This chapter reads the third of the study's three times. Chapter 3 concerns the time of the act, Chapter 4 the time of the evidence; the count has a time of its own. A count is not the same object as the thing counted. Someone must decide which cases belong together under one name, gather the results under that decision, and set them down in print or in an announcement later. Skeptron's records are its own readings of public sources, translated once for this study; the source texts that go with two of this chapter's cases are what their authors wrote at the time. Three cases carry the chapter: an edition of 1665 of John Graunt's readings of London's weekly bills; a museum's account of a hospital at Scutari in the winter of 1854 to 1855; and a reprint of 1858 that compared the mortality of the British army with that of the civil population. This study diagnoses what each record leaves out and what its denominator makes possible. It does not judge any count right or wrong. Numbers about the cases appear only where "The ledger's numbers" provides them, with what they count.
5.1Hacking and Porter
Two readings of what printed numbers do stand behind this chapter, and neither is about epidemics as such.
Ian Hacking, in The Taming of Chance (1990), holds that the categories used for counting shape what is counted. A count, on this reading, is never only a quantity: it is a quantity produced inside a set of names, and the names decide what arrives in the column.
Theodore M. Porter, in Trust in Numbers (1995), is the second reading behind this chapter, and his account is set out below, beside Desrosières's.
Notice: the scene below is synthetic, built from the source text's preface and its Chapter I, §§10–12, and adds no fact to either. A house in a London parish holds a printed sheet, bought at four shillings a year, that Thursday's distribution has brought to the family. A reader turns the sheet to the foot, where the burials rise or fall week by week, and to the week's casualties, so that the sheet may serve as a text to talk upon in the next company. What is waited for is the next Thursday, since on Wednesday the general account is made up and printed at the Hall of the Parish-Clerks; Graunt himself went to that hall and gathered as many of the bills as he could obtain before reducing them into tables. In plague-time the same foot is read as a signal: the rich judge of the necessity of their removal, and tradesmen conjecture what doings they are like to have in their respective dealings.
Skeptron's record for this case carries one date, and it is the date of printing. Its time field is 1665; its Scenario identifies the copy at the Wellcome as the 1665 fourth edition printed at Oxford, with 205 pages and folding tables, and adds a bibliographic note that the work has also been attributed to William Petty in addition to John Graunt's name. What the record leaves undated is everything the printed sheet contains but its own date: the period the tables cover, the years of the burials inside them, the years during which the searchers' opinions were formed, and the moment at which someone decided that a reported death of one name belonged with the deaths already under it. The record's answer states the discipline this section needs:
The edition discrepancy, the attribution note, the period covered by the tables, the mortality classification, and the actual onset time are kept distinct.
The act behind that answer is Graunt's own: confused volumes of weekly bills reduced into a few perspicuous tables, so that one year, season, parish or other division of the city could be compared with another, and printed for correction by readers. What was known when he did it was negative rather than positive. In the source text's Chapter II he writes that "the knowledg even of the numbers, which die of the Plague, is not ſufficiently deduced from the meer Report of the Searchers, which onely the Bills afford;", and he corrects the bill by comparing the plague with other casualties, suspecting that a quarter part more died of the plague than were set down. On Hacking's reading, the categories were under dispute while the table was still in use. On Porter's reading, the tables exist so that a claim about London's mortality can leave the hall and reach a council chamber, a reader or a printer's customer without the sworn woman who walked to the house being present.
Both lenses fit this case badly at one point each, and the reader should know it. Porter's note describes weak or contested experts and bureaucracies under outside pressure, and does not reach a parish apparatus of sworn searchers, a Tuesday return and a clerk's printing. Hacking's note concerns the printed avalanche from about 1820 to 1840, and does not say what happens to a count that a layman publishes against the habits of its own makers. This study therefore reads Graunt not as part of that avalanche but as a case where the printed table, the gathering rule behind it, and the dispute over its meaning were already in place.
What both readings give a reader of this chapter is a way to separate three things that a printed figure hides. There is the rule that decides which cases are alike — undated in every record this study holds, and visible only when someone argues about it, as Graunt does. There is the gathering that executes that rule on bodies, beds and streets — dated in the source text's weekly cycle, Tuesday night to Thursday, and undated in the record. There is the printing that makes the result public — the one date the record carries, 1665, and even that date belongs to a copy in a collection rather than to the years of the dead inside it. Each has its own time, and a reader who holds only the third can mistake the first two for a single fact.
