# When Medicine Was Mostly Memory

*Questions that died in the room, the slow search, and what a trained memory guarded*

Kanwar Partap Singh Gill, MD · November 7, 2026 · Becoming a Doctor

The word in this title that deserves the most attention is the small one. *Mostly.* A phrase like "when medicine was mostly memory" invites a familiar picture: the old physician who carried the whole of medicine in his head, who never needed to look anything up, and who was, by implication, more thoroughly a doctor than the people who now consult a screen between patients. The picture is flattering to the past and slightly insulting to the present, and I do not think it is true. Medicine before search was not a profession of walking encyclopedias. It was a profession in which most of the questions that arose during clinical work were answered from memory, or by asking someone, or not at all, and in which the third option was far more common than anyone liked to admit.

The place and period I have in mind are specific. I studied medicine at Sri Guru Ram Das Institute of Medical Sciences and Research in Amritsar from 1997 until I took my MBBS in 2002. I will not reconstruct a ward, a teacher or a particular night from those years; what I can honestly offer is a description of how clinical knowledge was generally arranged in teaching hospitals of that period, in India and elsewhere, and of what the evidence says about the arrangement. The years matter because they sit at a hinge. Public internet access had reached India in 1995, and I have written elsewhere about [what learning was like in that first year](../where-knowledge-lived-amritsar-1995/index.html). By 1997 a physician in Bethesda or Boston could search the indexed medical literature from a desk. That did not mean a house officer in a provincial teaching hospital could do so at a bedside at two in the morning. The existence of a tool somewhere and its presence in the room are different facts, and the gap between them is where most clinical life is lived.

So this essay is less about memory as a mental faculty, which I have treated in [a separate essay on the mind as the first database](../the-mind-as-the-first-database/index.html), than about memory as part of an information system: one component among several, with its own speed, range and characteristic failures, working alongside books, colleagues, paper and the patient's own recollection. I want to know what that system actually did with a clinical question, and what happened to the questions it could not answer.

## The Instrument at the Speed of Speech

Every clinical instrument has a range of operation. A stethoscope works at the distance of an arm. A thermometer needs a few minutes of contact. A laboratory test works at the speed of a courier, a centrifuge and a technician's shift. Memory has a range too, and its distinctive property is that it is the only reference that works at the speed of conversation.

A consultation is a conversation with a clock running. In 2017 a team from the primary care unit at the University of Cambridge, led by G. Irving, published a systematic review of primary care consultation length drawn from 179 studies covering more than twenty-eight million consultations in sixty-seven countries. The average ranged from 48 seconds in Bangladesh to 22.5 minutes in Sweden, and the authors estimated that eighteen countries, together holding about half of the world's population, gave patients five minutes or less with their primary care physicians. The review was not about memory at all, but it describes the conditions under which memory works. In a five-minute consultation there is no interval in which to consult anything else. Whatever the physician does not carry into the room does not enter it. A question that arises at minute three must be answered by minute four from what is already in the head, or deferred, or dropped.

This is why it is a mistake to think of clinical memory as a library that happens to be stored internally. It behaves more like an instrument used in real time, and like any instrument it is used selectively. Not everything remembered carries equal clinical weight, and the memory of a working physician in that era was, in practice, organised by urgency more than by subject.

There was, first, what might be called the danger list: the presentations that can kill quickly and must be recognised at once, before anything can be looked up, because by the time it has been looked up the moment has passed. Then came the common: the conditions and drugs met many times a week, whose ordinary doses and ordinary courses had to be available without effort or the day would never end. Then the sequences, procedural knowledge of the kind that lives partly in the hands, the order in which things are done when a patient deteriorates. And finally the mass of detail that was learned for examinations and seldom used, which memory kept in the loose and decaying way it keeps anything unrehearsed.

These four kinds of remembered knowledge were not equally reliable, and an experienced clinician knew it, though rarely in words. The danger list was rehearsed by fear and stayed sharp. The common was rehearsed by repetition and stayed accurate as long as nothing changed. The sequences were rehearsed by doing. The rest faded, and the fading was silent.

