This effect appears twice in this guide as well — chapter 25 covers it as a mechanism of belief, and it recurs here because it is the operating system of propaganda rather than one technique among many.
Hasher, Goldstein and Toppino's 1977 design was straightforward. Sixty plausible statements, some true and some false, rated for truth across three sessions two weeks apart. A third of the statements repeated. Repeated statements were rated as more true, regardless of their actual truth value.
What has been established since is what matters for this part.
Lisa Fazio and colleagues showed in 2015 that prior knowledge does not protect. Participants who could correctly answer a factual question when asked directly nevertheless rated a repeated contradictory statement as more true than an unrepeated one. Knowledge is available and fluency is immediate, and at the moment of judgment the immediate cue frequently wins.
A single repetition is enough to produce a measurable effect. The effect grows with further repetitions, with diminishing returns, and it persists across delays of weeks. It works on headlines, on advertising claims, on political statements, and it has been demonstrated for statements that participants rated as implausible on first exposure.
The mechanism is source monitoring failure, in Marcia Johnson's terms. The mind registers that a statement feels familiar and easy to process. It then has to attribute that ease to something, and the correct attribution — I have encountered this before — is less accessible than the convenient one — this is consistent with things I know to be true. Fluency gets read as prior verification.
Three practical consequences follow, and the third is uncomfortable.
First, for anyone producing propaganda, this is why you do not need arguments. A claim does not have to be defended; it has to be present, in many places, repeatedly. This is the entire logic of the talking point — a phrase distributed to dozens of speakers to be used verbatim, whose function is not to persuade any single listener but to accumulate exposures across a population.
Second, for anyone correcting misinformation, repetition of the false claim in the course of debunking it adds exposures. The literature on this is nuanced — outright backfire effects have proved hard to replicate and are less common than once thought — but familiarity accrual over time is real. The recommended structure is the truth sandwich: lead with the true statement, state the false claim once and clearly labeled, explain the specific technique or error that makes it wrong, and close on the truth.
Third, and personally: the sensation that a claim is obviously true, arrived at without recalling any evidence for it, is not self-validating. It is a report from the fluency system, and the fluency system is measuring exposure count. Somebody may have set that number.
The best available defense is prebunking, in chapter 328, precisely because it puts the fluency advantage on the correct side before the false claim arrives.
Cross-ref: Big Lie chapter following