5.2Graunt's bills
Skeptron's record for this case carries one date, 1665, and refuses to let that date do more work than it can.
The relevant copy at the Wellcome is the 1665 fourth edition printed at Oxford and includes 205 pages and folding tables, among other features. A bibliographic note is attached stating that it has also been attributed to William Petty in addition to John Graunt's name.
The edition discrepancy, the attribution note, the period covered by the tables, the mortality classification, and the actual onset time are kept distinct.
The answer names what must stay separate. The publication year of a copy in a collection cannot serve as the observation year of every table inside it; a table covering twenty years of burials and a bibliographic note about authorship are different objects with different dates. On the ledger's cells, the codings left Graunt's Bills unsettled on the count and unsettled on whether a gap in time is named in the case, while Scutari and the army mortality report were settled as showing no count at all. Those readings are in "The ledger's numbers" under the twenty-three cases this study reads.
The source text that goes with the case is Graunt's first edition of 1662, a diplomatic transcription from the printed page in which the long s and the original spelling are kept and the spacing is normalised. It answers a question the record does not ask: not how a table reached a collection, but how a death reaches a table.
In Graunt's account, when any one dies, the fact becomes known to the searchers either by tolling or ringing of a bell, or by a grave being bespoken of the sexton, who corresponds with them. The searchers then act: as Graunt writes, they "repair to the place, where the dead Corps lies, and by view of the ſame, and by other enquiries, they examine by what Diſeaſe, or Caſualty the Corps died".
Their report goes to the parish clerk, who every Tuesday night carries to the Clerk of the Hall an account of all the burials and christenings of that week. On Wednesday the general account is made up and printed; on Thursday it is published and dispersed to the several families who will pay four shillings a year for it. The weekly path is thus fixed at each step by office and by day: bell or grave-bespeaking to the sexton, viewing and enquiry by sworn women, a Tuesday night return, a Wednesday printing, a Thursday distribution.
A synthetic scene can be built only from what the text gives, and it gives a great deal of place. A parish clerk waits at the Hall for the return that a sexton's bell has set in motion. Two women walk to a house to view a body and to ask what the household believes killed it; their report is not a certificate but a sworn opinion formed on the spot. The clerk transcribes a week of such opinions into one general account, and the printer's forme turns them into columns that a family may hold in the hand. What is waited for at each step is the next week's column. Graunt himself went to the Hall of the Parish-Clerks and gathered as many bills as he could obtain, then reduced them into tables so that one year, season or parish could be compared with another.
Graunt did not treat the printed total as settled. He asked what the searchers' report could support, and his own answer was a partial negative, in his own words: "That the knowledg even of the numbers, which die of the Plague, is not ſufficiently deduced from the meer Report of the Searchers, which onely the Bills afford;".
What the bills afford must be added to, he wrote, by reasoning and by comparing plague with other casualties. He suspected that in years of plague a quarter part more died of the plague than were set down, and he tried to prove this by looking at what the other categories had gained while the plague's share shrank. His corrections also rest on judgments about which reported categories are trustworthy: some casualties are, in his phrase, matter of sense — a stillborn child, a man above sixty — while others, consumptions and infants' deaths, are confounded in ways he did not trust. He printed tables of notorious diseases and of casualties, with figures such as apoplexy, dead in the streets, drowned and hanged themselves, so that readers could compare each against the grand total.
Here the reading notes enter. Andrea A. Rusnock, in Vital Accounts (2002), is read in this study for the problem of vouching for a count. Graunt's work is the kind of object that reading prepares a reader for; the note does not mention Graunt among its cases, and this study reads his preface as an early instance of the same problem, which his own wording sets out — that a man who does not profess letters, offering tables and positions for correction, is asking a committee of readers to vouch for work made in a room alone.
What the record leaves unsaid about this case is considerable, and the chapter says so rather than filling it in. No searcher is named; no parish is compared with another inside the record; the number of burials left unregistered by the searchers' process is not in the record, though Graunt's own reading note lists unregistered burials among the things he criticised. The attribution note in the record — that the work has also been attributed to William Petty — is a later bibliographic judgment about authorship, and it bears on the count only by asking who stood behind the tables. A later reader who has the 1665 copy can separate the death, the searcher's view, the clerk's Tuesday return and the printed table. A reader who has only the year 1665 cannot.
