Why can I remember a fact but forget who told me?

Source monitoring is not simply checking a permanent label attached to every memory. We often make a judgment from whatever clues remain. An image, a setting, a remembered voice, or a trace of the effort involved in imagining something may help us decide what kind of experience produced the memory.

Marcia Johnson and Carol Raye described reality monitoring in 1981: distinguishing externally perceived events from internally generated thoughts and images. Perceived experiences often include sensory and contextual details. Imagined experiences may include traces of the mental work used to create them. These are useful clues, not guarantees. An imagined scene can be vivid, while a real conversation may leave only a vague impression.

Johnson, Shahin Hashtroudi, and D. Stephen Lindsay broadened the framework in 1993. We may distinguish two external sources, such as a doctor and a neighbor; two internal sources, such as a dream and a plan; or something imagined from something experienced.

Sometimes these decisions happen quickly. Other times, we deliberately search for context. Neither process makes familiarity and attribution interchangeable. Recognizing a statement answers whether it seems known; identifying its source asks how it became known.

A set of 1989 experiments by Larry Jacoby, Colleen Kelley, Judith Brown, and Jennifer Jasechko made this distinction especially clear. Participants encountered names belonging to people who were not famous. Later, previously encountered nonfamous names were sometimes more likely to be judged famous, particularly after a delay.

The names had not acquired public importance. They had acquired familiarity. At the later test, participants could mistake that familiar feeling for a sign of fame rather than connect it to the earlier list. This became known as the false-fame effect.

The result is narrower than a claim that familiar names always fool us. In this task, earlier exposure could influence a later judgment when its source was not correctly identified. A real signal of previous contact was being given the wrong explanation.

That distinction offers a useful everyday question: does this seem credible because I can identify its evidence, or because I have encountered it before? Repetition can support learning, but it does not independently establish trustworthiness. Seeing the same claim again does not tell you whether it came from a reliable source, an uncertain suggestion, or several retellings of one mistake.

In 1989, Alan Brown and Dana Murphy studied people taking turns generating examples from categories. Later, participants tried to remember their own contributions or produce new examples without repeating earlier ones. Despite those instructions, they sometimes reproduced contributions that other people had supplied.

This is cryptomnesia: remembering material while failing to recognize its earlier source, so that it can seem newly generated. The finding does not excuse deliberate copying or resolve every dispute over creative credit. It shows that an attribution error can occur without an intention to borrow.

Collaborating with AI adds another possible source. Imagine developing a product name through a long exchange. You suggest options, the assistant adds alternatives, and both sets are revised. Later, remembering the final wording may be easier than reconstructing who introduced each part.

A preregistered experiment reported by Tim Zindulka and colleagues in 2026 involved 184 participants generating and elaborating ideas with and without AI. A week later, participants identified which material had involved AI. Attribution was less reliable after AI use, especially when workflows mixed human and AI contributions. That finding concerns the tested collaboration tasks; it does not establish that every AI-assisted idea will be misattributed.

Stories contain a mixture of invented events and information about the real world. Readers can learn useful things from that mixture. The difficulty is that remembering a detail later does not necessarily bring back the context needed to assess it.

Elizabeth Marsh, Michelle Meade, and Henry Roediger examined this in their 2003 study, Learning Facts from Fiction. Participants read short stories containing correct and incorrect information, then answered general-knowledge questions. Story information influenced later answers, including answers containing misinformation. Exposure also sometimes increased the belief that the information had been known before the experiment.

The finding does not mean readers generally confuse novels with reference works. It suggests that information learned from fiction can enter general knowledge while some of its source context becomes less available. Remembering where a detail appeared can therefore matter as much as remembering the detail itself.

This distinction becomes especially important with first-person writing. A person describing their childhood, a fictional narrator recalling childhood, and a person reporting an imagined scene may all use similar language. The words alone do not establish whose life is being described. A vivid passage is evidence that the passage exists, not automatic evidence that its events happened to the person who wrote or shared it.

Maryanne Garry, Charles Manning, Elizabeth Loftus, and Steven Sherman tested a related question in 1996. Participants first answered questions about childhood events, then imagined selected events before rating their confidence again. Imagining an event increased confidence that it had happened, an effect called imagination inflation.

The important limit is that increased confidence is not the same as a detailed false autobiographical memory. The study did not show that everyone who imagines something develops a rich recollection of it. Its finding concerned a shift in belief or confidence after imagination.

Questions can also introduce information into remembering. In Loftus and John Palmer’s 1974 car-accident experiments, the verb smashed produced higher speed estimates than milder wording. In a later test, that condition also produced more reports of broken glass that was not present. Particular wording effects vary across settings and replications, so the lesson is not that every question rewrites memory.

A constructive response is to separate the remembered detail from its possible source. Did you see it, hear it later, infer it, or imagine it? Lindsay and Johnson’s 1989 eyewitness study suggests that attending explicitly to source information can reduce some kinds of suggestibility. It is not a universal safeguard, but it supports treating uncertainty about origins as useful information rather than something to conceal.