# A Record Repeated Is Not a Record Corroborated *How one assertion becomes six documents, and why copies must be weighed, not counted* Kanwar Partap Singh Gill, MD · November 5, 2026 · Practice Five people can repeat a statement without giving us five reasons to believe it. In conversation the point is obvious. Someone says he heard it from a colleague, who heard it from a neighbour, who read it somewhere, and we ask, reasonably, where the account began. The chain may carry something true. Its length does not make the first link any stronger. A medical record can make the same chain hard to see. Each repetition acquires a new date, a different heading, sometimes another clinician's name. The statement turns up in several places that look like separate sources, and a reader may fairly assume that someone along the way checked it. Picture a hypothetical patient filling in a questionnaire at her first visit to a new practice. Under past medical history there is a box marked *asthma or breathing problems*. As a child she had a cough that lingered every winter, and a doctor once listened to her chest and said something she no longer remembers exactly. She ticks the box. What follows is invented, and simple enough to set out as a list, with each document's source and what happened to the claim on the way. 1. **The questionnaire.** Its source is the patient's memory, passed through a category that joins two different things with the word *or*. The claim: something about breathing, once. 2. **The visit note.** Its source is the questionnaire, read by a clinician who writes *History of asthma*. The words *or breathing problems* have gone, and a vague childhood memory has acquired the name of a diagnosis. 3. **The problem list.** Its source is the visit note. The entry reads *Asthma*. Now *history of* has gone as well, and the condition sits in the present tense among the patient's active problems. 4. **A later visit note**, about something else entirely, carrying forward the previous assessment: *Asthma, stable.* Nobody examined anything to arrive at *stable*. It was inferred from silence. 5. **A referral letter**, generated from the problem list and sent to another office, where it arrives looking like a summary of the patient's established conditions. 6. **That office's consultation note**, written by a careful physician who read the letter: *Known asthma.* Six documents now agree. A reader opening the chart a year later will find the condition in the history, on the problem list, in a progress note, in correspondence and in a specialist's own words. It looks like a fact confirmed from several directions. It is one tick in one box, travelling. Count the appearances and there are six. Count the observations and there is one, and that one is a remembered remark filtered through a checkbox. Nothing in the trail tells us whether she has asthma; she may, and information absent from these six documents might settle the matter either way. What the trail does establish is narrower and, I think, more important: the wording grew steadily more definite while no new evidence entered the record at all. This essay is about the distance between those two counts, and about why repetition can feel like confirmation when it is only distance travelled. When several documents agree, how much more should we believe? The answer depends less on how many there are than on how they came to agree. ## The Grammar of Drift The first thing to notice about the imagined trail is that the copying was not exact, and that the drift ran in one direction. Qualifiers fell away; none were added. The words that disappeared are worth examining, because they were not decorative caution. Each answered a different question. *Or breathing problems* described a category: what kind of thing was being claimed. *As a child*, implicit on the questionnaire and lost by the second document, described time. *History of* described status: something past, recorded as past. The patient's own uncertainty, never written down at all, described the confidence of the source. Remove any one of these and every remaining noun can be accurate while the sentence as a whole says something nobody ever observed. The tense deserves particular attention. *Breathing problems as a child* is bounded in the past. *History of asthma* keeps the past but names it. *Asthma* on a problem list is present and continuing. *Asthma, stable* adds a judgement about the present that no one made. And *known asthma*, the most confident phrase in the trail, contains a word about knowledge. Known by whom? The adjective asserts that somebody knows, and it was added by the writer furthest from any evidence, for whom the knowing consisted of reading a letter. None of the people along the way did anything I would call careless. Translating a person's words into clinical vocabulary is ordinary and necessary work, and it was at those ordinary moments that the uncertainty leaked out. It helps to distinguish two acts that look alike in a record. Faithful copying carries a sentence forward unchanged; its risk is that an old statement goes on asserting itself in a new setting. Interpretive rewriting changes the sentence: a report becomes a diagnosis, a diagnosis loses its tense. The fourth and fifth documents in the trail were faithful copies; the second, third and sixth were rewritings. They carry different responsibilities, and blaming carried-forward