Punjabi, my first language, borrows its ordinary word for memory from Arabic by way of Urdu. ਹਾਫ਼ਜ਼ਾ (*hāfzā*, from *ḥāfiẓa*) is the faculty of remembering: someone with a sharp memory has a strong *hāfzā*, and an old person's failing memory is said to have grown weak. The root is the Arabic *ḥ-f-ẓ*, to guard or preserve, the same root that gives *ḥāfiẓ*, a guardian, and also the title of a person who has learned the Qur'an by heart. The etymology is more useful than it looks. It suggests that memory's first job is protective, which is close to what memory actually did in the clinic of that period. Its most valuable contents were not facts held for their own sake. They were guards posted at the points where a patient could be hurt: the drug that must never be combined with another, the finding that must never be dismissed, the symptom that means the clock has started. A memory that held every fact and none of the guards would have been a fine examination candidate and a dangerous house officer.

## Questions That Died in the Room

If memory was the main instrument, what happened when it failed? For a long time nobody measured this, because the failures left no trace. A question that was never answered does not appear in a chart.

In 1985 D. G. Covell and two colleagues published a small study in the *Annals of Internal Medicine* with a quietly devastating title: "Information needs in office practice: are they being met?" They asked forty-seven physicians to report the questions about patient management that arose during half a day of ordinary office practice. The physicians raised 269 questions. The questions covered every specialty and were highly specific to the individual patient in front of them. Only thirty percent of those needs were met during the visit, and when they were met, it was usually by another physician or another health professional. The authors also asked why the printed sources in the physicians' offices were not used, and the answers have the dull ring of truth: the textbooks in the office were old, journal articles were poorly organised, books and drug information sources were badly indexed, the physicians did not know which source would hold the answer, and finding it took time they did not have.

Fourteen years later John Ely and colleagues did something more rigorous. They sent investigators to sit with a random sample of 103 family doctors in eastern Iowa for two half-days each, collecting questions as they arose. The doctors asked 1,101 questions. The three most common kinds were what causes this symptom, what is the dose of this drug, and how should I manage this condition. Sixty-four percent of the questions were not immediately pursued. Of those that were pursued, eighty percent were answered, which tells you that the answers usually existed. The doctors spent, on average, less than two minutes pursuing an answer, and they used whatever print and human resources were within reach. Of 1,101 questions, two led to a formal search of the medical literature.

These were American physicians in American offices, and I would not transfer their figures to an Indian teaching hospital in the late 1990s as though they were measurements of it. The settings differed in almost every respect: the number of patients, the training of the doctors, the shelves, the hierarchy, the telephone. But the studies describe something structural, and the structure does not depend on geography. In any setting where clinical work moves faster than consultation of sources, a large share of the questions physicians raise are settled by memory or by a colleague, and a large share are not settled at all. They are asked silently, set aside, and lost.

What does a lost question look like? Usually it looks like nothing. The physician wonders, for a moment, whether the dose he has written is right for a patient of that weight or that age, decides it is probably fine, and moves on. Usually it is fine. The question has died without a witness, and its death is recorded nowhere, which is why the reputation of the memory era is better than its record. A system that cannot see its unanswered questions cannot count its near misses either. It remembers its successes as evidence of skill and has no file for its luck.

I find this the most important correction to the romantic picture. "Mostly memory" did not mean that physicians mostly knew. It meant that the cost of not knowing was mostly invisible. The physician who never looked anything up was not necessarily the one who knew most. He was, often enough, the one whose unanswered questions had been absorbed most smoothly into the flow of the day.

That is a hard thing to say about a generation of doctors who worked under conditions few physicians would choose today, and I want to say it carefully. The failure was not a moral failure of individuals. The studies point to time and access, not indifference. A question that would take forty minutes to answer, in a day that allowed four minutes per patient, was in practice an unanswerable question, and a doctor who stopped to answer every one would have abandoned the waiting room to answer them. The system decided which questions would die. The physician merely presided over the deaths.

## The Slow Search

It would be equally wrong to imagine that clinicians of that era worked from unaided memory because no references existed. References were everywhere. The problem was their speed.

