The continued influence effect (CIE) is the persistence of misinformation’s influence on reasoning after the reader has received a correction. In controlled research, people can remember that an earlier claim was retracted yet still use it to explain a later event. The classic 1994 experiments by Johnson and Seifert helped demonstrate this distinction between knowing about a correction and actually replacing an invalidated causal explanation.
Did you know that a simple ‘that report was wrong’ may leave a reader uncertain about what actually caused an event? When the original claim supplied a coherent explanation, its removal can leave a gap. A correction is often more useful when it clearly states the accurate alternative, if the evidence supports one. Vocabulary becomes a technology for reliable reasoning when reported, retracted and confirmed remain distinct.
This guide explains the original 1994 experiments, mental-model and memory-retrieval accounts, the 2012 and 2022 scientific reviews and evidence from correction meta-analyses. It distinguishes the continued influence effect from a true backfire effect and develops original examples for Primary and Secondary English, Science, Mathematics, History, General Paper, translation and workplace notices.
Editorial note: The library-closure examples below are constructed illustrations, not real incidents or new experiments. This article makes no claim that all corrections fail or that misinformation should be left uncorrected. Factual accuracy requires current credible sources.
Contents: why invalidated information can survive in an explanation
Open 54 chapters and correction checks
- 1. A correction can arrive without fully replacing a belief
- 2. Content memory and current validity can separate
- 3. Johnson and Seifert’s 1994 experiment
- 4. An evolving event report creates a mental model
- 5. The warehouse-fire story is a research example
- 6. Knowing the retraction is not always enough
- 7. The initial explanation can supply coherence
- 8. A missing cause is a coherence gap
- 9. A verified alternative can fill that gap
- 10. Experiment 3 tested replacement explanations
- 11. Immediate correction does not guarantee elimination
- 12. Incidentally mentioned details behave differently
- 13. Misinformation is not always malicious
- 14. Retraction and correction are different actions
- 15. Source-monitoring errors may contribute
- 16. Selective retrieval is another proposed mechanism
- 17. Source confusion and retrieval failure can coexist
- 18. The 2012 comprehensive misinformation review
- 19. The review distinguishes formation from persistence
- 20. Pre-exposure warnings can have a role
- 21. Corrective repetition can help in some designs
- 22. Emphasise the accurate account rather than the myth
- 23. The 2022 Nature Reviews Psychology synthesis
- 24. Corrections can be genuinely effective
- 25. The backfire effect is not a universal rule
- 26. The 2017 meta-analysis evaluated debunking
- 27. The 2023 science-misinformation meta-analysis
- 28. Belief accuracy and behaviour are different outcomes
- 29. A coherent false cause can influence a summary
- 30. An unsupported alternative would make matters worse
- 31. Causal networks clarify what changed
- 32. Situation-model updating is related but distinct
- 33. KReC focuses on knowledge revision components
- 34. An explicitly quoted false claim needs attribution
- 35. A current official source can supersede an old report
- 36. Source attribution alone may not repair the causal model
- 37. A strong correction should state the true proposition
- 38. Explain why the original claim appeared
- 39. Keep the correction tied to the source
- 40. Repetition of the false claim has a cost
- 41. A delayed transfer question is informative
- 42. A matched control condition matters in research
- 43. The effect can arise under cognitive load
- 44. Media literacy should separate stages
- 45. Primary English can use a fictional correction
- 46. Secondary English can track a changing claim
- 47. Science requires explanatory replacement, not a slogan
- 48. Mathematics requires correcting the model
- 49. History and GP need a correction ledger
- 50. Translation must preserve the retraction status
- 51. Workplace documents need current instructions
- 52. A five-column retraction audit
- 53. An eight-step correction reading routine
- 54. The final lesson is correction that reaches reasoning
Quick Test: a correction changes which causes are supported
| Stage | Fictional library incident | Correct reasoning |
|---|---|---|
| Early unconfirmed report | The room reportedly closed because a window was damaged | Treat the cause as provisional |
| Retraction | The windows were found undamaged | Remove the damaged-window cause |
| No replacement explanation yet | The reason for closure remains under investigation | Do not invent an alternative cause |
| Later verified update | A documented electrical inspection required closure | Use the supported inspection explanation |
| New comprehension question | Why was the room closed? | Answer using the latest supported account |
This is an invented example rather than a finding about an actual library. The key challenge is revising the explanation, not simply remembering that a retraction was printed.