text indiscriminately obscures where the drift actually happened. Scholars who study hand-copied manuscripts noticed this direction of drift long ago. A copyist who met an obscure word, a rough construction or a puzzling reading tended to replace it with something easier: the familiar word, the smoother phrase, the reading that made obvious sense. He seldom did the reverse. From this observation textual critics drew a working principle with a Latin name, *lectio difficilior potior*: other things being equal, the more difficult reading is the stronger, because scribes simplify far more often than they complicate. The chart drifts the same way. *Breathing problems as a child, details uncertain* is the harder reading. *Asthma* is the easier one, the reading that fits the field, the list and the referral form. The awkward, hedged original is more likely to describe the actual state of knowledge, and it is precisely what copying wears away. The opposite failure exists too, and it would be unfair to leave it out. A tentative phrase can survive long after the question it marked has been settled: *possible asthma* repeated for years after testing confirmed the diagnosis, so that a reader is denied a confidence the evidence has earned. The fault in both directions is the same. The record fails to show change. Words should match the state of knowledge, and the state of knowledge has a history that the record ought to be able to show. ## Weighed, Not Counted Textual critics did more than notice the drift. Over the nineteenth century, confronted with dozens or hundreds of manuscripts of a single classical or biblical text, each copied by hand from another and each carrying its own errors, they built a method for deciding what the copies were worth. It is associated with the German philologist Karl Lachmann, though the Italian scholar Sebastiano Timpanaro later showed how much of it Lachmann had inherited from predecessors, so that even the method's attribution is something of a copied simplification. Paul Maas set out its logic with severe economy in *Textkritik* (1927). Two of its principles bear directly on a medical chart. The first is that a manuscript copied from another surviving manuscript adds nothing to our knowledge of the original. If the exemplar survives, the copy is set aside as evidence for the text. This elimination of derivative witnesses sounds harsh only until one sees its logic. A copy can transmit what its source said, and add errors of its own. It cannot add independent testimony about anything earlier. The second is that shared errors reveal shared descent. If two manuscripts contain the same distinctive mistake, the same omitted line or the same misread word, they almost certainly descend from a common ancestor in which the mistake first appeared, and their agreement elsewhere counts, for most purposes, as one witness, not two. Agreement in error is the fingerprint of dependence. The maxim that gathers both principles is old and plain: manuscripts are to be weighed, not counted. A hundred late copies of one exemplar are, as evidence, a single witness. Two manuscripts from separate lines of descent may outweigh them all. Weighing is not the same as dismissing; the Italian philologist Giorgio Pasquali warned that later manuscripts are not necessarily worse, since a late copy may descend from a good source that is otherwise lost. The point is not to distrust copies. It is to know what kind of evidence each one is. Clinicians already practise a rough version of this method without its vocabulary. A paragraph that appears in ten consecutive notes with the same misspelling, the same outdated dose and the same oddly specific phrase is one paragraph written once and displayed ten times. The shared error gives it away. The imagined chart has its own fingerprint: the word *asthma* appears in every document after the first, and in each case it can be traced, by date and wording, to the second. Five documents, one witness, and that witness a clinician's reading of a checkbox. I grew up within a tradition that takes the faithful keeping of words very seriously. For most of its history the Sikh scripture was transmitted by hand. Tradition and scholarship both point to collections of hymns made in the time of the third Guru, Guru Amar Das, among them the manuscripts now known as the Goindval pothis, and to the volume compiled under Guru Arjan and installed in the Harmandir Sahib at Amritsar in 1604, with Bhai Gurdas as its scribe. Copies made from copies followed for generations before print. Scholars such as Gurinder Singh Mann have studied the early manuscripts to trace how the hymns were gathered and passed on, which is the critic's family tree by another name. Careful copying matters acutely in Gurmukhi because small marks carry grammar. The grammarian Professor Sahib Singh argued, in his studies of the grammar of Gurbani, that the small vowel signs at the ends of words often carry meanings of case and number, so that a stroke dropped by a tired scribe can change who is doing what to whom. I count that fidelity a real virtue. It is, however, the virtue of preserving a text whose authority was settled before the copying began. It offers no help in deciding whether an original deserved to be copied. Fidelity and verification are different duties, and a medical record, unlike a scripture, needs both. It must transmit faithfully what was