Correcting false claims
The experiments on whether correcting a false claim changes what people believe, and what a correction needs before it works.
hen someone believes something that is not true, what happens if you correct them? Does the correction stick, and does saying the false claim out loud to knock it down make it stronger?
Nyhan & Reifler (2010)
The four experiments that first reported a backfire effect, in which mock news articles gave people a politician's false claim, with or without a correction.
Unsupported block: pub.terrain.quote
Wood & Porter (2018)
Five experiments testing 52 different claims for a backfire effect, with more than 10,100 people taking part.
Unsupported block: pub.terrain.quote
Unsupported block: pub.terrain.quote
Unsupported block: pub.terrain.quote
Ecker, Lewandowsky & Tang (2010)
An experiment that counts how often people still bring up a retracted claim, under different kinds of correction.
Unsupported block: pub.terrain.quote
On Ecker's scale of references to the retracted claim: a bare retraction scored 4.04, no retraction at all scored 5.06, and a retraction with a plausible alternative scored 2.22.
Chan, Jones, Hall Jamieson & Albarracín (2017)
A meta-analysis of the studies on debunking messages, on what makes a correction work.
Unsupported block: pub.terrain.quote
Walter & Tukachinsky (2019)
A meta-analysis of 32 studies on what is left of a false claim after it has been corrected.
Unsupported block: pub.terrain.quote
Nyhan (2020)
A review by one of the authors of the original backfire finding, ten years later.
Unsupported block: pub.terrain.quote
Unsupported block: pub.terrain.quote
Unsupported block: pub.terrain.quote
Synthesis
The claims tested were ones people had just read, in a survey, on issues the researchers chose.
The correction in these studies is one message, delivered once, usually right after the claim.
Even the best correction left some people still referring to the claim they had been told was false.
Nothing here measures what happens when the false claim keeps being repeated after the correction.
It is not known how much of this carries into a campaign, where a correction is one voice among many and arrives months after the claim.
The comparison that would settle the practical question, a correction with a replacement account against no correction at all, measured on votes rather than beliefs, does not appear to have been run.
Linked notes
/notes/far-right-salience
/notes/persuasion
/notes/what-counts-as-evidence
Did you enjoy this article?
Recommend it — Standard Reader surfaces well-loved writing to more readers across the network.