5.3Scutari and the army's mortality
Skeptron's two records for this part of the chapter stand apart from each other in a way the study's question turns on. The first describes a hospital; the second describes a table.
Florence Nightingale went to Scutari in 1854 to care for the wounded of the Crimean War. The National Army Museum explains the high mortality rate at the beginning of 1855 alongside improvements to the nursing environment, as well as the sanitary commission's improvements to drainage and ventilation. It is a historical matter that examines the dedication of the individual hero separately from the structural conditions of the hospital.
The third sentence is the record's own instruction, and this chapter follows it. The record sets the dedication of an individual beside the conditions of a building, and does not let the first stand as the explanation of the second. The record also lists what is missing: changes in the composition of the patients and in the denominator of the mortality rate, and contemporary source materials that could distinguish the effects of the several improvement measures. That second item is a question about a count, and it is the one this chapter follows.
The 1858 London Harrison and Sons edition compares the death rates of the British military—domestic, overseas, and the Russian War—against those of civilians. It is a reprint of the Royal Commission report and includes tables and diagrams.
The observation period, population composition, death classifications, and the original report and the reprint are separated. The fact of being held in the Nightingale Collection alone does not establish a sole author.
The source text that goes with this case is a reprint from the Report of the Royal Commission appointed to enquire into the Regulations affecting the Sanitary State of the Army, printed in London by Harrison and Sons in 1858, with tables and diagrams.
The comparison the reprint makes is built, line by line, on denominators. Diagram A sets the Foot Guards against the English male population at corresponding ages. Black lines show the proportion of deaths per annum among 1,000 of the civil population for each five-year band from 20 to 40; red lines show the proportion of deaths per annum among 1,000 Guards of the same ages. The text states the selection openly: the Guards are the very select of the population, and any impediment or disease prevents a man being taken for service; none but the strongest and best-looking recruits are accepted by the Guards. The men rejected when the Guardsman was selected pass back into civil life; the civil population receives those not good enough for the army.
A second correction runs through the same comparison. Men discharged from the service by organic disease go on to die among the civil population, so their deaths are added to the black line and taken out of the red. The text says the Guards thus have a double advantage from their men who die invalided. Diagram B repeats the comparison for soldiers on home service with the same correction; Diagrams C and D set the whole army at home and abroad against Englishmen of the army ages living in healthy districts and against the general male population. In some years, the text says, mortality in the troops is from five to six times greater than in civil life. The diagrams also exclude deaths occurring for the first six months after men return from foreign service, which is another convention, and the text names it as one.
The denominators then change again when the comparison moves to war. Diagram H compares the English male population of the army ages with the soldiers of the British army in the East: for every Englishman of those ages who died at home, nearly twenty-three died in the East, and for every man who died at home from zymotic maladies, ninety-three died in the army in the East. Diagram G, on the classes of disease, reports that more than seven-ninths of the mortality in the Infantry on home stations come from two classes alone, fevers and similar diseases, and chest and tubercular diseases; and that the mortality from chest and tubercular diseases in the Infantry exceeds the total mortality from all causes among the civil population at the same ages. Diagram E restates the comparison in another unit: of 10,000 men between 20 and 40, about 8,500 lives in healthy civil populations survive the ages between 20 and 40, against 6,900 of the army's lives.
Scutari receives its own treatment in the same text, and its denominator is again different. The passage describes the transport ships as over-crowded and badly ventilated at first, and gives the losses: of 13,093 sick shipped for Scutari between the landing in the Crimea and the end of January 1855, 976 died on a passage of only 300 miles, which is 75 per 1,000. Of those who landed, two out of every five were lost among those treated in the hospitals of the Bosphorus during February 1855, and one out of every two at Koulali. The buildings, the text says, had sewers of the worst possible construction, open privies whose pipes ran sewer-air into the corridors and wards, walls needing constant lime-washing, and no means of ventilation; the population of the hospitals was increased while these conditions grew worse. A small diagram gives the annual rate of mortality per cent on the sick population of the hospitals: from the occupation of the buildings in October 1854, with few sick, the rate was already high; nothing was done; then ship after ship arrived, and in February 1855 the sick died at the rate of 415 per cent per annum. Slight improvements began in the beginning of March 1855, but not until the 17th were effectual means put into operation; by the end of June the improvements were nearly completed, and the mortality fell to less than a sixth part of what it had been in October 1854 and to a nineteenth part of what it had been in February 1855. The text's own comment on that fall is a statement about the denominator as much as about the remedy: the men were the same, the conditions only had been altered.