1. A correction can arrive without fully replacing a belief
The continued influence effect, or CIE, occurs when information that has been shown to be false continues to influence a person’s inferences or judgements after they have encountered a correction. The person may even remember that the original statement was retracted while still drawing conclusions that rely on it.
This is a specific experimental phenomenon, not proof that corrections never work. A correction can substantially improve accuracy while leaving some residual influence. The educational goal is to help readers connect the current evidence with the explanation they use when answering new questions.
2. Content memory and current validity can separate
A false claim may have become part of a coherent story. Later, a retraction changes its status, but the earlier narrative connections can remain accessible in memory and affect how the person explains an event.
This shows why vocabulary about status matters. Reported, alleged, corrected and confirmed are not interchangeable. A reader can accurately recall what was once reported yet still need to identify which proposition remains supported now.
3. Johnson and Seifert’s 1994 experiment
Hollyn Johnson and Colleen Seifert published Sources of the Continued Influence Effect: When Misinformation in Memory Affects Later Inferences in 1994. Their experiments asked why discredited information can continue to shape reasoning about an event.
They compared immediate and delayed corrections, examined whether misinformation contributed a causal explanation and tested replacement accounts. The research found that an alternative explanation could reduce a continued-influence effect under the tested conditions. It did not conclude that every simple retraction is always useless.
4. An evolving event report creates a mental model
In a classic paradigm, participants read a series of messages about an unfolding incident. Early messages suggest a cause; a later update retracts that information. The person then answers questions about what happened.
The method is designed to determine whether an invalidated explanation still affects reasoning. It is not a test of whether participants can find a misspelling in isolation. The critical outcome is what they infer about the event after receiving the correction.
5. The warehouse-fire story is a research example
A commonly used research narrative describes a warehouse fire, with an early report suggesting that flammable materials were present and helped cause it. Later messages retract that account and clarify that those materials were not there.
Participants may still refer to the initial materials when explaining the incident. The story is a fictitious experimental scenario rather than a report of an actual warehouse fire. The distinction matters because the paper studies reasoning processes, not the factual cause of a real disaster.
6. Knowing the retraction is not always enough
Some research found continued use of the earlier story even among readers who could remember the correction. This creates a crucial distinction between recognising that information was withdrawn and avoiding its use in a new inference.
A student may answer ‘Yes, that explanation was retracted’ yet later build a summary around the discredited cause. The repair must reach the structure of the explanation, not only a checkbox about which sentence was corrected.
7. The initial explanation can supply coherence
A causal explanation makes a sequence of events seem intelligible. If the later correction removes the supposed cause, the remaining account may feel incomplete unless an accurate alternative explains what happened.
Mental-model explanations of the CIE propose that readers can return to the old causal structure because it still connects events better than a bare negation. This is a theoretical account with experimental support, not a universal rule that people intentionally prefer falsehoods.
8. A missing cause is a coherence gap
Imagine an invented report that a library room closed because its windows were damaged. A later official update states that the windows were undamaged but provides no explanation for the closure.
The reader now knows one proposed cause is false but may still struggle to explain why the room closed. That gap is exactly the sort of situation in which an old explanation can remain tempting. The correct response is to mark the cause unknown, not silently reinstate the invalidated story.
9. A verified alternative can fill that gap
Suppose a further authoritative update says the room was temporarily closed for a planned electrical inspection, with supporting records. That account supplies a new explanation without relying on the discredited window claim.
The alternative should not be invented merely to create a satisfying narrative. A correction is stronger when it provides an accurate and supported explanation. If no verified cause is available, saying the cause remains under investigation is more honest than filling the gap with speculation.
10. Experiment 3 tested replacement explanations
Johnson and Seifert’s 1994 research reported that supplying a plausible alternative causal account mitigated one continued-influence effect. The result motivated later recommendations that corrections provide an explanatory replacement when evidence allows.
This does not mean any alternative story is better than admitting uncertainty. The replacement must be credible and connected to the actual incident. An appealing but unsupported explanation could create another misinformation problem rather than resolve the first.
