Chapter 13. The Findings Beside the Earlier Accounts

This chapter sets the ledger's findings beside the accounts of epidemics, numbers and decisions that the earlier chapters map. The findings are those Chapter 12 reports. The accounts are the works this study reads, used only at the level their notes give. The records bear on an argument when a case's own text holds what that argument asks about: a dated pair, a count with a scope, a class who bore a cost, a condition set in advance. They cannot bear when a cell is unsettled, when the ledger's own record field is empty, or when the argument concerns what an organisation thought rather than what an archive holds. Each section names both sides.

13.1Rosenberg's dramaturgy beside the three times

Chapter 3 sets out this dramaturgy beside the hour of a first act. The three times cut the same material at three points. The time of the act is when a public body closes a gate, orders a count, tries a remedy, notifies or declares. The time of the evidence is what was known about cause or effect at that hour. The time of the count is when the dead or the cases were tallied and published. Each is also a field in the ledger: act before proof, gap named, count. The findings bear on the dramaturgy's first act more than on the rest.

Synthetic scene. Country people crowd into the city in the heat. Physicians come to the sick and are the first to die; no remedy or human art avails. The dead lie unburied and the customary rites are thrown into confusion, and fear of gods and law ceases to restrain, since no one expects to live to be punished. What is waited for: the end of the pestilence, which no one can date, and a recognition that has not yet been made.

Rosenberg's progressive revelation describes recognition that comes late and reluctantly. The study's own reading finds it at the lenient reading only. Among cases that show a public act, the strict reading has 8 in support, 3 against, 61 unclear and 17 not applicable, a share of 73%; the lenient reading has 25 in support, 5 against, 20 unclear and 39 not applicable, a share of 83%. Chapter 12 gives the strict verdict as unclear, and says why: few cells were both found stated by the codings. Read across the three further families that read the same records, the strict verdict stays unclear, with 11 in support, 2 against, 37 unclear and 39 not applicable, and the lenient verdict supports, with 19 in support, 5 against, 6 unclear and 59 not applicable. Among the twenty-three cases read closely, thirteen carry the answer yes, five unsettled, four not said and one no. The dramaturgy's second act, the framing of an explanation that manages randomness, sits mostly outside this study's fields. The ledger asks what a case's own text contains; it does not ask what the actors believed, and it carries no cell for an explanation's shape. Rosenberg's note itself distinguishes two explanatory styles, in which disease is read as configuration or as contamination; neither is codeable from these records, and this chapter does not try to code them. What the records can carry is the moment before any explanation holds. Thucydides states that he leaves the plague's origin and causes to other writers, lay or professional, and confines himself to its nature and symptoms; and of the physicians he writes: "Neither were the physicians at first of any service, ignorant as they were of the proper way to treat it, but they died themselves the most thickly, as they visited the sick most often;" Chapter 4 treats the identification of past diseases, where a cause was settled centuries later and where the accounts of Morens and Littman, of Papagrigorakis and colleagues and of Shapiro and colleagues disagree. The record for the plague at Athens keeps the testimony apart from any genetic identification, which is what its own answer states.

The third act, the negotiation of a collective public response, is where the ledger meets the dramaturgy most directly, since restrictions and declarations are the study's read cases, treated in Chapters 7 and 8. Among the twenty-three cases read closely, six acts are notices, declarations or certifications and five are restrictions of movement, work or assembly; four are counts or published reports, three are institutions, compilations or gifts of medicines, three are trials or treatments, one is other and one unsettled. The reason cell agrees in 87 of the 89 cases; in 80 the reason is stated as recorded, in 2 implied, in 4 read as absent and in 1 not applicable, with 2 unsettled. The records carry the reason for a restriction more often than the series expected. The dramaturgy's last two acts, subsidence and retrospection, this study cannot reach. Its records end where the case ends; none holds an account of a society watching an epidemic fade. The retrospection present in them is the archive's later reading, which is what Chapter 11 shows for the field asking what would change the reading of an act: at the lenient reading 10 in support, 18 against, 22 unclear and 39 not applicable, a share of 36%, and at the strict reading 9 in support, 14 against, 49 unclear and 17 not applicable, a share of 39%. Thucydides writes that he sets down the plague's nature so that a student may recognise it if it should ever break out again, and says of his own purpose: "I shall simply set down its nature, and explain the symptoms by which perhaps it may be recognized by the student, if it should ever break out again." The ledger holds no such stated purpose for the other acts.