A synthetic scene for the hospital must follow that text exactly. A hospital stands above sewers; the wards hold sick men brought in by ships that have been at sea; the staff exert themselves to the utmost, and incur mortality themselves; the drainage and the ventilation are the objects of the commission's work in 1855; what is waited for, month by month, is the proportion of sick that survive a month of occupancy. The report attributes the fall to the sanitary works begun in March 1855 and gives no separate estimate of nursing's share; the nursing question is this chapter's own, as is the question of the record's silence.
The reading notes on this episode pull in one direction together. Hugh Small, in Florence Nightingale: Avenging Angel (1998), argues that the very high death rate at Scutari in the winter of 1854–55 came chiefly from the hospital's sanitary defects — drains, ventilation, overcrowding — and that the Sanitary Commission sent out in the spring of 1855 began to remedy them; and that Nightingale came to understand this fully only after the war, when she analysed the army's mortality statistics with William Farr, from which time she campaigned for sanitary reform. I. Bernard Cohen, in Florence Nightingale (1984), presents her as a statistician who used statistics as a tool of reform, presenting the army's mortality in graphical forms, among them the polar-area diagram, to persuade officials that most deaths in the Crimean War came from preventable disease. Both notes are consistent with the record's separation of hero and hospital; Small's note adds that the understanding came after the events, which is one of the study's three times made explicit in a secondary account.
The record keeps the hero and the hospital apart, and the denominator decides how far the two can be joined. A rate of 415 per cent per annum on the sick population of a hospital in February 1855 is not the same kind of statement as a rate per 1,000 among Guards between 20 and 40 at home. The first counts men already sick and already inside a building; the second counts selected men against a civil population that has taken back its rejects. The same report carries both, and the same reader can mistake one for the other. The record's own list of unknowns — changes in patient composition, and in the denominator of the mortality rate — names the trap without filling it. On the ledger's cells the codings read the Scutari case as showing no count with the gap unsettled, and the army mortality case as showing no count with the gap not named; these are readings of the records as written, not of the hospital or the report.
5.4Conventions of equivalence
Alain Desrosières, in The Politics of Large Numbers (1998), gives this chapter its vocabulary: conventions of equivalence, which decide which cases count as one kind. Those classifications are at once technical and political. Once established, they hold, and they become things that can be acted upon. James C. Scott, in Seeing Like a State (1998), adds the state's side of the same problem: legibility, reached by simplifying society into standard names, maps, measures and registers.
Applied to this chapter's three cases, the conventions become visible as work that had to be done before any comparison could be printed. For the army and the civil population to be compared at all, several things had to be made equivalent first. Ages had to be banded and matched — the text compares 1,000 men of the army ages with 1,000 civilians of the same ages, five years at a time. Periods had to be aligned: home service in time of peace, the army abroad, the Russian War, and the two years of the Crimea treated as separate objects. Death had to be classified, and the classification itself carried the argument: zymotic diseases against chest and tubercular diseases, wounds against disease, deaths in hospital against deaths on the field. The invaliding convention had to be stated, so that men discharged from the army and dying in civil life were counted in the civil column. Geographies had to be separated — home stations, the East, Manchester selected as a town for comparison — and a base had to be chosen, since the Scutari figure runs per 100 of the sick population and the Guards figure per 1,000 of a selected living group. Behind all of these sat one further equivalence: that a Guardsman and a civilian man of the same age belong, for this purpose, to one population whose life can be measured against the other.
Graunt's tables required the same kind of work at an earlier scale. A weekly bill could only be compared with another weekly bill if "aged", "infants" and "consumption" meant the same thing on both Tuesdays, and Graunt's own chapters show him doubting that they did. His corrections — treating some searchers' categories as adequate because they are matter of sense, and others as unreliable — are conventions of equivalence written down by the person who uses them. Desrosières's note says classifications hold once established; Graunt's text shows a classification under dispute while it is still being used. The two readings are compatible, and this study reads Graunt as a case where the convention is visible because the author argues about it.