11. Immediate correction does not guarantee elimination
In the 1994 experiments, correcting an initial statement immediately did not eliminate all use of the discredited content compared with correcting it later in the sequence. This challenged a simple account based only on the time available to construct inferences.
The result concerns the particular materials and outcome measures. It does not establish that timing is irrelevant in every communication environment. Early correction can still be useful, but an effective message may also need to address the reasoning supported by the original claim.
12. Incidentally mentioned details behave differently
The research found that misinformation serving an important causal role in a story could keep influencing inferences, whereas the same kind of material mentioned only incidentally was less likely to have that continuing effect under its tested conditions.
This suggests that narrative function matters. A false colour of a wall may have little explanatory importance, while an incorrect cause of a fire can organise an entire event account. The practical correction should therefore target the mistaken causal relationship, not only a peripheral word.
13. Misinformation is not always malicious
Incorrect information can result from early uncertainty, a transcription error, a misunderstood measurement or an intentionally deceptive claim. The CIE concerns subsequent influence after correction, not a requirement that the original source intended to mislead.
A good reader distinguishes false, mistaken and deliberately fabricated. Each makes a different assertion about content or intention. A journalist correcting an honest early report should not automatically be described as having committed deliberate deception.
14. Retraction and correction are different actions
A retraction indicates that particular earlier information should not be relied upon. A correction may provide a replacement claim or clarifying evidence in addition to invalidating the old one.
These processes can overlap but are not synonymous. A report can retract one proposed cause without establishing the true cause. A precise reader records the invalidated statement and the current supported status separately, avoiding the assumption that every retraction supplies a complete new explanation.
15. Source-monitoring errors may contribute
One proposed explanation for CIE involves source confusion. A reader might remember the initial claim but misattribute it to the final official report instead of the early message that was corrected.
The Source-Monitoring Framework studies how people assign memories to their origins. That can help explain some continued influence, but it is not the whole story. A person can remember the correct source yet still use an invalidated causal account in reasoning.
16. Selective retrieval is another proposed mechanism
Retrieval-based accounts suggest that both valid and invalid explanations can remain in memory and compete when a question provides relevant cues. The earlier claim may become more accessible than the correction under some conditions.
The reader might not consciously choose to ignore the correction. Nevertheless, later reasoning needs to select the current valid account. Source labels, reminders and clear replacement explanations can help, but no single technique guarantees perfect retrieval in every context.
17. Source confusion and retrieval failure can coexist
One reader may forget which report withdrew a claim. Another may remember the withdrawal yet spontaneously retrieve the old causal story because it fits a question. These mechanisms are distinct but can interact.
Educationally, the lesson needs two checks: can the pupil identify the currently valid source, and can they apply the corrected explanation? Correct source recall without accurate inference is incomplete, just as correct inference without traceable evidence may be fragile.
18. The 2012 comprehensive misinformation review
Lewandowsky, Ecker, Seifert, Schwarz and Cook published Misinformation and Its Correction: Continued Influence and Successful Debiasing in Psychological Science in the Public Interest in 2012.
Their review discussed misinformation formation, continued influence, mental-model and memory accounts, and methods of improving corrections. It drew on cognitive research and considered practical communication. The publication is a review, not one new experiment proving that every proposed debiasing technique works universally.
19. The review distinguishes formation from persistence
One question asks why a person initially accepts inaccurate information. Another asks why the claim keeps influencing reasoning after a correction. These problems have related but different mechanisms.
A correction strategy that reduces initial acceptance may not solve every later retrieval problem. Conversely, a source reminder may help a reader update an existing account without preventing all future exposure. Vocabulary such as prevention, correction and persistence describes different stages.
20. Pre-exposure warnings can have a role
Some studies have tested warnings that forthcoming information could be unreliable. A suitable warning may encourage a reader to treat a claim provisionally instead of incorporating it immediately as confirmed fact.
This is not a guarantee that every warning label prevents misinformation. The wording, source, audience and exposure context matter. For education, a sensible habit is to mark unverified early reports as provisional and check the later record before repeating them.
21. Corrective repetition can help in some designs
Repeated correction can increase familiarity with the accurate account and may reduce some misinformation effects. However, its effectiveness depends on initial exposure, the corrective wording and the outcome measure.
A teacher should not make unqualified promises that repeating a correction three times eliminates every false inference. Repetition works best when accompanied by comprehension and meaningful use. Students need to apply the correction to new questions rather than only recite a memorised line.