Textbooks existed, and they were heavy in both senses: in the hand and in authority. A standard textbook of medicine was revised every few years, and its authority attached to the volume, not to the edition, so that a copy several editions old on a duty-room shelf could still be consulted with the confidence its newer successors had earned. Drug indexes existed, pocket volumes listing the brands under which a molecule was sold, its preparations and its manufacturers, revised several times a year by commercial publishers; in a pharmaceutical market where one molecule could appear under many brand names, translating a brand on a patient's old prescription into a drug was itself a memory task, and such indexes did that translation better than they did pharmacology. Departments had their own protocols, typed or handwritten, and a protocol pinned to a wall can stay there long after the evidence beneath it has moved. And there was the library, which held the journals, bound by volume and shelved by title.

The library also held, in many medical colleges of the world, a long row of volumes that most clinicians never opened. *Index Medicus* was begun by John Shaw Billings in 1879, and it was published for 125 consecutive years: a printed index to the world's medical journals, arranged by subject heading, which a determined reader could use to find what had been published on a question. It was a search engine made of paper. Using it meant going to the library during its opening hours, finding the right volume for the right year, looking up the right heading, copying the citations, finding the bound journals on the shelves, and reading. A search of that kind took an afternoon if the journals were held locally and weeks if they had to be requested from elsewhere. The National Library of Medicine's own account of the index's decline is a short history of what changed. Its use fell slowly once the database behind it went online in 1971, fell faster in the 1980s as end users began to search for themselves, and dropped precipitously once that database was free on the internet in 1997. By 2003 the monthly index had 155 subscribers, and the library noted that even in the developing world the demand for it was almost nonexistent. The printed series ended with 2004.

So the idea that answers were once unsearchable is not quite accurate either. Answers were searchable. The search simply cost a day, and a clinical question that arises at the bedside cannot wait a day. Covell's physicians had their textbooks in the next room and did not use them, because the textbooks could not answer at the speed the question was asked. What separates that world from this one is the latency of retrieval more than its existence.

One more reference deserves mention because it was, in many Indian hospitals of that period, the most important of all, and it was not on any shelf. In many Indian public hospitals the outpatient record was not kept by the hospital at all. It was a slip or a card, the ਪਰਚੀ (*parchī*) of ordinary Punjabi speech, issued at registration and carried home by the patient, to be brought back at the next visit together with whatever earlier prescriptions and reports the family had kept. The patient's memory and the patient's plastic folder were the institution's archive. When the folder came, it held what no textbook could: what had been tried, in what dose, with what result. When it did not come, because it had been left at home or lost or never kept, the history began again from what the patient remembered, which might be exact, might be vague, and might be the confident recollection of a relative who had not been in the room. The system's memory was distributed across heads, shelves and kitchen drawers, and only one of those was in the physician's control.

## Asking Upward

Covell's physicians met most of their information needs by asking another physician, and Ely's doctors leaned on human resources as much as on print. Nothing suggests the habit was peculiar to America, and it has outlasted the period. The colleague was the fastest reference available. A colleague could hear a question in its full particularity, with the patient's circumstances attached, and answer it in the same breath. A colleague could also say whether the drug was a good idea at all, which no index of doses could.

Yet the colleague was a reference with social conditions attached, and those conditions shaped which questions were asked.

The first condition was availability. The right person had to be present, awake and willing. At night that might be one senior resident covering several wards, and a junior doctor calculating whether a question was worth the cost of waking him was performing a kind of triage on her own ignorance: this question is important enough to risk his irritation, that one is not. The calculation was rarely made on clinical grounds alone.

The second was hierarchy. Medical training in South Asia, which inherited much of its structure from British models, has long been hierarchical, often steeply so, and in a steep hierarchy a question travels upward at a price. To ask is to disclose that one does not know, and the disclosure is made to the person who writes one's assessment. Some questions therefore were never asked, and some were asked in a form designed to conceal how large the gap was. The answer that came back was shaped to the question that was asked, and so a disguised question could produce a precise answer to the wrong problem.

