011 min
Where the backfire effect shows up
A trust and safety lead at a social platform has to decide what to show under a viral post that contains a false health claim. One colleague argues that a correction label will make believers hold the claim more strongly, so the platform should add nothing. The colleague's argument is the backfire effect, used as a rule for a product decision.
The same fear appears wherever people correct others. A public health team wonders whether a myth-busting leaflet will do harm. A support team wonders whether telling a customer they are wrong about a feature will end the conversation badly. A manager wonders whether showing a team the data against its estimate will make the team defend that estimate harder. In each case the person doing the correcting expects the correction to move the other person in the wrong direction.
That expectation has a source in published research, and the research has been tested again since. This article follows the claim from its first experiment to its current standing, because the answer changes what the lead in the scenario should do.
022 min
Where the backfire effect comes from
Brendan Nyhan and Jason Reifler named the backfire effect in a 2010 paper in the journal Political Behavior, titled "When Corrections Fail: The Persistence of Political Misperceptions". They tested whether a correction placed inside a news story reduces a false belief. In the version of the paper posted on Nyhan's Dartmouth site, the authors report that corrections frequently fail to reduce misperceptions among the targeted ideological group, and that some corrections actually increase them.
The method was a survey experiment. Subjects read a mock news article in which a politician made a misleading claim. Some subjects read a version of the article that also contained a correction. After reading, subjects rated how much they agreed with the false claim. The authors split the results by political ideology, because they expected the group whose worldview favoured the false claim to resist the correction most. The participants were undergraduates at a Catholic university in the Midwest, and they took part in an online survey.
The first experiment ran in fall 2005 and was about the war in Iraq. In the correction version, a statement that could be read as suggesting Iraq had weapons of mass destruction was followed by a clarification that such weapons had not been found. The result for conservatives was the one that gave the effect its name. Among 33 conservative participants, agreement that Iraq had weapons of mass destruction before the US invasion rose from 32% in the control condition to 64% in the correction condition. Among the 97 non-conservatives, agreement fell from 22% to 13%. The correction lowered belief in the group whose worldview did not favour the claim and raised it in the group whose worldview did.
A second experiment used a different issue. Conservatives who were shown evidence that tax cuts do not increase government revenues ended up believing that claim more fervently than people who did not receive a correction. A later paper by other researchers summarised the original results this way. Wood and Porter describe the 2010 results as backfire on two issues that were highly salient, namely whether weapons of mass destruction were found in Iraq and whether tax cuts ultimately increase tax receipts, and no backfire on one less salient and more technical issue about stem cell research.
The reported effects were large, but the samples were small. The Iraq result for conservatives rests on 33 people. Later sections explain what larger samples found. The author of the original paper also counted how often the effect appeared. Nyhan later wrote that the effect appeared in two of the five studies and was not observed in the other three.
031 min
How the backfire effect spread
The backfire effect began as a result about two issues, in one sample, in one kind of experiment. It became a general claim about people. Nyhan wrote that later research and media coverage seized on the finding, distorting how general it was and exaggerating its role relative to other factors in explaining the durability of political misperceptions. The popular version said that facts do not work on people who disagree with them, and that showing facts can make things worse. The experiments never tested that general statement.
The effect also spread as a name. Reviews now separate two versions. The worldview version is the one Nyhan and Reifler described. The familiarity version is a later proposal that a correction fails because it repeats the false claim. The next section explains both. What changed most between 2010 and today is the weight of evidence behind them, which later sections cover.
042 min
Why the backfire effect was proposed
Two mechanisms have been proposed for the backfire effect. They predict different things, so they need to be kept apart. Swire-Thompson, DeGutis and Lazer, in their 2020 review, separate two versions of the effect: in the familiarity version, the correction itself repeats the false claim, and that repetition is presumed to be what raises belief.