22. Emphasise the accurate account rather than the myth
A correction should make the currently supported facts easy to find. Overly elaborate repetition of the original false claim can sometimes keep its content salient without improving understanding.
This principle needs nuance: clearly identifying the incorrect claim may be necessary to explain what was corrected. The aim is not to hide history or avoid all mention of misinformation. It is to structure the explanation so that the accurate account receives sufficient attention and detail.
23. The 2022 Nature Reviews Psychology synthesis
Ecker, Lewandowsky, Cook and colleagues published The Psychological Drivers of Misinformation Belief and Its Resistance to Correction in 2022. Their review examined cognitive, social and affective factors and strategies for reducing misinformation influence.
The review considered both proactive interventions and reactive corrections and discussed multiple accounts of continued influence. It is useful because it situates CIE within a broader evidence base rather than attributing the phenomenon to one memory failure. Its conclusions are conditional on the reviewed studies.
24. Corrections can be genuinely effective
The existence of residual influence does not mean that fact-checking always backfires or that people never update beliefs. Evidence from reviews of factual corrections shows that they can improve accuracy across diverse settings.
Porter and Wood’s 2024 review describes instances of a correction making belief accuracy worse as exceedingly rare and potentially sensitive to research design. A responsible guide should not repeat the simplistic claim that any attempt to correct misinformation makes it more powerful.
25. The backfire effect is not a universal rule
The term backfire is sometimes used when a correction makes a false belief stronger. That possibility drew considerable attention, but subsequent research has questioned how common and robust it is.
A CIE can occur without a backfire: a correction may reduce false belief yet leave some residual effect on later reasoning. The two outcomes should not be treated as synonyms. Teachers should correct inaccurate information using evidence rather than avoid correction out of exaggerated fear.
26. The 2017 meta-analysis evaluated debunking
Chan, Jones, Jamieson and Albarracín published a meta-analysis titled Debunking: A Meta-Analysis of the Psychological Efficacy of Messages Countering Misinformation in 2017.
The analysis synthesised fifty-two studies with 6,878 participants and reported substantial effects of both debunking and misinformation persistence across its included outcomes. This demonstrates that correction and residual influence can coexist. The results should not be translated into one personal success probability for every classroom message.
27. The 2023 science-misinformation meta-analysis
Chan and Albarracín published A Meta-Analysis of Correction Effects in Science-Relevant Misinformation in Nature Human Behaviour in 2023. It analysed 245 effect sizes across seventy-five reports and a combined sample exceeding fifty thousand participants.
The authors reported a small, statistically nonsignificant average effect in the science-relevant correction evidence they examined. This is a result for that defined body of studies, not evidence that every individual science correction is ineffective. Method, audience, message and outcome remain important.
28. Belief accuracy and behaviour are different outcomes
A person may accept a correction and answer a factual question accurately while still having an attitude or behaviour shaped by earlier experiences. Conversely, a decision may change for reasons unrelated to explicit factual belief.
A research report should state which outcome it measured: endorsement, inference, recall, intention or actual action. Educational lessons should not claim to change behaviour simply because students repeat a corrected definition. Each outcome requires separate evidence.
29. A coherent false cause can influence a summary
Imagine a fictional school report saying that a library room closed because of a damaged window, then later retracting the window explanation. A pupil might remember that no damage was found but still say the closure happened due to window repairs.
That answer illustrates continued influence: the false cause is still used after its status changed. The correction should prompt a new summary that excludes the invalidated cause. If the true reason is unknown, the answer should say so.
30. An unsupported alternative would make matters worse
A teacher should not replace the damaged-window story with an invented electrical fault merely because the narrative feels incomplete. That would solve the psychological coherence problem by adding another unverified assertion.
An accurate alternative requires evidence. Where the source has not yet established the reason, explain that the earlier proposed cause was withdrawn and the cause remains under investigation. Honesty about a gap is better than a complete but fictional explanation.
31. Causal networks clarify what changed
A reader can draw an event network containing the closure, alleged damaged window and supposed repair. The retraction invalidates the arrow connecting the alleged damage to the closure.
The student should remove or label that unsupported link and retain the confirmed events. If an alternative supported cause becomes available, it can be added with its source. The aim is to revise the explanation, not erase the entire history of the incident.
