Chapter 11. Conditions for Revisiting: What Could Be Observed

11.1The series' axis and Skeptron's fifth question

The series asks what must be decided now and what would make that decision change. The axis this chapter takes is the one that puts the question to the record rather than to the decision: what would change the decision, and could that change be observed? Chapter 2 sets out the whole set of axes and how each was coded. This chapter works on one of them, and keeps to what the records carry and what a record would have to carry.

A Skeptron case is built from a fixed set of questions. The closing sentences of every fifth question are among the sentences that recur across the cases with only names and dates changed: they are the frame, not the case. Chapter 2 explains why those sentences are Skeptron's method rather than anything any case shows, and records that the coders were told to read each case's own sentences only. Because those closing sentences recur, a count of conditions drawn from them would count the studio's method repeated eighty-nine times, not the contents of any record. The ledger holds the check beside the count: it reports how far a condition's wording would have to reach the studio's frame to be flagged — thirty per cent or more, on its reading — and with no conditions coded that count stands at zero of zero.

What remains of a case's own is what the case says about itself: the Scenario that sets the scene, the Answer that records what the source establishes, the Unknown that names what the source does not show, and the Decision window and Notice where the record has them. Beside these sit an event period written as a date and an as-of date for the studio's reading. Those sentences are Skeptron's reading of a public source, written now and dated now. They are not the words of the actors at the hour. This chapter works from that reading, marks it as such, and at no point treats it as the actors' own record. What such a reading can carry about the hour of an act is what the record leaves standing in its Unknown field: all eighty-nine cases list what is missing, and the items listed across them number two hundred and sixty-four. Where the Unknown names what was not then known, a later reader can see the hour of the act; where it does not, the reader supplies that knowledge from outside.

The record form dates its own materials and leaves the rest undated. It holds an event period, an as-of date, the sources consulted and a set of answers. It holds no moment at which a judgment could be opened again, because no record names a condition that would fix one. The ledger counts that absence in its own fields. The field for a condition of revisiting — what would change the judgment — stands at zero conditions in all eighty-nine records: zero of eighty-nine cases carry one, and the conditions listed across all cases number zero. The field for a decider named in the record counts the same emptiness in another place: zero cases. The two codings, which read the whole record rather than that field alone, settled on who decided in fifty-three of the eighty-nine cases, and thirty of those settled on unknown.

Set beside the condition field is another in the same ledger: a deadline in the record, counted in eighty-eight cases. A deadline names the date by which something is to be done, and its date belongs to the record's own day. A condition for revisiting names an observation that would make the judgment change — an observation no record names, and therefore one no record can date. The two answer different questions, and only the first survives in the record form as a thing to be counted. A form that carries eighty-eight deadlines and no condition can say when something must be done; it cannot say what would make a judgment be revised.

Conditions that the cases' own text does carry are counted in a different field, the study's seventh: what would change the reading, coded in the ledger as "revision in case". The two codings agree on that cell in sixty-seven of eighty-nine cases. Across all eighty-nine the cell reads: not applicable in thirty-nine cases, unsettled in twenty-two, named as stated in fifteen, not named in ten, named as implied in three. The ledger reports how far independent readings of this cell carry against one another, and among the seven fields it records the coefficient for this cell stands lower than any other in each of its further-reading tables. Whether a case's own text names what would change the reading is therefore the cell on which readers divide most: the two codings put fifteen cases in the named values, and those values do not carry unchanged from one reader of the same records to the next.

What could be observed therefore has two halves here. One is the ledger's count: what the records actually carry — eighty-nine listings of what is missing, eighty-eight deadlines, no condition, and a cell on what would change the reading whose values divide between readers. The other is the draft in section 11.5: what a record would have to carry, in terms its own day could supply, for a condition to be checked by a later reader. A condition for revisiting would name the moment at which the judgment could be looked at again; the ledger counts none in any record. The draft diagnoses what is present and what is absent. It does not judge any act against a condition the record does not hold.

11.2Conditions in the cases' own text

The studio's recurring sentences appear in every history case with only names and dates changed; Chapter 2 sets them out as Skeptron's method, and a count drawn from them would count the method repeated eighty-nine times rather than anything any case shows. E4 is asked of what the record fills in its own words.