The worldview version
This version comes from research on motivated reasoning, the tendency for a person's values and group identity to shape how they judge new information. A person who holds a belief tied to their identity may treat a correction as an attack on that identity. Swire-Thompson, DeGutis and Lazer write that a possible reason for the backfire effect is that people generate counter-arguments consistent with their pre-existing views to contradict the new information or correction. If the counter-arguments are strong enough, the person ends with more reasons to believe the claim than before the correction.
This is related to cognitive dissonance, the discomfort of holding two conflicting ideas. A correction creates that conflict, and one way to reduce it is to reject the correction.
The familiarity version
This version comes from memory research. The illusory truth effect is the finding that people rate a statement as more likely to be true after they have seen it before. A correction such as "eating apricot seeds does not cure cancer" repeats "apricot seeds" and "cure cancer". The proposal is that the repeated words make the false claim feel familiar, and familiar claims feel true.
This version is different from the first in one important way. It does not need the person to hold a strong worldview about the claim. It predicts a backfire on any false claim, including a neutral one about health or science. Brashier, Eliseev and Marsh (2020), as the review describes their work, found that the illusory truth effect was wiped out when participants were simply asked to rate statements for accuracy. That finding matters for the familiarity version because it suggests that the repetition effect can be switched off by how a person reads the correction.
052 min
Where the backfire effect evidence is contested
The effect has been tested again, with much larger samples. The result is the main reason the effect is now called contested. Wood and Porter ran five experiments with more than 10,100 subjects and tested 52 issues of potential backfire. Across all experiments they found no corrections capable of triggering backfire, even on the polarized issues where backfire should be expected. The result was also similar across the political spectrum. Among liberals, 85% of issues saw a significant factual response to correction, among moderates 96% of issues, and among conservatives 83% of issues. No backfire was found for any issue in any ideological group.
The authors did not claim the effect is impossible. Wood and Porter wrote that their results do not lead them to conclude that backfire is categorically impossible. They argue instead that backfire is much less common than earlier research suggested.
Part of the original result came from the question
One of their experiments reused the Iraq article and correction from 2010. They randomly gave subjects either the original survey item or a shorter one that only asked whether US forces found weapons of mass destruction after the invasion. In one of their experiments, the question wording strongly conditioned the level of factual backfire on the Iraq item. Wood and Porter point out that the original item presents multiple ways to account for the failure to find weapons of mass destruction. The authors suggest that respondents may have read the long item as an invitation to judge the war in general, not as a question about what was found. Question wording is a known source of different answers, as research on the framing effect shows, so a backfire that appears only with one wording is weak evidence about beliefs.
Measurement problems
A 2020 review looked at how the studies measured belief. They also found that 81% of the backfire effects in their review of the worldview and familiarity literatures came from single item measures. Single items are noisy, and the review notes that random error can produce results that look like backfire. The review's overall view is cautious. Swire-Thompson, DeGutis and Lazer suggest that backfire effects are not a robust empirical phenomenon.
Where this leaves the field
Nyhan, the first author of the original paper, now agrees that the effect is rare. Nyhan writes that an emerging research consensus finds that corrective information is typically at least somewhat effective at increasing belief accuracy when received by respondents. The original authors and the critics now agree on the main point: a backfire is not the general rule.
062 min
What explains false beliefs that last
If corrections rarely backfire, why do false beliefs last? Nyhan's review gives a different answer, and it is the one a practitioner can act on. Nyhan writes that the accuracy-increasing effects of corrective information like fact-checks often do not last or accumulate, and frequently appear to decay or be overwhelmed by cues from elites and the media promoting more congenial but less accurate claims.
A second real case: the birth certificate
The rumour that Barack Obama was not born in the United States gives a clear example. In April 2011, Barack Obama released his long form birth certificate from the state of Hawaii. Belief that Obama was born in the U.S. increased immediately after the release but reverted to prior levels by 2012. Belief rose, so the correction worked. It then fell back, so the correction did not last. Neither result is a backfire.