32. Situation-model updating is related but distinct
Situation-model updating describes how readers revise an integrated representation when the described state changes. A CIE involves a particular difficulty: invalidated information can still influence interpretation or inference.
A person can update a character’s location correctly without confronting any misinformation. Conversely, a corrected causal explanation can remain influential despite explicit retraction. The theories overlap but address different research questions and should be described separately.
33. KReC focuses on knowledge revision components
The Knowledge Revision Components Framework proposes that coactivation, integration and competition between old and accurate representations contribute to correcting misconceptions.
This relates to the CIE but is not the same phenomenon. KReC emphasises mechanisms of conceptual revision in learning, while CIE research examines persistent influence after a correction across misinformation paradigms. A teacher can use insights from both while assessing actual transfer.
34. An explicitly quoted false claim needs attribution
When an article discusses misinformation, it may quote an initial false claim to explain why it was retracted. The quotation should not be read as the author endorsing the claim.
Source monitoring helps preserve who originally said the statement and what the later evidence established. A pupil summarising the article should use precise reporting terms such as initially claimed and later corrected. These small lexical distinctions protect factual meaning.
35. A current official source can supersede an old report
A library might issue an early timetable and later replace it with an authorised final schedule. The older schedule may remain widely circulated and familiar after it is no longer operative.
A reader should verify effective dates and controlling status before acting. The CIE is a psychological research effect, not proof that every remembered old schedule caused a memory-based error. The practical task is simply to use the current reliable record.
36. Source attribution alone may not repair the causal model
A pupil could accurately name the newspaper that published an incorrect account but still use the article’s withdrawn explanation in a later answer. Correct attribution helps but does not necessarily settle reasoning.
Ask which causal link the retraction invalidated and what the current evidence now supports. This moves beyond a bibliography exercise. Accurate vocabulary must represent both provenance and the content’s current epistemic status.
37. A strong correction should state the true proposition
A practical message begins by clearly identifying the accurate claim. It may then name the specific old claim as incorrect and explain why it is unsupported, using a suitable replacement account when evidence permits.
The structure should be concise and comprehensible to its audience. A lengthy correction containing many competing details may create unnecessary confusion. The best wording depends on the topic, available evidence and reader’s needs rather than a fixed magic sentence.
38. Explain why the original claim appeared
Sometimes a mistaken early report arose because an initial observation was misinterpreted or a preliminary estimate was used. Explaining that origin can help readers understand why the false claim was encountered.
The explanation should be supported, not an invented excuse. A careful account distinguishes an honest measurement error from deliberate manipulation without guessing at motives. Knowing how the error arose can make the corrected narrative more coherent and easier to retain.
39. Keep the correction tied to the source
A current official statement may retract one part of an earlier announcement while leaving other parts intact. Students should record exactly which proposition changed and which remains valid.
A broad statement that the entire announcement was false could be as inaccurate as the original error. Vocabulary like withdrawn, amended, partially corrected and unchanged helps represent the scope of revision.
40. Repetition of the false claim has a cost
When discussing misinformation, an educator should avoid repeatedly presenting unsupported claims as free-standing headlines without clear corrective context. A student might remember the familiar formulation while losing the surrounding warning.
This is a risk to consider, not a reason to conceal necessary details from analysis. Place the accurate account clearly, mark the invalidated claim and explain the correction. The goal is to make the correct information more accessible during later reasoning.
41. A delayed transfer question is informative
A student may correctly answer the immediate question ‘Which claim was retracted?’ but later rely on the old cause when writing a story summary. A delayed inference question can reveal whether the correction changed the working explanation.
Such a classroom assessment is not a formal measure of the CIE without controlled conditions. Still, it is more educationally useful than testing only whether the pupil remembers the word retraction. The aim is accurate independent use.
42. A matched control condition matters in research
CIE experiments often compare groups exposed to misinformation and retraction with suitable controls. The difference in later false inferences helps determine the residual influence under the task.
A vivid anecdote about one person repeating an error cannot establish the size of a causal effect. Experimental conclusions depend on comparable stimuli, response rules and appropriate analysis. A credible public guide should not invent population percentages from a single fictional example.