The ledger's counts go the other way. In the strict reading, nine cases support E4, fourteen go against it, forty-nine are unclear and seventeen fall outside it; the ledger's share stands at thirty-nine per cent. In the lenient reading, ten support, eighteen go against, twenty-two are unclear and thirty-nine fall outside; the share stands at thirty-six per cent. In both readings more cases go against the prediction than stand beside it: the case's own text names what would change the reading more often than was predicted. The verdict on E4 itself is Chapter 12's business; what this chapter asks is what such a condition is, and when it was named.

The group columns divide the cell differently. The medicine series holds seventeen cases and its act cell reads yes in all seventeen; there E4 stands against the prediction in both readings, with shares of twenty per cent strict and twenty-one per cent lenient, and the cell for what would change the reading reads named in eleven cases, not named in three and unsettled in three. The disaster and industrial series do not settle: both stand unclear in both codings — fifty-seven per cent in each reading for the disaster series, sixty-seven and fifty per cent for the industrial one.

Skeptron's records are the studio's readings of public sources, made in 2026 and translated for this study. Their Scenario and Answer sentences are the studio's words, not the actors'. A condition written in such a sentence is named after the outcome was known: it can name what would have had to be observed, because the reader now knows what was observed, and it cannot be what the actors set at the hour, because no condition of the kind appears in the record and because the words carrying it are not theirs. This study reads the named conditions as a later reader's conditions, set to the reading rather than to the act.

One record shows that distance in dates. The Minamata record is fixed at the first notification:

The reading note on Timothy S. George, Minamata: Pollution and the Struggle for Democracy in Postwar Japan (2001), places the tracing of the cause by researchers at Kumamoto University in 1959, the national Government's official recognition of the cause in September 1968, and the company's own physician's evidence, from experiments on cats, that was not made public. The separation between event, evidence and act is legible to a reader of 2026 and was fixed by no one at the hour of the act.

Three kinds of condition appear in the cases' own text. The first separates two events a casual reader would merge, and three records hold that separation in their own titles:

Each time field fixes one of the two dates only. Philadelphia is fixed at 28 September 1918, and the reading note on the 2007 study places the first civilian cases in mid-September and the interventions in early October. The pandemic characterisation is fixed at 11 March 2020, and its title keeps the date of the characterisation apart from the start of the infection it describes. The Life Esidimeni record is fixed at 1 February 2017, and the ledger's count cell carries provisional or revised in stated form in one case of eighty-nine; that case is this one.

The second kind states that a date must not carry the knowledge of a later report. The Broad Street record is fixed at 8 September 1854:

Snow's own text dates his appearance before the Board of Guardians of St James's parish to 7 September and the removal of the handle to the following day, and adds that the attacks had so far diminished before the use of the water was stopped that it is impossible to decide whether the well still contained the poison in an active state. A reader who takes 8 September as the date on which the outbreak was understood has merged the removal of a handle with the establishment of a cause no observation that day could settle.

The third kind gives the interval between an event and its record:

The record is fixed at 6 July 1885; Pasteur's report was read on 26 October 1885. Lind's comparison of twelve sailors began on 20 May 1747, Old Style, and its treatise appeared in 1753; Jenner's inoculation is dated 14 May 1796 and the Inquiry that argues from it was printed in 1798. Each record names two dates for different work: the hour of an act, and the hour at which the act could be read.

None of these sets a date at which an act would be looked at again, and none names an observation that would make the act be revised.

The ledger's own field counts no condition in any of the eighty-nine records. Its table of conditions by horizon lists zero conditions per case in every band it names, from days to months through decades to centuries and more. That table is a statement about the record form, not about the horizons themselves. The records are histories, not decision records, and their form leaves the condition out entirely. What a record form leaves out is not something the cases show, and it is not a pattern among them.

Where the case's own text names a condition, a later reader can see what the archive had to keep apart in order to write its reading, and can ask whether the actors ever saw the separation at all. Where it names none, the reader must supply one from outside the record, and at that point the study's reading becomes the source of the condition rather than a report of the record's contents. Among the twenty-three cases read closely, fourteen carry a named condition, five do not and four the codings could not settle. The five without are the Edwin Smith Papyrus, the Ordinance of Labourers, Lind's twelve sailors, the WHO Constitution and the Ebola committee tally; the four unsettled are the plague at Athens, the Shosoin record, Jenner's inquiry and the army mortality comparison of 1858; the fourteen named run from Florence in 1348 and Ragusa's month apart to the polio certification, the MERS outbreak in Korea and the pandemic characterisation. The field is the study's own; it counts what Skeptron's text says, not what the actors said.