The third condition was the hardest to see from below. An answer from a senior colleague arrived without a citation. It came with the authority of the person, and the authority of the person was considerable, often deserved and always difficult to question. If the senior said that a drug was safe in a particular circumstance, the junior had no easy way to know whether that statement came from a recent trial, from an old textbook, from a single memorable case, or from a habit picked up from someone else's senior twenty years before. And in a hierarchy, challenging the answer carried a cost that challenging a book never did. One may disagree with a textbook in private. Disagreeing with a consultant is a public act.

Knowledge also flows downward in a hierarchy, and much less often upward. A house officer who had read something newer than the consultant knew had to choose between keeping it to himself and offering a correction to a person who did not ask for one. The second course required tact of a kind that is rarely taught. The result was a system in which new knowledge entered most easily at the top, through the people least likely to be reading the newest material at three in the morning, and least easily at the bottom, through the people most likely to have just been examined on it.

None of this makes the colleague a bad source. Colleagues remain among the best sources a physician has, and the speed and particularity of a good answer from a trusted senior is something no database fully replaces. But a system that depends on asking is only as good as the conditions under which asking is safe. Where a question costs its asker standing, the questions that matter most, the ones that reveal the largest gaps, are the ones least likely to be voiced.

## A Night Without an Answer

Consider, as a hypothetical and not as a remembered case, a house officer in a hospital of that era, on duty overnight in a medical ward. It does not matter which hospital; the circumstances were common to many. A patient whose kidneys are working poorly has developed a fever, and the plan from the evening round is to start an antibiotic if the fever recurs. It recurs. The house officer knows the drug. She has prescribed it many times at its usual dose. She also knows, with the half-certainty that unrehearsed knowledge has, that this drug is one of the many that are cleared by the kidneys and whose dose must be reduced when the kidneys are failing. She does not know by how much.

What are her options? She can give the usual dose and trust that a single dose will do no harm, intending to sort it out in the morning. She can give a reduced dose, choosing the reduction from a vague memory of halving. She can delay the first dose until morning, accepting the cost of an untreated fever for some hours. She can wake the senior resident, who is covering three wards and may or may not know the answer himself. She can walk to the duty room and open the pocket drug index, which may or may not give adjustments for kidney function. The library is closed.

Each of these choices converts the missing answer into a particular kind of cost. Giving the usual dose converts it into risk, borne by the patient. Delay converts it into a different risk, also borne by the patient. Waking the senior converts it into social cost, borne by the house officer. Searching converts it into time, borne by every other patient on the ward who might need her in the next half hour. There is no option in which the missing answer costs nothing; the system's only choice is where to put the cost.

What is interesting about the hypothetical is which piece of memory did the most work. The dose was missing; what she had was the guard: the half-remembered knowledge that a question existed at all. A house officer who had not learned that this class of drug depends on the kidneys would have felt no hesitation, given the usual dose, and slept. Her night would have been easier and her patient's possibly worse. The memory that troubled her was the valuable one. The *hāfzā* had done its job. It had stopped her.

And then, most often, the night would have resolved itself without event. She would have chosen a reasonable course, the morning team would probably have checked and adjusted it, and the patient would have recovered. That is the ordinary outcome, and it is why the cost of the memory era is so hard to see. The question was raised, the answer was approximated, nothing happened, and nothing was recorded except a dose and a time. The system learns nothing from a night like that. Nor does the house officer, unless she goes to the library on her afternoon off and looks it up, which is the only step that would improve the next such night and the one step nothing in the system required.

## The Singer Who Never Changed a Word

There is a further problem with memory as an instrument, and it is subtler than forgetting. Memory changes what it holds while reporting that nothing has changed.

In the 1930s Milman Parry, a young American classicist trying to understand how the Homeric poems could have been composed without writing, went with his assistant Albert Lord to what was then Yugoslavia to record the singers of oral epic who still performed in Bosnia and Montenegro. The most remarkable of them was Avdo Međedović, a singer then in his sixties. In 1935 he dictated, over five days, a version of a traditional song, "The Wedding of Smailagić Meho", that ran to more than twelve thousand lines. He had taken the story partly from a printed version, read aloud to him, which ran to 2,160 lines. In the printed text the opening scene, an assembly of lords, took 141 lines; in Avdo's it took 1,053. When Lord recorded the same singer singing the same song in 1950, it came to 8,488 lines.