43. The effect can arise under cognitive load
Comprehension and memory retrieval compete with other tasks for attention. Research has examined how divided attention, including listening to information while performing a driving task, changes the effectiveness of corrections.
Those studies provide controlled evidence about workload and message processing. They do not justify making safety-critical decisions while distracted. A practical lesson is to read consequential updates in suitable conditions and use authoritative written records to verify the conclusion.
44. Media literacy should separate stages
A useful lesson distinguishes exposure to an initial report, the later corrective message, the current evidence and the learner’s final inference. These stages should not be compressed into one undated claim.
A timeline allows pupils to see why an early assertion was reasonable to report provisionally, when it became unsupported and what is known now. The objective is not to shame people for reading an earlier notice, but to prevent its false content from controlling current decisions.
45. Primary English can use a fictional correction
Tell pupils that a school character initially thought the missing library book was under a chair, then learned it was on a shelf. Ask what the character believed first and where the book was actually found.
The classroom should establish the corrected location clearly and test it in a new sentence. This is a safe age-appropriate analogy to updating an event explanation, not a CIE experiment or a diagnosis of children’s memory. The teacher can focus on the words thought, found and corrected.
46. Secondary English can track a changing claim
A Secondary learner reads a fictional incident report with an initial proposed cause, an explicit retraction and a later supported explanation. Ask which facts remained true and which explanatory links must be removed.
A strong summary includes the latest version without silently restoring the withdrawn cause. The pupil should cite the correcting sentence and name any remaining uncertainty. This teaches narrative accuracy as well as critical reading of evolving documents.
47. Science requires explanatory replacement, not a slogan
A student may learn a scientific misconception from a familiar explanation. A correct lesson should explain the actual mechanism and give an example that the old model cannot handle.
The principle resembles the coherence-gap account of corrections: readers need an accurate causal alternative rather than only a bare denial. But the scientific explanation must be grounded in reliable knowledge, and transfer must be tested. This is not a licence to invent neat but false explanations for unfamiliar phenomena.
48. Mathematics requires correcting the model
A pupil might assume that multiplying always produces a larger result. A teacher can show that multiplying a positive number by a fraction between zero and one yields a smaller positive result.
The revision should explain the relevant number relationship, not merely present one counterexample without context. Ask for an unfamiliar numerical case to test transfer. The CIE analogy concerns persistence of a wrong rule, while mathematical truth follows from the actual operations and definitions.
49. History and GP need a correction ledger
An initial historical report or policy claim may be updated as new records emerge. Students should identify the earliest claim, its evidential status, the correction and what remains contested.
In General Paper, distinguish what a source originally alleged from what later evidence established. An argument can be coherent while relying on invalidated premises. The correct response removes or replaces those premises with supported ones, rather than simply adding a footnote and continuing the original inference.
50. Translation must preserve the retraction status
A translated news report may state that an earlier claim was withdrawn, was revised or was disproved. These words have different strengths and should not be substituted casually.
The translator must preserve which proposition was invalidated and what evidence supports the new statement. A fluent target-language summary can spread misinformation if it omits the correction or keeps the older cause as confirmed. Accuracy requires checking the full timeline and source.
51. Workplace documents need current instructions
An earlier operational notice may have authorised one method that was later replaced. Staff who remember the old method can continue applying it despite knowing that an update was issued.
The relevant real-world action is to consult the current authorised procedure and verify that the replacement instruction is understood. A psychological account of continuing influence does not determine the compliance rule. Current records, permission and safety procedures must govern the work.
52. A five-column retraction audit
Create columns for initial claim, original source, correction, supported replacement explanation and current open question. Ask the learner to mark each as confirmed, withdrawn, provisional or unresolved.
The ledger externalises what memory can blur. It helps prevent an old causal explanation from re-entering a summary unnoticed. The table is a practical learning aid rather than an EEG measure or a standardised scientific test of continued influence.
53. An eight-step correction reading routine
First identify the current reliable source. Second record the exact earlier claim and its date. Third state what was invalidated. Fourth remove the old claim from the causal explanation. Fifth add a supported replacement if available.
Sixth mark unknown causes as unknown. Seventh test the corrected understanding with a fresh question. Eighth revisit the source if new evidence appears. This sequence is a classroom protocol, not a claim that every participant in a psychological experiment consciously executed eight steps.