said, and it must show, alongside, whether anyone has since had reason to believe it. ## Independence Is a Relationship What would genuinely independent support look like in the imagined chart? The question is worth asking precisely, because independence is not a property of a document. It is a relationship between a claim and the route by which each piece of support for it was produced. A test of lung function would be one kind of support: a different method, applied now, capable of contradicting the label. Records from wherever the childhood illness was treated would be another: a contemporary account by someone who examined her at the time. The most available source, and often the most neglected, is the patient herself, asked plainly. Her answer is not independent of the questionnaire, since she was its source, but it is a chance to recover what the checkbox lost: what was actually said to her, by whom, and whether anything was ever done about it. A new author does not make a source independent; neither does a new date, a new institution or a new kind of document. The consultation note in the trail had a different author, letterhead and specialty from everything before it, and it was the most derivative document of all. Conversely, a single careful note may contain several independent kinds of evidence: what the patient reported, what the clinician examined, what a test showed. The useful unit is not the document. It is the basis on which a particular assertion rests. Even genuine agreement needs thought. Two clinicians who each wrote *asthma* after hearing the same patient say *asthma* have produced two documents and one witness. Two clinicians who examined her separately and reached the same conclusion have produced something better, though if both applied the same assumption to the same ambiguous finding, their agreement still shares a root. Independent collection does not guarantee independent interpretation. Corroboration is a reason to look at the basis of one's confidence. It does not end the looking. Other trades know this pattern well enough to have named it. Intelligence analysts and journalists speak of circular reporting: information that seems to come from several independent sources and in fact comes from one, passed around until it returns to its starting point looking like confirmation. Scholarship has a cautionary tale of its own, and it is a good one. For decades, textbooks and journals repeated that spinach owed its reputation for iron to a misplaced decimal point in an old laboratory analysis, a tidy parable about careless copying. In 2010 the criminologist Mike Sutton went looking for the evidence behind the parable and found none. The decimal-point story had been spread largely by an unsourced 1981 article in the *BMJ*, cited by writers who cited one another, and the Norwegian scholar Ole Bjørn Rekdal later traced how that chain of citations turned a rumour into what he called an academic urban legend. The best-known story about a copied error was itself a copied error. Why does repetition persuade so well? Psychologists have studied the question for nearly fifty years. In a 1977 experiment, Lynn Hasher and her colleagues found that people judged statements more likely to be true simply because they had encountered them before, and later work on what came to be called the illusory truth effect found that repetition can sway people even when they have the knowledge to know better. These were laboratory studies of how people judge the truth of statements, not studies of clinical records, and they say nothing about how often charts are wrong. What they explain is the feeling. A chart that repeats a statement six times is, among its other functions, an efficient machine for producing the sense that something has been established. ## The Good Reasons Information Travels None of this makes carrying information forward a vice, and the case in its favour is stronger than critics of copying usually allow. A longitudinal record exists so that knowledge can outlive the encounter that produced it. A patient should not have to recount a painful history, a stillbirth, an assault, a cancer, at every visit to every new clinician. A severe allergy, carried faithfully from one note to the next and one institution to the next, may one day protect a person too ill to mention it. Most of what is copied in a chart is true, and much of it is true because someone, once, took care to establish it. A record that had to rediscover everything at every visit would throw away the continuity that makes it worth keeping, and a clinician who insisted on verifying every inherited statement from its origin would never finish a morning's work. Demanding original verification of everything is not rigour. It is a way of making the record unusable. There is also a perfectly rational form of trust that does not require re-examination: trust in a process. A laboratory result that travels from the laboratory's system into the chart automatically, carrying its date, its method and its source, can be relied upon not because anyone rechecked it but because the route it travelled is known and sound. That is the right model for much of what a record carries. The trouble begins when a statement travels by a route nobody can see. The published evidence supports this balance, not either extreme. A 2017 systematic review by Amy Tsou and colleagues in *Applied Clinical