Lord's book on this work, *The Singer of Tales*, published in 1960, drew a conclusion that changed the study of epic poetry. The singers were not reciting memorised texts. They were composing in performance, rapidly, from a stock of formulas, phrases that fitted the metre and expressed recurring ideas, and from a stock of themes, the recurring scenes of the tradition. No singer sang a song the same way twice. And yet, in Lord's account, singers would insist with complete sincerity that they did. Asked whether a song was the same one they had heard from another singer, or the same one they had sung last year, they would say yes, word for word, line for line. The recordings said otherwise. What the singer meant by "the same" was the same story, the same essential ideas, the same song as a singer understands songs. He was not lying. He was describing his memory from the inside, where it felt like fidelity.

I find this a better model of clinical memory than any library. A physician who has practised for years does not retrieve a stored text when he decides what to do for a common condition. He composes, fluently, from formulas and themes: the usual dose, the usual investigation, the usual reassurance, assembled in the moment to fit the patient in front of him. The fluency is the expertise. It is what lets a good physician manage twenty patients in a morning without consulting anything, in exactly the way that Avdo could dictate for five days.

But Avdo's song grew, contracted, changed its emphasis and its length between 1935 and 1950, and from the inside it felt like the same song. The clinical equivalent is a habit that drifts without announcing its drift: a dose that has crept, a threshold remembered from an older edition, a practice learned from a senior who learned it from someone else. Each repetition feels identical to the last, because the feeling of sameness belongs to the singer, not to the song. Memory, as I have argued elsewhere, keeps no version date, and the only way to discover that one's own song has changed is to compare it with something written down.

This is where the old system was weakest, and where the slow search mattered most. A physician who never consulted a source could not discover his own drift, because the instrument that had drifted was the same instrument he would have used to check it. The source on the shelf, even an old one, was valuable precisely because it did not change with him. A textbook, whatever its age, sang the same song every time it was opened. Its words did not move with the reader's habits, and among singers that stubbornness is the one virtue a book has that no singer can supply.

## The Case for Friction

A defender of the old arrangement would reply that I have described its costs and not its gifts, and he would have a point. The memory era, on this view, produced physicians who had integrated what they knew. Knowledge carried inside the head is connected to everything else inside the head; it is available for comparison, for analogy, for the sudden recognition that two things learned in different chapters bear on each other. Knowledge consulted on a screen is borrowed for a moment and handed back. The friction of the slow search, the defender would add, was itself a teacher. The physician who spent an afternoon in the library came back knowing the answer and also the shape of the controversy around it, and remembered both for years because they had cost him something. A profession that can look everything up may stop carrying anything, and a clinician who carries nothing has no material from which to notice anything.

Most of this seems to me right, and none of it rescues the romantic picture. Integration is a real good, but a body of knowledge can be integrated around a drifted dose as easily as around a correct one; coherence is not accuracy, as Avdo's songs show. The afternoon in the library was a fine teacher for the few questions that received one, and Ely's two formal searches out of 1,101 questions are a reminder of how few that was. What the defender has identified is a virtue of effort, and scarcity was only one way of imposing effort, a crude one that rationed it by accident. The physician who looks up an answer at the bedside and then, at the end of the day, reads the source behind it gets the integration without the night of uncertainty. Whether anyone will choose that when nothing forces it is a question about habit and time. The age of the technology does not settle it.

## What Retrieval Changed

Retrieval at the bedside is now ordinary. A clinician can look up a dose adjusted for kidney function in less time than it takes to describe the problem, from a reference revised far more often than any printed index ever was. Much has improved, and I do not want to understate it. The house officer of the hypothetical night would today have an answer in a minute, and the drift of habit can be checked against a current source without a trip to a library that closes at five.

