KurisutinaIdentity and status. Marcia K. Johnson, Shahin Hashtroudi, and D. Stephen Lindsay, Source Monitoring. Psychological Bulletin 114(1), 3–28, July 1993. Peer-reviewed theoretical review, not a new primary experiment or systematic meta-analysis. DOI: 10.1037/0033-2909.114.1.3; PubMed metadata; author-hosted scanned publisher PDF. Accessed 22 September 2026.
Factual core
The framework proposes that people attribute memories to sources using available perceptual, contextual, semantic, affective, and cognitive-operation information, rather than ordinarily reading an unambiguous origin label. Relatively rapid assessments and more deliberate retrieval/reasoning can check one another; both depend on criteria influenced by goals and beliefs. Source attribution can be partial and graded. Reviewed experiments relate source errors to feature similarity, encoding conditions, and test procedures. Recognition and source judgments can dissociate or covary, depending on the information each task uses. Indirect performance can also retain source sensitivity without successful explicit attribution. The review discusses misinformation, familiarity errors, inadvertent plagiarism, and clinical cases within this framework. It does not uniquely localize source memory to one brain region. Its proposed component architecture remains provisional. Figure 1 illustrates differentiation and overlapping task demands; it is not a fitted response-time curve. These are framework claims and a synthesis of cited work, not independent verification of every cited experiment. Full review, pp. 3–21.
Evidence audit
The review reports both reductions in misleading-source claims after source-oriented testing and cases where source errors persist despite explicit instructions. Neither finding warrants a universal corrective-interview rule. Those primary reports have not been fully read for this note. Several supporting references were conference presentations, manuscripts in preparation, or in press when cited; for example, the response-signal work cited on p. 9 was a 1992 conference presentation. The clinical discussion and proposed neural division of labor require their own primary-source audit. This reading supplies no new pooled effect size, formal identifiability analysis, or modern replication assessment.
Original interpretation for extraction and transfer
Separate provenance from a person's attribution. A system can reliably log that a sentence came from an interview while remaining uncertain whether the person originally saw, inferred, imagined, or heard its content. Those are different source questions. Store the acquisition record, the person's dated attribution, and the system's attribution estimate separately. An explicit software source field is useful engineering; the psychological argument against a universal origin-label readout does not prohibit it. Conversely, preserving an interview's filename does not establish historical truth or reproduce the person's source judgment.
Feature richness is not source verification. Vividness, contextual detail, and a plausible causal story are candidate predictors, not evaluator ground truth. If an elicitation system supplies those details and later retrieves its own elaboration, a fluent, detailed answer can be supported by the interaction rather than the original event. A provenance-aware recipient should identify the supporting observations and mark its own generated additions. Whether this improves forecasts of the person's later errors is a separate empirical question from whether it improves documentary accuracy.
A changed answer need not mean a changed stored event. As an exact illustration, stipulate unchanged source evidence summarized by p=.6, and a decision to assert source A only when p>c. Changing c from .5 to .8 changes the overt answer without changing the stipulated evidence. This is our toy, not the review's computational model. Actual instructions might also change attention, retrieval, or the evidence available, so an instruction effect alone does not identify a pure criterion shift. More generally, attribution, accessibility, and lasting content change need distinguishable predictions before fitted parameters receive those names.
Do not equate a generated error with clinical confabulation. The useful analogy is functional: a system may retrieve a claim yet misidentify why it is available, or fail to check it against incompatible evidence. Shared terminology does not establish shared pathology, subjective conviction, or a common biological mechanism. Likewise, successful source classification does not establish transfer of the subjective feeling of remembering.
A falsifiable extension, not a study performed
Use neutral controlled episodes with evaluator-held event and source records. Cross source distinctiveness during acquisition with a later manipulation intended to alter attribution criteria, such as instructions or response costs. Keep the response interface and scoring fixed for that contrast, retain novel lures and an explicit uncertainty response, and measure the joint pattern of content, origin claims, confidence, and subsequent choices. Do not assume that a feature manipulation affects only source encoding or that instructions affect only criteria. Use prespecified competing measurement models and test their predictions on new source combinations and test contexts; a richer model must beat an equally informed simple baseline.
Separate initial tests from delayed-only conditions because asking for source can itself change later retrieval. Compare paired recipients that share an acquisition history but receive different portions of its report and contextual evidence. Then separately randomize acquisition policies to estimate their effects on the person. Hold evaluator truth out of extraction inputs; an oracle record is a distinct reference. Forecast the person's future source-response distribution before seeing those answers, and have recipients take their own actions under matched continuation conditions. Score event-optimal actions, calibrated forecasts, and recipient behavioral distributions separately. A recipient that corrects the person's attribution error can improve historical accuracy while departing from personal behavior; neither target should silently substitute for the other.
Primary experiments most useful next — not full-paper reads
Subsequent reading: Lindsay 1990 now has a full primary audit. The descriptions below preserve the access/screening stage of this review; Lindsay and Johnson 1989 remains unread beyond metadata and abstract.
- Lindsay and Johnson (1989), The eyewitness suggestibility effect and memory for source. Memory & Cognition 17, 349–358. DOI: 10.3758/BF03198473; PubMed. Read next to determine exactly what changed between the recognition and source-monitoring tests and whether response format, demand, retrieval orientation, and criterion explanations can be separated. The publisher page was inspected for metadata and abstract and currently presents subscription access; no full-text read or verified download is claimed here.
- Lindsay (1990), Misleading suggestions can impair eyewitnesses' ability to remember event details. Journal of Experimental Psychology: Learning, Memory, and Cognition 16(6), 1077–1083. DOI: 10.1037/0278-7393.16.6.1077; author-hosted PDF. Read next as a boundary test for source-oriented warnings and to distinguish original-detail loss from intrusion of later information. The author listing and indexed first-page text were inspected, not the complete methods or results. Check the bundled timing/context manipulations before attributing an effect to a single source-discriminability mechanism.
Reading and archive record
Read all main text, footnotes, theoretical discussion, evidence descriptions, qualifications, and Conclusions on journal pp. 3–21. Read the bibliography through p. 28 without treating its cited papers as completed reads. Visually inspected PDF pages 1, 3, 7, 9, 13, 17, and 18, covering metadata, criteria, the sole figure and caption, misinformation, clinical interpretation, and proposed architecture. Figure 1 is conceptual; no graph digitization was appropriate. The final page's subscription form is not research content. No supplementary material was identified in this article or author listing, and none was read. No raw data or code were retrieved.
The retained 26-page PDF is byte-identical to the verified queued author download. Its image pages lack usable embedded text. The readable file was generated locally with Tesseract 5.5.3 at 180 dpi using automatic page segmentation; page boundaries and journal-page numbers were added. It is uncorrected OCR with occasional word and punctuation errors, not an authoritative transcription. All 26 page sections are nonempty; consequential passages were checked visually as listed above. Copyright is ©1993 American Psychological Association; no open reuse license was identified.
Local PDF, readable OCR, and provenance and hashes.
Subsequent primary reading — checkpoint7. Lindsay and Johnson1989 is now fully read, including the narrative Appendix and all six tables. The source-focused test reduces false picture attributions and correct endorsements, while the study's two “new” response categories answer different questions. This updates the earlier queue status without extending the review itself into a primary experiment.