Informatics* examined copy-and-paste practice in electronic records and described its risks, among them redundant notes, conflicting information and the spread of errors, while stressing how limited the evidence linking the practice directly to patient harm remained. Its recommendations included making copied material identifiable and its source available. A report from the National Institute of Standards and Technology the same year, NISTIR 8166 by Svetlana Lowry and colleagues, examined the problem as a matter of human factors, identifying difficulties with the volume of text, unclear attribution and inadequate review of reused material, and addressing how interface design could make safe use easier. Neither source tells me how often a particular error occurs in any particular practice, and neither shows that copying is unsafe as such. What both support is traceability: a reader should be able to see where a statement came from. The standard I would propose is proportion. Scrutiny should follow importance, uncertainty and consequence, taken together. A statement that will shape a decision today, and whose origin is unclear, deserves a look at its source. An old description with no bearing on the present can be left in peace. A visible inconsistency deserves attention whatever its importance, because it is a sign that something in the chain has slipped. These are principles for spending attention, not an algorithm, and their purpose is to make reliance cheaper, not to make doubt universal. The narrow difficulty that remains is this: a borrowed statement should not quietly change its status by becoming familiar. Copying preserves an assertion. It does not test it. ## A Colophon for a Sentence Scribes in many manuscript traditions left a note at the end of their work, the colophon, recording who had copied the book, when, and sometimes from what exemplar and for whom. It was not always present and not always truthful, but where it survives it is often among the first things a textual scholar reads, because it says what kind of witness the manuscript is. A colophon does not make a copy more accurate. It makes the copy's relation to its source visible. Clinical sentences rarely carry anything of the kind. The note as a whole has an author and a date, but the sentences inside it arrive without histories of their own, and the reader cannot tell which were made today and which were inherited. Small phrases can do a colophon's work: *patient reports*; *per outside records, reviewed today*; *carried forward, not reassessed*; *confirmed by testing*, followed by a year. Each is short, and each tells the next reader what kind of witness the sentence is. The objection is obvious, and fair. A record in which every sentence wore its provenance would be unreadable, and a display that flagged every copied word would bury the few differences that matter under a blizzard of markings. Qualifications attached by habit become decoration, and decoration is ignored. The art lies in choosing where the colophon matters: at the claims a decision will rest on, at the claims whose origin is uncertain, and at the point where a claim changed, from a report to a diagnosis or from suspected to confirmed. There, a few words can save a later reader a long reconstruction. Elsewhere, silence is kinder. Something like this is also a matter for the people who build records, not only for those who write in them. If the software already knows that a sentence was carried forward from a note three years old, it could say so quietly, in a way a reader can consult without being shouted at. That is the human-factors argument in small: carefulness should not depend entirely on the stamina of the person typing. ## Four Dates A date at the top of a note creates an impression of freshness. It tells us when that document was written. It may tell us much less about when the information inside it was observed, when it was last checked, or whether it still applies. In reading a record I find it useful to separate four dates that a single sentence can carry: when the thing happened, when it was reported, when it was entered, and when it was later relied upon. This is a way of reading, not a description of how any particular system stores its fields, and in most records only one of the four is printed. A worked example, again invented, shows what the separation reveals. A man falls in his kitchen at the end of January. He mentions it in April, at a visit about his blood pressure, almost as an afterthought, and the clinician writes *fall at home* in the April note. In August a summary prepared for a referral lists *fall (April)*, taking the date from the note. In October the specialist's letter refers to his *recent fall*. Read as a timeline, the record is orderly and wrong. It places the fall in April and calls it recent in October, when the event happened in January, nine months before the word *recent* was written. Nothing was invented. A chronology of knowledge, the order in which the practice learned things, has quietly been read as a chronology of events. For many purposes the difference would not matter. For a reader trying to connect the fall to a new medicine started in March, it would matter a great deal. Time also changes the truth of statements that were accurate when they were made. A note records that a woman lives alone. A year later her sister moves in. Every faithful repetition of the