One place where the dramaturgy fits badly should be said plainly. It presumes a response narrative, an epidemic that a society publicly answers. On the agreed reading, 36 of the 89 cases show a public act on disease or health, 39 carry no such act and 14 are unsettled. Nearly half of the ledger's material stands outside the dramaturgy's form. The dramaturgy remains a good account of how an epidemic is made sense of; it is not an account of what a public archive keeps of the acts themselves.

13.2Turner and Vaughan beside the gaps between event and notice

A synthetic scene.

Hands and eyes: the company's own physician keeps records from experiments on cats, which are not made public; researchers at Kumamoto University work through the town's cases in 1959 and trace the illness to organic mercury discharged from the Chisso factory. What is waited for is an official statement of the cause, and it comes only in September 1968, after the factory has stopped the process that discharged the mercury. Through that period the people who fell ill wait for recognition and for compensation, by protest and by lawsuit.

Vaughan's The Challenger Launch Decision (1996) shows the same shape inside NASA. Chapter 1 maps both works in the fourth literature. Both works are read here only at the level their reading notes give. The codings agreed in 76 of the 89 cases: 39 stated a named gap, 34 not named, 3 not named at the implied reading and 13 unsettled. Among the twenty-three read cases, 13 name a gap, 8 do not and 2 are unsettled. The hypothesis about gaps in time finds support at the lenient reading, 21 in support, 13 against, 16 unclear and 39 not applicable, a share of 62% among cases with an act, and is unclear at the strict reading, 21 in support, 11 against, 40 unclear and 17 not applicable, a share of 66%. This is among the steadier cells of the ledger: across the further readings, agreement on it runs from κ 0.54 to κ 0.80 against the two codings, and the two further readers, where both read a case, agree at κ 0.87.

The fit is real in one direction. Turner and Vaughan both explain how a gap comes to exist; the ledger records only whether a text holds a dated pair. Where the records bear is the second half of their accounts, the reconstruction after the event. The cases that name gaps are largely cases where the gap runs between an event and its official recognition or settled cause: twelve years between a notice of unexplained cases in 1956 and a government's view of the cause in 1968; years between the last case of smallpox and the declaration of 1980; years between a last wild case in 1991 and a certification in 1994. Turner's point is that reconstruction is what an inquiry does afterwards. These records show that the reconstruction is sometimes already inside the archive a later reader holds.

The fit fails in the other direction, and the reason is structural. Turner and Vaughan are both about the internal life of an organisation: what people saw, what they minimised, what the rules allowed them to ignore. The ledger's records are public histories. The ledger's own check finds a named decider in the record's own field in 0 of the 89 cases, and the coders divided on the decider cell in 36 cases; of the remaining 53 cases, 30 read as unknown and the other 23 read as institution or state, a team or organisation, or a multilateral body, some at the stated and some at the implied reading. There is no cell in these records for which warning sign was discounted, or by whom. The series' argument about warnings cannot be tested here. What this section holds is Skeptron's reading of a public source, made once for this study, and the facts its reading notes give.

There is a second place where the records cannot bear. Turner describes warning signs accumulating before an event and Vaughan describes danger signals coming to be treated as normal; this book reads their implication as one that a decision should be revisited when signs accumulate. Neustadt and Fineberg make the same demand elsewhere in this study; here the point is that neither confirmation nor denial is available from these cases. The gap-named cell supplies neither: a record that sets a 1956 notice beside a 1968 recognition shows that two dates stand apart in the archive, and shows nothing about whether anyone in 1957 read the gap as a warning. The reading notes say what a government and a university did across the years; they report no body's view while the gap was open. What would change the reading of an act is the field Chapter 13.4 takes up.

One thing the records do show that the two works might recognise. Turner describes rigid beliefs; Vaughan describes rules that normalise a deviation. Neither argument predicts that a public history record will state the reason for an act, and where these records do state it, the discounting they describe would leave no trace in it. The silence of a record is not evidence of what an organisation failed to see, and this chapter draws no such inference.

13.3Porter and Desrosières beside the counts

Chapter 5 sets out both works and where Graunt's bills and the report of 1858 bear on them.