54. The final lesson is correction that reaches reasoning
The continued influence effect demonstrates that a corrected sentence can lose its factual authority while retaining some influence on the explanations people generate. Research has identified several mechanisms and useful corrective strategies, while also showing that outcomes depend on the design and context.
For vocabulary education, the most important achievement is not merely remembering that an error occurred. It is using the current supported account to answer new questions, while keeping the earlier report correctly labelled as withdrawn. Accurate words, sources and causal relationships make correction meaningful.
Full worked case: remove the false causal link
A fictional pupil reads an early school report claiming that library damage caused a room closure. An update says no damage occurred, and a later verified notice establishes that the room closed for an electrical inspection. The pupil can still remember the old claim, but should no longer use it as an explanation.
Ask two separate questions: which earlier statement was retracted, and what now explains the closure? A good answer gets both right and does not cite the withdrawn report as evidence for the current cause. This tests meaningful revision rather than simple recall of a correction label.
Eight-stage accurate correction workbook
- Identify the earliest claim and label its original evidential status.
- Find the correcting source and record exactly what was invalidated.
- Distinguish a retraction from a verified replacement explanation.
- Remove unsupported causal arrows from the event model.
- Add an accurate alternative explanation only when trustworthy evidence provides it.
- Mark remaining unanswered questions honestly instead of inventing details.
- Test the new account with a fresh inference or transfer question.
- Use current source attribution and correct status vocabulary in the final summary.
Frequently asked questions
1. What is the continued influence effect?
The lingering effect of corrected misinformation on later reasoning or judgement.
2. Can someone remember a correction and still be influenced?
Yes. The original research found that explicit awareness does not always remove all misleading inferences.
3. Who published the classic 1994 experiments?
Hollyn Johnson and Colleen Seifert.
4. What is a causal coherence gap?
An explanatory gap left when a withdrawn cause is removed from the reader’s event model.
5. Do alternative explanations help?
Supported alternatives have reduced continued influence in some controlled studies.
6. What if no verified replacement cause is available?
Say the cause is unknown rather than inventing a new story.
7. Is CIE identical to a backfire effect?
No. Residual influence can remain even when a correction improves accuracy.
8. Do corrections typically make misinformation stronger?
No. Reviews find genuine factual corrections can improve accuracy, and strong backfire is uncommon.
9. What is a source-memory explanation?
One possible mechanism involves confusing the origin or status of an earlier false claim.
10. What is the best learning outcome?
Use the current accurate explanation in a new question, while remembering that the old claim was withdrawn.
Foundational research, contemporary reviews and connected reading
- Johnson and Seifert (1994): Sources of the Continued Influence Effect — original experiments and alternative causal explanations.
- Lewandowsky, Ecker, Seifert, Schwarz and Cook (2012): Misinformation and Its Correction — comprehensive review.
- Ecker, Lewandowsky and colleagues (2022): The Psychological Drivers of Misinformation Belief and Its Resistance to Correction — modern cognitive and communication review.
- Chan, Jones, Jamieson and Albarracín (2017): Debunking — A Meta-Analysis of Psychological Efficacy — correction and residual effects.
- Chan and Albarracín (2023): A Meta-Analysis of Correction Effects in Science-Relevant Misinformation — scope of science-related studies.
- Porter and Wood (2024): Factual Corrections — Concerns and Current Evidence — accuracy improvement and rare backfire.
- Irving, Clark, Lewandowsky and Allen (2022): Correcting Statistical Misinformation About Scientific Findings — correlation versus causation.
- Listening to Misinformation While Driving: Cognitive Load and the Effectiveness of Repeated Corrections — attention and correction conditions.
- Vocabulary | What Is the Knowledge Revision Components Framework? — coactivation and correction of misconceptions.
- Vocabulary | What Is the Source-Monitoring Framework? — source misattribution as one contributing possibility.
- Vocabulary | What Is a Causal Network in a Story? — correcting causal event relationships.
- Vocabulary Learning Hub — responsible source reading and accurate language use.
The continued influence effect reminds us that a corrected sentence can remain influential when it is still connected to a familiar explanation. Its practical educational lesson is to correct the causal account, not simply the label attached to the error. The strongest reader identifies the invalidated claim, understands the supported replacement and uses the current evidence reliably in new reasoning.