Source confusion is a memory-attribution error. A fact, phrase, image, or idea feels familiar, but you assign it to the wrong speaker, document, place, or experience. The content may be accurate even while its origin is wrong.

That distinction matters. Forgetting a number and remembering the right number from the wrong report are different failures. The second is a source-monitoring error.

Three maps, one familiar number

The following scene is illustrative. Jalen, the municipality, the maps, Station 8, the files, the meeting, and the outcome are invented; they do not describe a study or real incident.

Jalen works on a fictional stormwater planning team. Before a meeting, he reviews three maps. They use the same colors and layout, but they serve different purposes:

  • a verified sensor report;
  • a simulated training file;
  • a colleague's draft forecast.

Only the simulation says Station 8 failed twice.

At the meeting, Jalen remembers the number two and attributes it to the verified report. He recommends an inspection. The memory feels specific because the number really was on one of the maps. Its familiarity does not reveal which map carried it.

When the team opens the source files, the mistake becomes visible. They do not conclude that the station is safe. They mark the claim unverified, inspect the live sensor log, and redesign the templates so reports, simulations, and forecasts have distinct headers and persistent provenance fields.

The critical error was not inventing the number. It was placing remembered content in the wrong external source.

How source monitoring works

The source-monitoring framework describes how people judge the origins of memories, knowledge, and beliefs. It does not assume that every memory carries a source label that can simply be read at retrieval. Instead, people infer origin from available features and from the demands of the current decision, as Marcia Johnson, Shahin Hashtroudi, and D. Stephen Lindsay explain in their framework review.

Those features can include:

  • perceptual detail, such as a voice, font, color, or sound;
  • spatial and temporal context;
  • semantic or emotional detail;
  • traces of the mental work used to generate an idea;
  • familiarity or memory strength.

The judgment also depends on what you are trying to decide. "Have I encountered this?" asks for recognition. "Which document contained it?" asks for source attribution. A person can answer the first correctly and the second incorrectly.

Source cues can be weak because they were poorly encoded, became less accessible over time, were disrupted by interference, or resemble cues from another source. Retrieval instructions and response criteria matter too. None of these possibilities can be diagnosed from one wrong answer, and source confusion does not require later information to overwrite the original memory.

External, internal, and reality monitoring

Researchers separate several source decisions that everyday language often blends together.

External source monitoring distinguishes one outside origin from another: which speaker made a claim, which map reported a failure, or which article contained a statistic.

Internal source monitoring distinguishes among internally generated acts: whether you merely considered sending a message, rehearsed it, or decided to send it.

Reality monitoring distinguishes internally generated information from externally perceived information: whether a detail was imagined or actually seen.

These tasks share an attribution problem, but they are not interchangeable. A study comparing two speakers does not automatically tell us how people separate imagination from perception.

What experiments can show

Source confusion is clearest when a study separates memory for an item from memory for its origin. Several paradigms do that in different ways.

When familiarity outruns recollection

In Larry Jacoby and colleagues' false-fame experiments, participants first read nonfamous names and later judged whether names were famous. The three experiments included 36, 32, and 16 participants. In the first experiment, earlier exposure made a nonfamous name more likely to be judged famous after 24 hours, while the corresponding immediate test did not produce that false-fame pattern. Later experiments changed whether participants were asked to recollect the list source or could use that source to correct the judgment. The results show how familiarity can influence a decision when its experimental origin is not spontaneously recovered (Jacoby et al., 1989).

The boundary is important. This was an artificial name task with small student or volunteer samples. It does not estimate how often people misremember real public figures, and false fame is not another name for source confusion. It is one outcome that can reveal a source-recollection failure.

When a suggestion is assigned to the event

Postevent information creates a different attribution problem. An eyewitness may remember a suggested detail but later treat it as part of the witnessed event. Experiments by Lindsay and Johnson found that source similarity can make this discrimination harder (Lindsay & Johnson, 1989).