A synthetic scene. A searcher, an ancient matron sworn to her office, walks from her parish to the place where a dead body lies, looks at it, and writes down by what disease or casualty the corpse died; a parish clerk carries that return into a weekly bill. In Oxford in 1665 a printer sets a fourth edition of a reader's observations of those bills, and the reader criticises their reliability in the same pages where he draws regularities from them. A commission on the sanitary state of the army receives tables and diagrams comparing the mortality of soldiers at home, abroad and in the Russian war with that of civilian men of the same ages; a reprint of 1858 carries the material to a reading public under a title page that names no author. What is waited for, in both scenes, is a total that can be set against another: a denominator, a scope, a period, and the date on which the total stands.

Desrosières's convention of equivalence is the finding's best fit. The qualifications the ledger records are the decisions he describes: which cases are of the same kind, on what scope, over what denominator. The army and civilian mortality comparison of 1858 turns on the denominator that makes the two figures comparable, yet the count cell for that case records no count at all, which is what both codings read there. The Ebola tally of August 2014 includes probable and suspected cases, which the record's own answer states; the Korean MERS total joins counts from two countries. In the Ebola tally and in the Korean MERS total the qualification is inside the case's own text, and those records bear on Desrosières directly: they show the equivalence being made, or its scope being stated. The agreed values stand as they are given there: no-count 48, no-count at the implied reading 14, unsettled 11, qualified by definition, scope or denominator at the stated reading 7, settled at the stated reading 5, no-count at the stated reading 3, and provisional or revised at the stated reading 1.

Porter's question is only partly answerable here. He asks what makes a count travel, who vouches for it and why a distant reader accepts it. The ledger's cells carry the shape of a qualification, not the negotiation behind it; there is no field for who vouched. Where the records bear is the publication side: the bills of mortality that London's parish clerks compiled from the searchers' reports, which Graunt read and criticised, and the reprint of 1858 that carried the army's tables to a reading public. Chapter 5.2 and Chapter 5.3 treat both. Graunt's own text shows the vouching at its source, in the searcher's office: "The Searchers hereupon (who are antient Matrons, ſworn to their Office) 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." And it shows the warrant he found there wanting: "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;". The 1858 material carries a different kind of warrant, recorded in its reading note rather than in its pages: the title page names no author, the work is held in Nightingale's collections and associated there with Florence Nightingale and with the statistician William Farr, and the material was prepared for the Royal Commission on the sanitary state of the army. In these two cases a count travels by a chain of persons and offices; the ledger has no cell that holds such a chain.

The study's own hypothesis E2, that a case's own text qualifies its count rather than giving it as settled, wherever a case gives a count at all, was registered before coding. It returns unclear at the strict reading and support at the lenient: strict, 8 in support, 5 against, 25 unclear and 51 not applicable, a share of 62 per cent; lenient, 8 in support, 5 against, 11 unclear and 65 not applicable, a share of 62 per cent. Group by group the ledger gives the hypothesis support in the medicine history cases at both readings, 6 in support, 2 against and 4 unclear, a share of 75 per cent; it returns unclear in the disaster group at both readings, 1 in support, 2 against and 13 unclear at the strict reading and 1 in support, 2 against and 4 unclear at the lenient, a share of 33 per cent; and unclear in the industrial group, 1 in support, 1 against and 8 unclear at the strict reading and 1 in support, 1 against and 3 unclear at the lenient, a share of 50 per cent. On this cell the chapter model's two codings agree at 88 per cent, κ 0.69; the further readers agree with them at κ values running from 0.55 to 0.74, and among themselves at κ 0.62, 0.79 and 0.80 on the cases both read. Agreement is not accuracy: readers can agree in error.

Where the fit is bad should be stated. Desrosières argues that statistics are bound up with the making of states and that counting rests on conventions of equivalence deciding which cases are of the same kind. The ledger's counts sit in reports, certifications and provisional totals, each qualified differently, and none of these records says why the classification was chosen. Chapter 5.6 sets out the three biases of counting during an outbreak, and Chapter 10 applies them to the two count cases. The counts in these records are what a later reader was given, and the ledger reports what they hold.

13.4Neustadt and Fineberg beside the conditions

Chapter 11 treats both works beside Koselleck's experience and expectation, and what those works demand is that a decision state in advance what would change it.

A synthetic scene. The scene below is this study's, assembled from the reading note on Richard E. Neustadt and Harvey V. Fineberg, The Swine Flu Affair: Decision-Making on a Slippery Disease (1978). Place: the United States in 1976, around the post at Fort Dix where a swine influenza had been found, and the federal department that was to run an immunisation programme against it. Roles: the Secretary of Health, Education and Welfare, to whose office the study was written; the officials and advisers who shaped the programme; and the people who would be inoculated if it went ahead. The programme went ahead, and no pandemic followed.