So it is natural to expect that the unanswered question has disappeared. It has not. In 2014 Guilherme Del Fiol and two colleagues published a systematic review of studies that had examined the questions clinicians raise during patient care, drawing on seventy-two studies. Pooling the eleven that used similar methods, which together elicited more than seven thousand questions through short interviews after visits, they found that clinicians raised about one question for every two patients they saw. They pursued about half of those questions. When they did pursue them, they found answers about four times in five. The main barriers to pursuing a question were lack of time and doubt that a useful answer existed. And the authors observed that this picture had been fairly stable over time, despite the broad availability of online resources that could answer such questions.

That finding should give pause to anyone who tells the history as a story in which memory was replaced by search. The proportion of questions answered when pursued was high in Ely's Iowa and remains high now. The proportion pursued has risen far less than the tools have improved: about a third of questions in Ely's study, about half in the pooled studies. The barrier was never only the slowness of the library. It was the clock in the room and the clinician's judgement, made in seconds, about whether an answer was worth the interruption. Faster retrieval lowered one cost and left the other in place.

Retrieval has also added demands of its own. An answer that arrives in seconds arrives without the friction that once made its source visible. The Index Medicus reader, by the time he had found a paper, knew its journal, its date and roughly where it stood among other papers on the subject, because he had walked past them. A modern answer can appear on a screen detached from all of that, fluent and current-looking, and the work of asking where it came from and whether it applies to this patient must be done deliberately. The skill of appraisal, which once arrived as a by-product of searching slowly, now has to be practised on purpose. And memory has not been retired by any of this. The question still has to be noticed before it can be searched, and noticing is still the work of the guard.

## Keeping the Guard

I said at the beginning that the important word in the title is *mostly*, and I want to return to it with a slightly different understanding.

Medicine before bedside search was mostly memory in the sense that memory was the only instrument that worked at the speed of the encounter, and so it did most of the work. It was never only memory. It was memory supported by books that could not keep up, colleagues whose answers came with social conditions attached, paper records that the patient often carried home, and an unknown number of questions that were silently set aside. The strengths of that system were real: fluency, speed, a clinician who could act without waiting for anything. Its weaknesses were mostly invisible, because they consisted of answers that were never sought and drifts that were never measured.

What it taught, at its best, was the difference between the parts of memory that should be trusted and the parts that should be checked. The guards deserve trust, because they are what tell a clinician to stop. The fluent compositions of habit deserve periodic comparison with something written down, because they change without saying so. The details learned once for an examination deserve no trust at all until they have been looked up. A physician trained in that era who carries this distinction into the present has something of value that does not depend on nostalgia.

What the present makes possible is a habit that the old system could not support: letting questions survive. A question noticed at the bedside need no longer die there for want of a source. It can be answered in a minute, or, if there is no minute, written down and answered at the end of the day, so that the next occasion finds a better-informed physician. Del Fiol and his colleagues suggested that technology should help clinicians track their questions as well as answer them, and that seems to me exactly right, because the questions are the most honest record a clinician has of where his memory stands.

The best physicians I can imagine from that era, and I do not mean any particular one, were not those who never looked anything up. They were those whose memory was organised as a set of guards, who knew which of their own songs had been sung too long without a text, and who, when the library opened, went. The ਹਾਫ਼ਜ਼ਾ was a guard post. It was never meant to be the archive.