Scott's note about legibility fits the army comparison more awkwardly than it fits a state's registers. The 1858 report is not a scheme to improve the human condition by simplifying society; it is an argument, made through simplifications, that a society's own institution is killing its selected young men. Where Scott's lens helps is on the question of what such simplifications cannot see: a table that counts only men still in service cannot see the men discharged from it, and the report's own correction for invaliding is an admission that the register as drawn would have hidden them. Where it fits badly, this study says so; the note does not discuss military mortality statistics, and it would be a misuse to ask it to.
5.5The debate: numbers as trust or as the making of states
The two readings of printed numbers can be set beside each other without inventing an exchange between their authors. Porter's note and Desrosières's note are separate books making separate claims; this chapter places them together because each says something about what a count does beyond counting.
Porter's account makes the rules of counting a substitute for trust. Where experts are weak, contested or under outside pressure, they pursue mechanical objectivity — rules rather than judgment — so that a result may travel to people who have no reason to trust the person who produced it. Desrosières's account makes the rules constitutive: the equivalence conventions are what produce the statistical object, and they are technical and political at once. Scott's account supplies the institutional side — the state that needs a legible society and the simplifications it imposes. Rusnock's note sits between the two, and it names both the gathering and the vouching; persuasion is this chapter's own addition. The note can be read against either Porter or Desrosières without being reduced to either.
Where do this chapter's records bear on the debate, and where do they not? Graunt's bills bear on it at the point where trust and rules meet. The searchers are sworn to their office; their report is the only source the bills afford; the clerk's return, the printing and the sale at four shillings a year are the rules that make the number travel beyond the households who knew the dead. Graunt's own work is the contest over those numbers: a layman reducing confused volumes into tables, inviting correction, and telling his readers that the numbers show things the city's officials should consider. On Porter's reading, the tables let a claim about London's population and its mortality reach a council chamber without the person who walked to the house being present. On Desrosières's reading, the tables are the thing that exists only because "aged", "consumption" and "plague" were made to hold across years and parishes. The record for the case adds a third element that neither reading has yet: an attribution note, later, asking who stood behind the work. A later bibliographic judgment does not change the count, but it changes who vouches for it.
The 1858 report bears on the same debate with a different shape. It is a reprint of a royal commission's report, not the commission's report itself; the record's answer says the original and the reprint must be separated, and that holding the copy in a collection does not establish a sole author. It is also a work of persuasion: diagrams are used, the text says, because ideas about vital statistics reach the eye more readily than figures reach the mind. On Porter's reading, the diagrams are a technology of distance, letting an argument about barracks, drains and hospitals travel from a commission's tables to officials and readers who were not there. On Desrosières's reading, the finding is produced by the conventions — the matched ages, the invaliding correction, the classification of deaths — and the diagrams display those conventions rather than the deaths directly. Rusnock's note again provides the middle term: the trust in the comparison depends on how the numbers were gathered, and the report names its own gathering, its corrections and its exclusions in the course of making the argument.
What the records cannot settle is the question the two accounts actually disagree about. Skeptron's records are readings of what a source says, not transcripts of the argument behind a table; the Scutari record lists missing materials that could distinguish the effects of measures, and the army record separates the observation period, the population composition, the death classifications and the reprint. Neither record says what any official thought of the equivalence conventions, or whether a diagram persuaded anyone. Where this chapter's cases therefore stop short: they show conventions in use and claims made through them, and they cannot show how often such claims were accepted, which is what a Porter-style argument about the social life of numbers would need and what a Desrosières-style argument about the making of objects would need in another form. The verdict on the debate is left to the reader; this chapter reports what each position holds and what the three records can bear.
5.6Counting while it happens
Marc Lipsitch, Christl A. Donnelly, Christophe Fraser and colleagues, in Potential Biases in Estimating Absolute and Relative Case-Fatality Risks during Outbreaks (2015), warn that an outbreak's case-fatality ratio is still being built while it is read: the deaths arrive after the cases, so a figure taken early sits too low; the mildest cases never reach the reports, so a figure built on them sits too high; and no two places or weeks count the same cases in the same way, so a figure from one cannot stand beside a figure from another until all three are allowed for. The note closes with that requirement, the one this study's question keeps returning to.