This study's fifth axis and its last field come closest to the demand those works make. The ledger records no conditions at all: cases with a condition for revisiting the judgment, 0 of 89, and conditions in all, 0. The study's own hypothesis on that field was predicted to hold. It goes against at the lenient reading, 10 in support, 18 against, 22 unclear and 39 not applicable, a share of 36%; at the strict reading it is unclear, 9 in support, 14 against, 49 unclear and 17 not applicable, a share of 39%. Chapter 12 reports the verdicts; this section reads the two works against them.

The apparent fit dissolves on reading. Neustadt and Fineberg's condition has to be set before the outcome, observable, and capable of changing the decision. The conditions the ledger finds named are named after the outcome was known: Chapter 11 shows that the text naming them is Skeptron's reading of 2026, not the actors' record of the hour. A condition set after the fact is a record of what happened next, not a point of reconsideration. Where the ledger seems to contradict their argument, the contradiction is with the prediction this study registered, not with their claim. Neustadt and May, in Thinking in Time: The Uses of History for Decision Makers (1986), make a neighbouring demand. The same gap stands in the way of that demand here: these records hold no story of an issue, only dated acts, and the ledger's deadline cell, present in 88 of the 89 cases, is a date an act waits for rather than a condition that would reverse it.

The cases where a condition was set in advance are the certification cases, and they sit outside the ledger's fields. These are conditions set in advance, observable, and capable of being failed. Chapter 8 and Chapter 11.3 treat them: the records hold the dates, the notes hold the criteria. This study reads those criteria as Koselleck's horizon of expectation written into a public instrument — the future anticipated now, in the terms the note gives the phrase — while the ledger, which counts a condition only where a case's own text sets one for revisiting a judgment, counts none.

Where Neustadt and Fineberg also fit badly: their study is of one programme's decision process, traceable from announcement to inoculation, written for the office that ran it. The ledger's cases are histories of acts at a distance, with no process record and with no decider named in the record's own field in any case. Their argument concerns decisions; this ledger concerns records. A record can show that an act was dated, that its grounds were stated, that a count followed it and that a deadline governed it; it cannot show what was weighed before the act, which is what their method turns on. The certification cases come nearest, and even they reach the reader through a commission's published criteria rather than through a decision's deliberation. What this section adds beside Chapter 11 is the comparison itself: the demand the works make and the count the records return are not measurements of the same thing, and the ledger's zero is a statement about what an archive holds, not about whether any decision was ever revisited.

13.5Scott, Foucault and who waits

Chapter 5.4 treats conventions of equivalence and legibility beside the army mortality comparison. Chapter 6.3 sets out where Foucault's model fits and fails this study's acts.

A synthetic scene. What is waited for: service at the wages customary before the pestilence, and the judgement of those who refuse it, which the ordinance makes punishable.

The bearer cell's codings agreed in 52 of the 89 cases: 38 read as unknown, 37 unsettled, 8 those nearest the stoppage at the implied reading, 5 third parties or the vulnerable at the implied reading and 1 stated. The book's own bearer field agrees in 70 of the 89 cases: 40 not applicable, 19 unsettled, 18 not named as stated, 3 not named as implied, 6 named as stated and 3 named as implied. Among the twenty-three read cases, 13 do not name the bearer, 5 are unsettled and 5 name them. The hypothesis that most acts do not name who bore the wait is supported at the lenient reading, 21 in support, 9 against, 20 unclear and 39 not applicable, a share of 70% among cases with an act, and unclear at the strict reading, 17 in support, 6 against, 49 unclear and 17 not applicable, a share of 74%.

Scott fits where the records show a register being made. In Seeing Like a State (1998) states make society legible by simplifying it into standard names, maps, measures and registers, and schemes that impose such simplifications over local, practical knowledge, which Scott calls mētis, have failed, sometimes disastrously. The counts, notifications and declarations of this study's read cases are legibility work in that sense: a category is fixed, a population is counted, a date is published. The ledger can show that a classification exists and how a count is qualified; it cannot show what local knowledge the classification displaced, because no case in this ledger records what a community knew and did. Paul Richards's account of how communities in Sierra Leone learned how Ebola spread and changed their own practices, set beside the debate on the justice of restriction in Chapter 6.6, is the nearest thing to mētis in this study's readings, and it is a work's argument rather than a cell in the ledger. Where the fit is bad is stated here: Scott's argument concerns what a classification does to the knowledge it overrides, and this ledger holds only the classification, its scope and its date.