The people who need the correction do not see it
A fact-check only helps if it reaches the people who saw the false claim. In the weeks before the 2016 US election, only 3% of those who read an article from an untrustworthy website that had been fact-checked also read the fact-check. This is a targeting problem. The correction exists, but the audience for it and the audience for the false claim are mostly different people.
The two cases have one variable that differs. The original experiments forced every subject to read the correction. Real life does not. The experiments therefore measure what a correction does when it is read, and the birth certificate case shows what happens afterwards, when a correction competes with ongoing messages from other sources.
072 min
How to correct claims after the backfire effect
Back to the trust and safety lead. The research gives a clear answer for the decision. The same review states that fact-checkers can rest assured that it is extremely unlikely that their fact-checks will lead to increased belief at the group level. The fear that a label will make believers worse is not a good reason to add nothing. Swire-Thompson, DeGutis and Lazer conclude that practitioners should not avoid giving corrections to any specific subgroup of people.
The research also changes what the lead should work on. The risk to manage is that the label is not seen or does not last, not that it backfires. These steps follow from the evidence above:
Show the correction to the people who saw the claim.
Place it where the false post appears, not in a separate page that few people open. This addresses the targeting problem.
Make the correction clear first.
They also write that misinformation should always be clearly and saliently paired with the corrective element, and that needless repetitions of the misconceptions should still be avoided. State the false claim once, mark it plainly as false, and put the accurate statement next to it.
Plan for decay.
Repeat the correction when the false claim returns. One correction often fades, as the decay findings above show.
Measure with more than one question.
If you test a correction in your own product, use several questions about the same belief, because single items are noisy.
Test before you assume.
If a backfire matters for your audience, run the test on that audience. The group-level result from large samples is not the same as a guarantee for one subgroup, one issue or one wording.
081 min
Common misunderstandings about the backfire effect
"Corrections make people believe the false claim more." This is the popular version, and the large tests do not support it. The correct statement is narrower: a backfire was reported under specific conditions and has rarely been reproduced.
"A correction that works has changed the person's mind." This is also wrong, in the other direction. A correction can raise the accuracy of a factual belief without changing the opinion built on it. Nyhan writes that people's opinions about an issue sometimes do not change even if their factual beliefs become more accurate, though findings vary on this point. A fact-checker who measures only belief accuracy can miss that the opinion stayed put.
"If there is no backfire, corrections are enough." The decay and targeting findings above show that a single correction often is not. The research shifted the problem; it did not remove it.
091 min
The backfire effect vs. nearby concepts
The backfire effect is often mixed up with the biases that may cause it. The axis that separates it from each of them is the direction of the change in belief.
| Concept | What it describes | How it differs from the backfire effect |
|---|---|---|
| PsychologyConfirmation bias | ||
| PsychologyCognitive dissonance | ||
| PsychologyReactance | ||
| PsychologyIllusory truth effect | ||
| Not in the library yetBelief decay |
?7 questions
Questions people ask
What is the backfire effect in psychology?
Is the backfire effect real?
What causes the backfire effect?
Do corrections and fact-checks work?
Should you avoid correcting people who strongly disagree?
Does repeating a myth in the correction make it stronger?
How is the backfire effect different from confirmation bias?
§4 sources
Sources
Show all 4 sourcesHide sources
Nyhan, B. and Reifler, J. (2010). When Corrections Fail: The Persistence of Political Misperceptions. Political Behavior, 32(2), 303-330. .
Wood, T. and Porter, E. The Elusive Backfire Effect: Mass Attitudes' Steadfast Factual Adherence. Political Behavior. .
Swire-Thompson, B., DeGutis, J. and Lazer, D. (2020). Searching for the Backfire Effect: Measurement and Design Considerations. Journal of Applied Research in Memory and Cognition, 9, 286-299.
Nyhan, B. (2021). Why "backfire effects" do not explain the durability of political misperceptions. Proceedings of the National Academy of Sciences.