11.3Certification as the clearest condition

Among the acts these records carry, certification holds the clearest conditions and the clearest check on them. Two cases show it: the polio certification of 1994 and the smallpox declaration of 1980. Chapter 8 reads both records on the ledger's cells. This chapter reads only their conditions: what each condition names, when it was fixed, what a later reader can check against it, and what the record leaves outside it.

The polio case carries a public report, Certification of the Eradication of Indigenous Transmission of Wild Poliovirus in the Americas (1997) by Frederick C. Robbins and Ciro A. de Quadros. The record states its act in its own title:

The reading note on that report gives the condition's parts. Each element is checkable: the years that ran, the standard of surveillance named for them, and the laboratory work required of a case found. The condition was fixed before the act it governs, and the report gives its terms.

The smallpox case has no source text available to this study: the World Health Assembly's declaration of 1980 is one of the cases whose original could not be reached, so what survives here is the record and the reading notes. The record names the act:

The notes on Fenner and others, Smallpox and Its Eradication (1988), and on Henderson's Smallpox: The Death of a Disease (2009), carry the condition. Certification came before the declaration. This condition is known at one remove, through this study's reading notes rather than through the document that stated it.

Notice. The scene below is a synthetic scene written for this study from the reading notes and the ledger. It adds no event, number, date or outcome to them, and names no individual, because the notes do not name the members of the commission.

The review room

Place: a room set aside for the review, with returns arriving from national surveillance systems rather than from a single ward. Roles: members of an international commission, and the surveillance staff who prepared the material. Hands: turning pages of accumulated years, checking the date on which a case was found against the laboratory result attached to it. Eyes: comparing one year's returns with another's, and each against the standard the commission was given. What is waited for is not the last case, which the notes place in 1991, but a run of years under surveillance the commission judges adequate. When the review closes, the act it supports is an announcement; the date of the announcement and the dates of the process it certifies are not the same dates.

A condition set in advance has two properties that a condition set afterwards does not. It is observable: a reader can check whether the years ran, whether the surveillance met the standard, whether the search continued and whether laboratory work was done on the cases found. And it is fixed before the outcome: the standard is not adjusted to fit the last case found. Certification conditions are the one place in these records where the governing terms of an act are visible to a reader who has none of the actors' reasoning at all. The terms are written out in a document; a later reader can read them beside the dates and see whether the act met its own standard.

What such a condition does not cover is as plain as what it covers. A certification condition names the years before the act, not the state after it. Neither the report of 1997 nor the notes on smallpox says what observation would withdraw a certification once given. The closing sentences of every fifth question in Skeptron's cases touch on what would withdraw a comparison; those sentences are frame, and Chapter 2 explains why they cannot be read as any record's own condition. A reader can check the search. The record does not say what would reopen the conclusion.

The ledger's own fields let this study read the two cases apart. Both are read as acts taken; both name the gap between event and record; both carry no count of their own; both are read as naming what would change the reading. They differ where the wait and the proof are concerned. On who bore the cost while the years ran, the codings could not settle the polio case and read the smallpox case as naming no one. On whether the act came before proof, the codings read the smallpox record as not taken before the cause or effect it addressed was established — a value that appears once across all eighty-nine cases, read as implied. That difference is what certification is: an act whose governing terms were written down before the announcement, so that a later reader can date them without the actors' testimony. None of this judges the certification or the declaration; it says only what the record and the notes carry.

The asymmetry of the evidence belongs in the diagnosis. The polio condition survives in a report this study can read, and can be checked against what was written at the time. The smallpox condition survives only as a summary in the reading notes, and this study reports only what those notes say. The ledger's field for conditions of revisiting counts none in any of the eighty-nine records: zero cases carry one, and the conditions listed across all cases number zero, with zero per case in every horizon band the table names. The horizon cell is one where the codings divided — agreed in thirty-two of eighty-nine cases, unsettled in fifty-seven — and no claim in this chapter rests on the horizon a record assigns to an act. What certification gives a later reader is not a verdict but a check: a set of terms, a set of dates, and the question of whether the announcement met the standard written down before it.