Foucault's model fits differently. The household level is the level his account needs, and that is where this study's records part from it. What the records can show is that a cost fell on someone, and the class the records use to name who bore it: survivors ordered to serve at the wages customary before the mortality, sick refused entry, or patients moved. Neither the Florence record nor the Ordinance of Labourers of 1349 holds the household list, the daily inspection or the register passed upward, and this study draws no such inference from their absence. The records cannot show how a cost compared with what an act prevented, and this study does not weigh it.

One further limit stands on the section. The place cell, whether a decision rests with a person, a machine or an automatic rule, agrees in 37 of the 89 cases, with 52 unsettled, 33 human-decides at the implied reading, 2 ai-or-rule-decides at the implied reading, 1 human-decides at the stated reading and 1 unknown. The human role is specified in 62 of the 89 cases, and in all 62 the answer is that the role is not specified, 58 at the implied reading and 4 at the stated. Both readings ask what the archive names rather than who held authority in the room, and where a cell is unsettled this chapter reports the settlement as unset. So the question of who waits is, in this ledger, a question about what an archive holds. Chapter 6's reading of the plague orders and of the 2017 inquiry announcement is where the burden is shown case by case; this section adds only the shape of the two accounts against which that showing is read.

13.6Snowden, Harrison, Beck and Jasanoff: epidemics as mirrors, risk as definition

Chapter 3.5 sets out Ackerknecht's thesis and Baldwin's answer beside Ragusa and Florence. Chapter 8 treats the declaration cases with Gostin, Fidler and Wilder-Smith and Osman.

One of the declaration cases this section discusses is the World Health Organization's characterisation of 11 March 2020. Skeptron's records hold one case on COVID-19, the World Health Organization's statement of 11 March 2020; lockdowns, vaccines and national responses are not in them, and this study does not use them as evidence.

Harrison's argument bears where a record shows a restriction whose cost to trade is visible in its own terms. Chapter 3.3 sets out Ragusa's decision of 1377 and the history of the trentino and the later forty days; the record's answer fixes the one month and the vote. A merchant republic choosing separation before any theory of disease is the strongest case this ledger has for Harrison's bargain. What the ledger cannot show is the argument behind the act: the record carries no commercial objection, no negotiation and no named proposer. Harrison's tension appears in these cases as a consequence rather than as a documented process.

Beck and Jasanoff meet the declaration cases, which are acts of definition under uncertainty: the entry into force of the WHO Constitution in 1948, the declaration on smallpox of 8 May 1980, the certification of polio in 1994 and the characterisation of 11 March 2020. The ledger can date the act and the count that follows it, and Chapter 8 keeps them apart. Beck's question of who defines a risk, and on what evidence, is one this ledger cannot answer; what it can show is that a definition act is dated, and that the evidence and the count are dated separately.

Snowden's mirror question, what an epidemic shows about the society that lived through it, this ledger answers only sideways. The groups are 17 medicine cases, 47 disaster cases and 22 industrial cases, and Chapter 12.5 sets out what the findings look like in each. The pattern worth noting here is that the group with the clearest acts, the medicine group, whose acts are restrictions, trials and compendia, is also the group where the hypotheses reach a verdict most often, and the group where the prediction on E4 fails most clearly, at 20% on the strict reading and 21% on the lenient. At the lenient reading, acts before proof hold in that group with 11 in support, 2 against and 4 unclear, a share of 85% as the ledger's table gives it for that group, while the disaster group returns unclear at both readings and the industry group returns unclear at the strict reading and support at the lenient. The count argument behaves the other way: the hypothesis that a count is qualified rather than settled returns support in the medicine group at both readings, 6 in support, 2 against and 4 unclear, a share of 75%; it returns unclear in the disaster group at the strict reading, 1 in support, 2 against and 13 unclear, a share of 33%, and unclear in the industry group at the lenient reading, 1 in support, 1 against and 3 unclear, a share of 50%. A society's structure shows itself in these numbers as a property of the archive: which acts are recorded, which are counted and which are left as dates.

The six accounts, and where each stands on this ledger:

The table summarises this chapter's diagnosis only. It assigns no weight to any account and judges no act. The conclusion follows.