The note concerns a ratio — deaths over cases — still forming in an outbreak. None of this chapter's three counts is that ratio: Graunt's tables set burials against christenings, the Scutari diagram deaths against the sick population of a hospital, the 1858 comparison deaths against deaths in two populations. The first two biases do not reach them; the third does, and this study reads the three cases as showing what it names — the definition that decides what the count contains.
Graunt's case dates one thing and leaves the rest undated. Its time is 1665, and its Scenario identifies the copy at the Wellcome as the fourth edition printed at Oxford; the tables are the 1662 first edition, and the plague years they compare ran from 1592 to 1636. What the record leaves undated is what the tables hold: the years of the dead inside them, the weeks on which the searchers formed their opinions, and the moment a reported name was placed under a category. Graunt's own text dates the weekly cycle and dates the correction as reasoning after the fact, "in Years of Plague a quarter part more dies of that Diſeaſe then are ſet down". The thing counted had stopped moving before Graunt counted it; what was known when Graunt worked was negative — the searchers' report alone did not deduce the plague's numbers, which Graunt reached by comparing the plague with other casualties. Graunt framed the shortfall as misattribution between categories rather than as missed cases, which stands outside the note's first two biases. Graunt's answer:
The edition discrepancy, the attribution note, the period covered by the tables, the mortality classification, and the actual onset time are kept distinct.
This study reads Graunt's distinction between casualties that are matter of sense and casualties that are confounded as the same problem the note names in its third bias, in the terms of a parish bill.
Scutari's record is dated 1854 to 1855 and leaves its denominator undated. In February 1855 the sick died at the rate of 415 per cent per annum on the sick population of the hospitals. A rate per annum drawn from one month expresses a population that ships were still enlarging as though it were fixed, and the record's own list of what is missing names the problem without closing it: changes in the composition of the patients and in the denominator of the mortality rate.
The 1858 case dates a reprint, which its Scenario identifies as the 1858 London Harrison and Sons edition, a reprint of the Royal Commission report with tables and diagrams. The report dates the conventions as well as the figure: deaths of men discharged by organic disease are added to the civil column and taken out of the army column; deaths in the first six months after return from foreign service are excluded; ages are banded five years at a time and matched. Under the frame's third bias, whose death counts decides what a ratio means, here the definitions are printed rather than silent. The text gives no count of cases — mild or severe — that would turn a death-to-death comparison into a death-to-case ratio. The record keeps the pieces apart:
The observation period, population composition, death classifications, and the original report and the reprint are separated. The fact of being held in the Nightingale Collection alone does not establish a sole author.
None of this makes any of the three counts wrong; each text states, in its own way, the convention it relies on, which is more than a bare figure does.
On "The ledger's numbers", no case carries a condition for revisiting its judgment: 0 of 89, and the conditions-per-case column reads 0.00 in every horizon band that table gives, including centuries or more. The horizon runs to centuries or more for Graunt's case and is unsettled for Scutari and the army report. Every case in the ledger lists what is missing, 264 items in all; a list of what is missing is not a written condition under which a count would be read again. The case-by-case table of the twenty-three reads the revision cell as named for Graunt's Bills and for Scutari and unsettled for Army Mortality 1858. This study reads that spread as the ledger's answer to what the frame asks — what would make a count be read again — and as a sign that naming what is missing and setting a trigger are different acts.
A later reader holding Graunt's tables can tell what convention produced each column, because Graunt argued about it in print, but not how many burials went unregistered, a figure in none of the tables. Holding the Scutari diagram, the reader can tell that the base was the sick population of the hospitals and that the base was still moving, but not how that population was composed; holding the 1858 reprint, the reader can tell that the comparison's conventions were printed and named, but not how often the same convention would have been applied elsewhere, since the text compares two populations. The third bias is answered only as far as the text names its definition; the first two lie outside these counts by construction, because none of the three counts deaths against cases.
Chapter 10 applies the three biases to counts made during an outbreak — the Ebola emergency committee's tally and the Korean MERS report. Chapter 6 takes up the question this chapter circles: when a count arrives late, provisional, or on a denominator that has changed, who is waiting on it in the meantime, and who bears the cost of the delay. The records of this chapter name the conventions; only the Scutari record, as "The ledger's numbers" reads it, names who bore the wait, and Graunt's record and the army mortality record leave the waiters unnamed.