original sentence preserves its words and loses its usefulness, and the relevant question is no longer whether it was ever true but when it was last confirmed. That changes the purpose of a review. Not every questionable entry is an error waiting to be pinned on someone. Sometimes the record needs updating because life went on. The fourth date, the moment of reliance, deserves more weight than it usually receives, because it is when a claim begins to do work. A label can sit harmlessly in a record for years. Then someone relies on it: a clinician choosing between two medicines, a surgical team weighing risk before an operation, or the patient herself filling in a form for a job with medical standards, a diving course or an insurance policy, where the same word may be read as a settled fact about her lungs. The weakness of the claim did not matter until that day, and on that day it may matter a great deal. This suggests where scrutiny is best spent. Not at the moment of copying, when nothing depends on the claim and no one has reason to question it, but at the moment of reliance, when something does. The useful question then is not only whether the statement is true, but when, and by whom, it was last established. Borges has a story that makes the point with comic precision. In "Pierre Menard, Author of the *Quixote*" (1939), a minor French writer of the early twentieth century sets out not to copy Cervantes but to compose, by his own route, pages that coincide word for word with *Don Quixote*; he completes the ninth and thirty-eighth chapters of the first part and a fragment of the twenty-second. Borges's narrator then compares a passage in the two versions. In Cervantes, the phrase calling truth the daughter of history, *la verdad, cuya madre es la historia*, is mere rhetorical praise of history. In Menard, a contemporary of William James, the identical words make an astonishing philosophical claim: that history is not an inquiry into reality but its origin. The joke has a serious edge. The meaning of a sentence depends on who wrote it, when, and on what basis. A chart is full of Menard's chapters. *Lungs clear* written on a March morning by a clinician who has just listened is an observation. The identical words carried into a September note, in a visit at which nobody listened, are a different sentence: an inheritance wearing an observation's clothes. Nothing on the page distinguishes them. Only the date of the act, and the act itself, would. ## Correction Without Erasure Suppose the patient in the first example is asked about her asthma, and the answer turns out to be uncertain. What should happen to the six documents? Medieval scribes had a way of dealing with unwanted text that a medical record must not imitate. Parchment was expensive, so a manuscript no longer wanted could be scraped or washed and written over, producing a palimpsest. The most famous of these is the Archimedes Palimpsest. In the tenth century a scribe copied several treatises of Archimedes; in the thirteenth, another scribe took the book apart, scraped its pages and wrote a prayer book across them, finishing the work in 1229. The Danish philologist Johan Ludvig Heiberg recognised the faint mathematics beneath the prayers in 1906, and after the book was sold at auction in 1998, imaging scientists spent years recovering what the scraping had nearly destroyed, including the only surviving copy of the treatise known as *The Method*. We have that work because the erasure was imperfect. A record's correction should be the opposite of a scraping. The convention of the paper chart was sound: a single line drawn through the mistaken entry, so that it remained legible, with the corrector's initials, the date and, where needed, a reason. Electronic records keep version histories and addenda that can serve the same purpose. The reason for keeping the mistaken text visible is not sentiment. The trail of how belief changed is itself evidence. A reader who meets *asthma* in an old referral letter needs to be able to find the later entry that revised it, and to see why, or the old letter will go on testifying unopposed. Patients have a part in this that is sometimes formal. Under the federal privacy rule, as I understand it, a patient may ask a practice to amend information in her record; the practice may decline if it judges the record accurate and complete, and she may then submit a statement of disagreement to be kept with it. That is a legal floor, not a description of how any particular office handles such requests, and it is no substitute for the ordinary correction that happens in conversation. The commoner route is simpler. A clinician asks, listens and writes a better sentence. How the question is asked matters. *The chart says you have asthma. Can you tell me about that?