking. It also advises choosing your battles. A myth that almost nobody has seen may be better left alone, since correcting it lends it reach. And it recommends prebunking where possible, which means warning people about manipulation techniques before they meet them.

Some further points apply to conversations between individuals:

• Correct the claim, not the person. People update more readily when doing so does not require admitting that they are foolish or that their group is bad.
• Use sources the listener trusts. A correction from someone seen as speaking against their own side's interest carries extra weight.
• Remember the audience. In a public thread, the person you are replying to may never concede, but many silent readers will see the accurate information.
• Accept partial movement. Moving from "certain" to "not sure" is a real change, and often the most a single conversation can achieve.

The backfire story spread for the very reasons this site documents elsewhere. It was vivid and counterintuitive, and it flattered the teller, since it is always other people who are immune to facts. It also excused inaction. There is an irony in the fact that a claim about the futility of correction was itself corrected, and that people who follow the research updated their view accordingly.

It is worth noticing who gains from the belief that facts are useless. High-volume propaganda strategies such as the firehose of falsehood rely on defenders giving up. The evidence says they should not. People are more responsive to accurate information than the pessimists assumed, across parties and across issues. What they are reluctant to do is change their loyalties because of a single fact, which is a different matter and, arguably, not unreasonable.

For practical scripts, see Defenses, and for the companion piece on how false claims are spread in the first place, read The Disinformation Playbook.