Contaminated record · genuine or implanted?
Continued Influence Effect
Misinformation continues to influence reasoning and beliefs even after it has been clearly retracted or corrected, because the original false information leaves a persistent trace in memory that is not fully overwritten by the correction.
Exhibit · the account as remembered
The continued influence effect was named and experimentally demonstrated by Holly M. Johnson and Colleen M. Seifert in 1994 using a warehouse-fire paradigm: after participants learned that paint and gas canisters had caused a fire and then were told this was false, they still used the discredited cause when making inferences. The effect has proven remarkably durable across hundreds of studies; a 2012 review by Lewandowsky, Ecker, Seifert, Schwarz and Cook in Psychological Science in the Public Interest found that even explicit warnings and provision of alternative accounts reduce but typically fail to eliminate the effect entirely. Real-world examples include persistent belief in Iraqi WMDs a decade after their non-existence was confirmed. ← reconstructed, not recorded
How the contamination enters
Two major accounts compete to explain the continued influence effect. The 'mental model' account (Johnson & Seifert, 1994) proposes that people construct event models on the fly and prefer an inconsistent but complete model over a correct but incomplete one; a retraction that leaves a causal gap is therefore resisted unless an alternative cause fills the void. The 'selective retrieval' or dual-process account (Ayers & Reder, 1998; Ecker et al., 2011) suggests that misinformation and corrections compete for automatic activation, but the correction requires controlled strategic retrieval that can fail under cognitive load or time pressure. A computational sampling model has been proposed in which each memory representation of misinformation can be offset by one retraction token, explaining why repeated retractions help only when multiple memory traces exist.
Competing account — The mental-model (gap-filling) account and the selective-retrieval (dual-process) account are the two leading explanations; both are supported by evidence, and they may operate in parallel.
Seams the examination found
Participants continued to rely on the retracted misinformation for inferences even though they remembered the retraction. Providing an alternative cause (arson) reduced but did not always eliminate the effect.
Johnson & Seifert · 1994 · n = 96 undergraduates across experiments
A specific warning reduced reliance on misinformation as effectively as providing an alternative cause, but even combining both manipulations failed to eliminate the CIE entirely; more than one-third of participants still referenced corrected misinformation.
Ecker, Lewandowsky & Tang · 2010 · n = 152 undergraduates
Stronger retractions reduced the CIE only when misinformation was strongly encoded. Even strong retractions failed to eliminate continued influence associated with weakly encoded misinformation, supporting an 'irreducible persistence' account.
Ecker, Lewandowsky, Swire & Chang · 2011 · n = 128 (Exp 1), n = 130 (Exp 2)
Where provenance breaks
Watch for situations where a person continues to cite or act on information they have already acknowledged was false or retracted, such as referencing a discredited study or debunked news story in reasoning about current events.
Authenticity — robust. The effect has been replicated dozens of times across multiple paradigms (warehouse fire, minibus crash, terrorism scenarios) and labs worldwide. A 2020 meta-analysis by Walter and Tukachinsky in Communication Research confirmed a significant overall effect size. However, the effect is moderated by factors such as the presence of alternative causal accounts, warning specificity, and (in political contexts) whether the misinformation aligns with pre-existing attitudes. The backfire effect (corrections increasing misbelief) has been observed in politically charged settings but is not a universal feature of the CIE.
Records later found altered
politics · 2004-2014
Persistent belief in Iraqi WMDs after the 2003 invasion
Despite the Duelfer Report (September 2004) confirming Iraq had no WMDs, polling data reviewed by Jacobson (2010) showed that around 60% of Republicans and 20% of Democrats continued to believe Iraq possessed WMDs through 2009, with little decline over time. A December 2014 poll still found 51% of Republicans and 32% of Democrats holding this false belief.
medicine · 2010-present
MMR vaccine and autism misinformation persistence
The false claim that childhood MMR vaccination causes autism, originating in a 1998 Lancet paper that was retracted in 2010, continued to influence parental immunization decisions and vaccine-preventable disease rates for years after retraction. Lewandowsky et al. (2012) cited this as a paradigmatic real-world instance of the continued influence effect.
To authenticate the memory
Provide a specific warning about the CIE up front, combined with a clear alternative causal account that fills the explanatory gap left by the retraction. Repeat the correction when misinformation was strongly encoded. Design corrections that are visually and structurally distinct from the misinformation to minimize recursive reminding.
Ecker et al. (2010) found that specific warnings were as effective as providing an alternative cause; Ecker et al. (2011) showed that stronger retractions help when misinformation is strongly encoded. The Lewandowsky et al. (2012) review synthesized evidence showing that simple retractions are insufficient.
Easily confused with
- Misinformation effect (Loftus)siblingBoth involve false information affecting memory, but the Loftus misinformation effect focuses on altering memories of witnessed events, whereas the CIE concerns retracted misinformation continuing to influence reasoning despite known invalidity.
- Belief perseverancesiblingBelief perseverance is the broader tendency to maintain beliefs after evidence for them has been discredited; the CIE is a specific mechanism involving continued reliance on corrected misinformation in inferential reasoning.
- Backfire effectmechanistically-linkedUnder some conditions (particularly politically charged topics), corrections can paradoxically strengthen misbelief rather than reduce it, though this is not a universal feature of the CIE.