* is no challenge to the record. It is a courtesy to the person the record describes, who usually has no idea how far one sentence about her has travelled. Her account is not infallible; people forget the names of tests and the years of illnesses, and so do clinicians. But she is a source, not an annotation, and on some matters, how a medicine is actually taken at home, what an instruction was understood to mean, whether an appointment was ever possible to keep, she is the only source there is. A phrase like *declined follow-up* may conceal a whole event: what was offered, what was heard, what got in the way. Beneath this lies a harder question: how to ask about an inherited statement without implying that everyone who wrote it was careless. Baba Farid, some of whose couplets the Guru Granth Sahib keeps, has a shalok that counsels a person of fine understanding not to write black marks against others, but to bow his head and look into his own collar first. Farid's concern is the soul, not the chart. Yet the counsel describes the right posture. The reader who finds a doubtful entry is not outside the chain of copyists. She has copied, summarised and rewritten too, at ordinary moments, under ordinary pressure. The question that does the work is old and unglamorous, *how do we know this?*, and its plural pronoun is the point. Asked with that pronoun, it includes the asker. It separates what was observed from what was reported, what was reported from what was inferred, and all of those from what remains uncertain, and it accuses no one. A correction also needs somewhere to go. A better sentence in today's note does not reach the problem list, the referral already sent, or the record in another system that imported the old one. Requested, entered, communicated and received are different stages, and a correction that stops at the first can leave the patient believing her whole record mended when a single page has changed. None of this argues for a bureaucracy around every trivial discrepancy. It argues that when an inherited statement matters, its correction should be made where the next reader will meet it. ## The Machine That Counts the Copies Software that summarises records makes the difference between counting and weighing more pressing, though it does not create it. Imagine, hypothetically, a summarising system reading the six-document chart. It finds *asthma* in five places, in documents of four kinds by several authors across two institutions. If it treats those appearances as independent support, it may write something like *well-documented history of asthma*, a phrase more confident than any document in the trail, produced by counting what should have been weighed. If it can trace the appearances to a single source, the repetition becomes what it is: a feature of the record's structure, not evidence of a diagnosis. This is an illustration of a design question, not a claim about how any particular product behaves. But the risk has a second stage that deserves attention. Once a clinician reads such a summary and copies it, it appears in a new document under a human name. A later system may then meet its own earlier interpretation as though it were independent confirmation. That is the spinach legend at machine speed: a chain in which the citations cite one another, each pass adding a little confidence and no evidence. A trustworthy summary would reveal three things about whatever it compresses. First, source: where the earliest statement came from, and what kind of statement it was. Second, qualification: whether that statement was a report, an observation or an inference, and how sure its author was. Third, unresolved disagreement: where two accounts conflict, the summary should keep the conflict open until there is a basis for resolving it, because choosing the version that appears more often manufactures clarity without adding knowledge. Smooth prose gives its reader no way of telling whether any of these survived compression. There is a hopeful reversal here. Detecting shared descent is exactly the kind of work machines do well. Identical strings, shared misspellings, the same oddly specific phrase recurring across years of notes: a system could find the textual critic's fingerprints of dependence in seconds, gather the copies together, and show a reader the earliest witness and its line of descent. A tool that did that would be helping its reader weigh instead of count. I would welcome it, and I would judge it by whether it made the record's evidence easier to understand, not by whether it made the summary shorter or more confident. ## The Seventh Document In the hypothetical, suppose the patient answers honestly when she is asked. She isn't sure. It was a cough when she was small, every winter, and a doctor said something about her chest. No one has ever tested her breathing. The physician adds a line: *Reported childhood breathing problems; asthma not confirmed by testing; patient unsure of the original diagnosis.* It is a less confident sentence than any of its predecessors and a more accurate one, and it restores what the trail had lost: a category, a time, a source and an uncertainty. The seventh document in the trail is the first that says how anyone knows. That sentence does not settle whether she has asthma. It settles something else: what the record actually knows, and what it would take to know more. A later reader can now weigh the six earlier documents correctly, as one witness repeated, and can see what kind of evidence would supply a second. If a test is done and confirms the diagnosis, the record will at last contain an independent observation, and *asthma* will deserve its place on the list, this time with a basis a reader can find. The habit I would want any reader of a record to carry, clinician or patient, is small. Before a repeated statement is allowed to grow more persuasive by repetition, ask what supports this particular assertion, and whether the support is independent. Most of the time the answer will be reassuring, and the work of earlier hands will deserve the trust it is given; faithful transmission is how knowledge outlives the people who produced it, and nothing here is meant to disparage it. Sometimes the answer will show that a single uncertain remark has travelled a long way, gaining confidence at every step and evidence at none. The number of documents cannot tell those cases apart. Only their relationship can. Records, like manuscripts, have to be weighed. ## Evidence and Editorial Note **Documented findings.