11.4Points of reconsideration and thinking in time

Each of the three works treated here offers a question that could be put to the acts in these records; none is a verdict on any act.

Neustadt and Fineberg's remedy is: ask what would change the decision, and build points of reconsideration in. A point of reconsideration is written into an act. A deadline is not the same thing: a deadline says when something must be done; a reconsideration point says when something must be asked again, and what would be observed when it was. A reason is not a reconsideration point either: the ledger records a stated reason in eighty of the eighty-nine cases. A reason tells a later reader why an act was taken at its hour; it does not say what would end it.

Snow's case shows what the difference costs a reader. The record fixes 8 September 1854, the day the pump handle was removed. The same text records that the attacks had so far diminished before the use of the water was stopped that it is impossible to decide whether the well still contained the poison. No sentence in the actor's account names what would make the removal be reviewed or reinstated. A later reader can date the removal and the fall in attacks; the record cannot say what would have made either be kept or undone.

One record in this study carries the opposite shape. The instrument's own terms carry a reconsideration of a kind the records rarely hold. Article 21 lets the Health Assembly adopt regulations including "sanitary and quarantine requirements and other procedures designed to prevent the international spread of disease;" and Article 22 provides that regulations come into force for all Members after due notice has been given, "except for such Members as may notify the Director-General of rejection or reservations within the period stated in the notice." (Constitution of the World Health Organization, Articles 21 and 22.) A Member thus has a period, named in the instrument itself, within which it may decline what the Assembly adopts. The record dates the coming into force; it does not date the moment at which each Member accepted it, and that interval separates an act whose reconsideration is written into its own terms from one whose terms a later reader must supply.

The Ordinance of Labourers, dated June 1349, shows what the act presumed and when it was taken; it does not date the scarcity it speaks of, and nothing in it marks when the issue began for the labourers or for the masters. On Minamata the record dates the initial notification to 1 May 1956 and holds the Government's view of the cause from 1968; these two dates stand apart, and neither tells when the issue began for anyone. Placing an act in its own history asks which past the actor stood inside. On the international health that preceded the Constitution's adoption, this study's set holds two readings — Valeska Huber's, which reads the sanitary conferences as a slow and political process in which agreement came late, and Mark Harrison's, which reads them as an attempt to agree on measures protecting health while sparing trade — and neither is a record of what the drafters of 1946 presumed.

An act on disease is taken in a present where the evidence is thin and the expectation is wide; the record later keeps the thinness out and leaves the expectation unstated. A condition for revisiting would have to name what the actors expected to observe and when. What the Ragusa council expected its month to achieve, and when it expected to see it, is not in the record.

This study reads these three works as three questions that any record of an act on disease could be asked to answer. Neustadt and Fineberg's question: what observation was named at the hour that would make this act be revised, and if none was named, what would the actors have named? The ledger's counts answer the first part with zero conditions in eighty-nine records; the second part the records cannot answer at all. Neustadt and May's question: when did the issue begin for the actors, and what did they presume about it? The records give dates of events and dates of documents, and where a presumption appears it appears as a recorded reason; the nearest thing they give to an answer to the first half is the case where a description was written so that a later reader might recognise the disease again — Thucydides states that he shall set down the plague's nature "and explain the symptoms by which perhaps it may be recognized by the student, if it should ever break out again" (Thucydides, History of the Peloponnesian War, Book II, chapter 48, translated by Richard Crawley), a text kept for a reader who has not yet met it. Koselleck's question: what horizon of expectation did this act carry, and how far did it draw from experience? This study reads a rule made before any theory of how the disease spreads as carrying a horizon that no observation inside its own record would confirm or withdraw. Whether a record lets a later reader judge an act in its own time says nothing about whether the act was right.

Each question asks of a record something the record does not hold. That absence is a diagnosis of the record form rather than of the actors. A later reader who wants a condition for revisiting an act must write one from the record's own terms, and it then carries the reader's date rather than the act's. Neustadt and Fineberg's remedy was written for a decision in progress; read against these records it points at a form — a record carrying, beside the date of an act and the reason given for it, an observation that would end it and the date on which that observation would be made.