That finding does not mean every misinformation effect is source confusion. Nor does it establish that the suggestion erased the original event memory. A later report can also reflect response criteria, demand, uncertainty, or other memory processes. The safe claim is narrower: wrong-source attribution is one route by which postevent information can affect a report.

When another person's idea feels self-generated

Cryptomnesia narrows the problem further. In three category-generation experiments, Alan Brown and Dana Murphy observed participants reproducing previously encountered responses despite instructions to avoid repetition (Brown & Murphy, 1989). The laboratory outcome supports the possibility of inadvertent reuse. It does not establish intent, legal plagiarism, or a real-world rate of copied creative work.

Age changes the task, not the definition

It is tempting to turn lifespan findings into a simple ranking: children and older adults confuse sources; other adults do not. The evidence is more conditional.

In an aging study, younger and older adults heard words spoken by two people, then identified the speaker. The researchers varied how distinctive the voices and the accompanying mental operations were. Older adults performed worse when the relevant cues were similar, but the age difference was absent when either cue was more distinctive. Combining cues did not uniformly help and could disadvantage the older group relative to a single distinctive perceptual cue (Johnson et al., 1995).

That result makes cue design part of the finding. It does not support an inevitable decline in every kind of source memory or the idea that adding more context is always better.

Developmental research is equally sensitive to the test. In a study of 132 children aged 3 to 8, Debra Poole and D. Stephen Lindsay trained children to distinguish witnessed events from events they had only heard described. Training reduced false reports to direct questions among 7- and 8-year-olds, but not among the younger children (Poole & Lindsay, 2002).

The result was age- and transfer-dependent. It does not show that younger children cannot distinguish reality from imagination in general. Interview language, named alternatives, delay, source similarity, and the demand to apply a strategy in a new task can all change performance.

Clinical findings require a separate boundary

Source-monitoring tasks are also used in psychosis research. A 2022 systematic review and meta-analysis screened 5,256 records and included 44 studies with 1,566 patients and 1,175 controls. Across measures, source-monitoring performance was lower in psychosis groups, with larger differences in some internal-monitoring and imagined-stimulus conditions (Damiani et al., 2022).

Those group averages are not a diagnostic test. The studies varied in symptom profiles, stimulus modality, task design, medication, and comparison groups. Associations with cognition and symptoms were based on smaller, heterogeneous subsets.

An earlier meta-analysis provides a useful check. It found a large deficit for pair recognition and a medium deficit for source recognition in schizophrenia samples, but no evidence that internal, external, or reality-monitoring source types differed from one another (Achim & Weiss, 2008). That pattern is compatible with a broader associative-memory difficulty rather than a unique reality-monitoring impairment.

An everyday wrong-source memory therefore does not imply psychosis, predict hallucinations, or diagnose a neurological problem.

Source confusion and nearby memory effects

The same examples often circulate under several labels. Keeping the measured outcome in view prevents them from collapsing into one vague idea.

  • Source amnesia means retained information cannot be linked to where or when it was learned. Source confusion usually involves assigning it to the wrong origin rather than reporting no origin.
  • Misinformation effect means misleading postevent information changes a later memory report. Source misattribution can contribute, but the effects are not synonymous.
  • Cryptomnesia is inadvertent production of encountered material as if it were new or self-generated. It is one specific attribution outcome.
  • False memory is broader. Someone may remember a detail or event that did not occur without assigning an existing memory to the wrong source. Source content can also be true while its attribution is false.
  • Imagination inflation is increased confidence that an event occurred after imagining it. Source confusion is one proposed pathway, but the classic experiment measured confidence, not independent proof of a detailed false memory (Garry et al., 1996).
  • Sleeper effect concerns delayed persuasion under specific message and discounting-cue conditions. A meta-analysis covered 24 reports and 72 independent experiments or data sets, finding that cue timing, message impact, ability, and motivation shaped the effect (Kumkale & Albarracin, 2004). An attitude shift is not the same outcome as assigning a fact to the wrong source.

Sources