** The essay relies on two facts from the author's public biography: he is a Sikh, and he grew up in Amritsar. Amy Y. Tsou and colleagues, "Safe Practices for Copy and Paste in the EHR: Systematic Review, Recommendations, and Novel Model for Health IT Collaboration," *Applied Clinical Informatics* 8(1), 2017, 12–34: used narrowly for documented risks (redundancy, conflicting information, error propagation), the limits of the evidence on direct patient harm, and recommendations on identifying copied material and making its source available; no prevalence estimate is inferred. Svetlana Z. Lowry and colleagues, "Examining the Copy and Paste Function in the Use of Electronic Health Records," NISTIR 8166, National Institute of Standards and Technology, January 2017: used for human-factors concerns about text volume, attribution and review, and the role of interface design. Both were checked directly and are used only for the narrow points stated; the four-date reading and the other distinctions are the author's analysis, not findings of either report. The 1977 study of repetition and judged truth: Hasher, Goldstein and Toppino, *Journal of Verbal Learning and Verbal Behavior*; later findings that knowledge does not fully protect against the illusory truth effect: Fazio and colleagues, 2015. The spinach decimal-point story, Mike Sutton's 2010 finding, the unsourced 1981 *BMJ* article by Terence Hamblin, and Ole Bjørn Rekdal's 2014 analysis in *Social Studies of Science*: Inside Higher Ed and FiveThirtyEight. Circular reporting: standard reference usage. Textual criticism (the method associated with Karl Lachmann; Sebastiano Timpanaro's study of its origins; Paul Maas, *Textkritik*, 1927; the elimination of derivative witnesses; shared errors as evidence of common descent; *lectio difficilior potior*; the maxim that manuscripts are weighed, not counted; Giorgio Pasquali's caution that later manuscripts are not necessarily worse): standard scholarship, stated from standard knowledge. The Goindval pothis, the compilation of 1604 installed in the Harmandir Sahib with Bhai Gurdas as scribe, Gurinder Singh Mann's manuscript studies, and Professor Sahib Singh's work on the grammar of Gurbani: standard Sikh history and scholarship, stated in general terms. Borges, "Pierre Menard, autor del Quijote" (1939), its plot and the passage from Cervantes (*Don Quixote*, Part One, chapter 9) quoted in Spanish: standard knowledge of the texts. The Archimedes Palimpsest (tenth-century copy, overwritten as a prayer book completed in 1229, identified by Heiberg in 1906, sold in 1998, recovered by imaging, sole surviving copy of *The Method*): standard history of the manuscript. The scribal colophon recording copyist, date and sometimes exemplar: standard manuscript studies. The paper-chart convention of single-line correction: standard documentation practice. The HIPAA Privacy Rule's right to request amendment and to submit a statement of disagreement (45 CFR 164.526): stated in general terms from standard knowledge; this is not legal advice. The shalok of Baba Farid on black marks and the collar: described, not quoted; standard translations. **Vendor representations.** None. Summarising software is discussed generically, and no claim is made about any product's behaviour. **Recollection and tradition.** That the author is a Sikh who grew up in Amritsar is part of his published biography; the remark about the tradition he grew up in is a general statement, not a remembered scene. The account of the scripture's manuscript transmission is Sikh history and tradition stated in general terms. The patient, the questionnaire, the six documents, the man who fell, the woman who lives alone, the summarising system and the corrected entry are hypothetical and illustrative; they do not describe a real patient, a real record, an actual chart error or any real dispute. **Analysis and forecast.** The observation that qualifiers tend to fall away as a statement is copied is a reading of the hypothetical example and an analogy with manuscript copying, not a measured finding. The application of textual-critical principles to clinical records, the four dates, the reading of Farid and Borges, the standard of proportionate scrutiny and the suggestions for correction and for summarising software are the author's interpretation and preference as of November 6, 2026. **Independence.** This essay is not sponsored by any institution, company, health system or technology developer. It is personal reflection, not medical advice; readers seeking care should contact their own physician's office. ## Related reading - [The Note That Grew Longer and Said Less](https://personal.kpsgill.com/essays/the-note-that-grew-longer-and-said-less/) - [Wrong Beautifully](https://personal.kpsgill.com/essays/wrong-beautifully/) - [What the Chart Cannot See](https://personal.kpsgill.com/essays/what-the-chart-cannot-see/)