11.5Conditions a record could carry

The records code their acts by kind, and five kinds appear among the twenty-three cases read closely: restrictions of movement, work or assembly; notices, declarations and certifications; trials and treatments; counts and published reports; and institutions, compilations or gifts of medicines. One further case, Nightingale at Scutari, is coded as other, and this chapter draws nothing from it; Chapter 5 reads that case. For each of the five, the table asks what an observable condition for revisiting would name, in the terms the records use.

Restriction. Five restriction cases in the twenty-three are dated in their records, from 1348 to 1918. An observable condition for revisiting would name the review date, the counting system by which the restriction would be judged, and the state of that count that would end it. The counting system belongs inside the condition, because a count made during an outbreak changes with the definitions and the testing available, and one word can name different things at different dates. The records give the start, and in places the cost. Across the twenty-three, the ledger's cell on who bore the wait reads named in five cases, unsettled in five and not named in thirteen. A record that gives a start date and no review date cannot be read for what would have ended the restriction, and the cost of the interval stays with whoever happened to bear it, whoever that was. This is a diagnosis of what a restriction record could carry, not a verdict on any restriction.

Declaration. The declaration cases divide in a way their conditions show. A certification asserts a state of affairs and names the years, the standard and the search standing behind it; section 11.3 sets those out. A declaration that opens a process asserts a characterisation of something already under way, and the records give the date of the assertion. An observable condition for the second kind would name the criterion asserted, the evidence standard behind it, the date of review, and what would change it. The records name the date and leave the other three out. A declaration that closes on something requires the search to be checked afterwards; a declaration that opens something requires the change in the evidence to be named in advance. Chapter 8 reads these records on the ledger's cells. What this chapter adds is only that the conditions a later reader would need for either kind are absent from the record's own terms, and that the reader supplies them from outside.

Trial. The three trial cases name their comparison in the record: the patients and their common diet, the treatments or inoculations they were given, and the dates of inoculation and of later exposure. An observable condition for revisiting would name the outcome observed, the date on which it was observed, and the result that would end the practice. What such a condition makes possible is what a comparison with no date for its observation cannot give: a point at which the comparison stops being current. Chapter 4 reads these trials on the axes, including the discipline that makes a comparison checkable at all. The condition drafted here is what a record would have to carry; it is not a claim that any trial was conducted or reported badly.

Count. The count cases carry conditions of a different shape. Where a count qualifies itself in its own terms it states its own revision condition. A count called provisional names what would change it: the final figure, still to be made. A count that states its scope — a named category of cases, from a named place or a named date — names the boundary within which it holds. A count that joins two separate counts names what it joins, and a later reader can see that the total is not one system's total. Among the cases read in Chapters 5, 6 and 10, this qualification appears as a word inside the count's own record rather than as a date for recomputing it. A condition drafted for a count would name the denominator, the case definition, the reporting lag and the date of recomputation. The records give a denominator in some cases and a case definition in others, and neither as a standing rule for revision. Chapter 10 sets these counts beside one another.

Compendium. The three remedy records carry their own terms in three different ways. One arranges medicine under categories and cites the sources it draws on, which makes a new source visible as new and gives a later reader something against which to check the compilation. One lists the items of an offering record, listing the gift itself; the list records the gift and does not record use or effect, so no condition on the record's own terms can be observed, and the question of what the list was for cannot be answered from it. One carries case entries with one of three verdicts, and the reading note gives treatments for the treatable cases but no outcome for any case; what happened to the person judged untreatable the note does not give. Chapter 9 reads the three on the ledger's cells. The condition drafted for a compendium names its sources, its revision date, and what new remedy or new case would trigger revision. The three records carry part of that, and none of it as a rule.

What is common to the five kinds is a shape rather than a substance. An observable condition for revisiting names an observation, a person or body that makes it, a date on which it is made, and what the observation would do to the act. The ledger's field for conditions of revisiting counts none in any record, and the horizon table lists zero conditions per case in every band. A later reader who wants such a condition has to write it, and it then carries the reader's date rather than the act's. This chapter has written five of them, in the terms the records use, to show what that writing involves and what it costs the reader who does it.