## Evidence and Editorial Note

**Documented findings.** Public biography relied on: medical studies at Sri Guru Ram Das Institute of Medical Sciences and Research, Amritsar, 1997–2002 (MBBS 2002); Punjabi as first language. Primary care consultation length: Irving G, Neves AL, Dambha-Miller H and colleagues, "International variations in primary care physician consultation time: a systematic review of 67 countries," *BMJ Open* 2017;7:e017902 (PubMed abstract, PMID 29118053): 179 studies, more than 28.5 million consultations, 48 seconds in Bangladesh to 22.5 minutes in Sweden, eighteen countries holding about half of the world's population at five minutes or less. Covell DG, Uman GC, Manning PR, "Information needs in office practice: are they being met?" *Annals of Internal Medicine* 1985;103:596–599 (PubMed abstract, PMID 4037559): 47 physicians, 269 questions in half a day, 30 percent of needs met during the visit, usually by another physician or health professional, and the stated reasons print sources were not used. Ely JW and colleagues, "Analysis of questions asked by family doctors regarding patient care," *BMJ* 1999;319:358–361 (PubMed abstract, PMID 10435959): 103 family doctors in eastern Iowa, 1,101 questions, the three commonest generic types, 64 percent not immediately pursued, 80 percent of pursued questions answered, under two minutes spent, two formal literature searches. Del Fiol G, Workman TE, Gorman PN, "Clinical questions raised by clinicians at the point of care: a systematic review," *JAMA Internal Medicine* 2014;174:710–718 (PubMed abstract, PMID 24663331): 72 studies; 7,012 questions in eleven studies; 0.57 questions per patient; 51 percent pursued; 78 percent of pursued questions answered; lack of time and doubt that an answer exists as the main barriers; stability over time despite online resources; the suggestion that technology help clinicians track questions. *Index Medicus*: National Library of Medicine, "Index Medicus to Cease as Print Publication," *NLM Technical Bulletin* 2004 May–Jun;(338):e2 (https://www.nlm.nih.gov/pubs/techbull/mj04/mj04_im.html): begun by John Shaw Billings in 1879, published for 125 consecutive years, MEDLINE online from 1971, decline in the 1980s and after free internet access in 1997, 155 monthly subscribers by 2003, negligible demand in the developing world, end of print publication with 2004. Avdo Međedović and the Parry–Lord recordings: Wikipedia, "Avdo Međedović," for the 1935 dictation over five days of more than twelve thousand lines, the printed source of 2,160 lines read to him, the assembly scene of 141 versus 1,053 lines, and the 1950 version of 8,488 lines; Wikipedia, "The Singer of Tales," for Lord's 1960 book, composition from formulas and themes, and the conclusion that singers never tell a story twice with the same wording. Lord's account of singers sincerely asserting that their songs were unchanged "word for word, line for line" is standard reference knowledge of *The Singer of Tales*, paraphrased. Public internet access in India from 1995 is treated in the linked essay. Punjabi and Urdu ਹਾਫ਼ਜ਼ਾ (*ḥāfiẓa*), its Arabic root *ḥ-f-ẓ*, and *ḥāfiẓ* as guardian and as one who has memorised the Qur'an: standard reference knowledge of the languages. Patient-held outpatient slips (ਪਰਚੀ) in many Indian public hospitals, commercial pocket drug indexes revised several times a year, the dependence of many antibiotics on renal clearance, and the general hierarchy of South Asian medical training: standard reference knowledge, stated in general terms and not as claims about any particular hospital. **Vendor representations.** None. No reference product or publisher is named.

**Recollection and tradition.** The author's training years and institution are his published biography. No ward, teacher, colleague, patient, night or library visit is recalled, and no claim is made about the references, protocols or records of Sri Guru Ram Das Institute or any other particular hospital. The house officer and her patient are a hypothetical illustration, introduced as such; the closing reference to the best physicians of the era is likewise imagined, not a recollection of any person.

**Analysis and forecast.** The description of memory as an instrument with a range, the four kinds of remembered clinical knowledge, the reading of ਹਾਫ਼ਜ਼ਾ as a guard, the argument that the cost of the memory era lay in unseen unanswered questions, the comparison of habitual clinical composition with oral-formulaic singing, the reply to the case for friction, and the conclusion that faster retrieval moved the barrier from access to time and appraisal are the author's interpretation as of November 6, 2026. The American studies are not presented as measurements of Indian practice. **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

- [Rounds in a Teaching Hospital in Amritsar](https://personal.kpsgill.com/essays/rounds-in-a-teaching-hospital-in-amritsar/)
- [The Handwritten Medical Chart](https://personal.kpsgill.com/essays/the-handwritten-medical-chart/)
- [The Professors We Never Forget](https://personal.kpsgill.com/essays/the-professors-we-never-forget/)