The historical claim of this section is that the federal weekly reports of autumn 1918 recorded the spread of influenza and not the public orders taken against it, and that for Philadelphia they record nothing that a comparison of cities by the timing of their measures could use. The weekly reports knew what had been telegraphed to Washington, said plainly that this was incomplete, and named no condition under which anyone should act differently. The section adds to the timing debate a reading of what the government's own running account could and could not show while the season unfolded, and to the precaution debate an instance of signs collected without any statement of what they were for.
4.1The issue of 27 September
The first of the four items opens with an explanation of its own limits. Influenza, it says, was not a disease that most states required to be reported, and so the Service could not obtain a full picture of its spread among civilians. To fill the gap, the Surgeon General had wired the health officers of the states, and the Service's own officers, asking them to reply by telegraph. What follows is a digest of the answers that had arrived by the time the issue went to the printer.
The opening paragraph thus states, before any entry, three facts about the knowledge it carries. The knowledge was partial, because the legal machinery of notification did not reach the disease in most states. It was solicited, because the Surgeon General had asked for it rather than receiving it as routine. And it was dated by the press, not by the events: the entries are what had been received, not what had happened, up to a moment the paragraph does not fix.
The entries that follow are uneven in kind: some counts, some impressions, some single sentences of report at second hand. Connecticut describes the disease as widespread in the east and south of the state and growing. Massachusetts, whose health officer had telegraphed on 24 September, lists towns with cases and deaths, and gives a week's registered deaths for Boston from influenza and from pneumonia separately. Vermont's health officer had received vague reports from many places, with no diagnosis confirmed. Kentucky reports from Louisville a few cases that might be influenza. The longest entry is Massachusetts's list of towns. Another is clinical: a steamship that had sailed from Boston on 1 September reached the quarantine station at New Orleans, coming from Tampico, on 14 September, with deaths and illness on board, and the entry lists the crew's temperatures and complaints.
Pennsylvania's entry, in this issue, is short and reassuring. Among civilians, it says, relatively few cases had been reported so far; there had been a handful at a place the entry calls Hog Island, and Pittsburgh was said to have some. The entry does not name Philadelphia.
Placed beside the platform record, this entry raises the first question of the case. The record gives the first civilian reports in Philadelphia as 17 September. The federal digest printed on 27 September, on the eve of the parade, does not mention them. The two need not conflict: the digest carries what Pennsylvania's officials had sent to Washington before the issue went to press, not what had happened in the city. The record's date comes from a study published in 2007. What the digest shows is that, at the federal level, the signs a later study treats as the start of Philadelphia's epidemic had not yet become a statement about Philadelphia.
4.2The issue of 4 October
The second item, dated 4 October, just after the date the record gives for Philadelphia's closing order, opens differently. Reports reaching the Service, it says, show that influenza is spreading fast. It has been noted in most of the states and in the District of Columbia; it is heaviest on the Atlantic coast and along the Gulf, and it is moving west.
The tone of the entries has changed. Alabama, Maine, Montana and others report epidemics by place. Massachusetts gives an estimate, made on 1 October, of the cases in the state outside the military camps. Illinois quotes its health officer's telegram of 1 October on improvement near the Great Lakes naval station and spread along the main lines of travel. New York reports outbreaks at Buffalo and in New York City, with the disease said on 30 September to be declining in the city.
Pennsylvania's entry is now a single sentence: in both Philadelphia and Pittsburgh, it says, the disease is epidemic. It gives no count, no date, and no mention of any measure. The order that the record dates to 3 October does not appear, though the issue is dated the next day.
The silence is not evidence that no order was given. The digest compiled telegrams about prevalence, and its entries for other states say nothing of measures either. It recorded where the disease was and, where it could, how much of it there was; it did not record what cities did about it. A later reader cannot date Philadelphia's order from this journal.
4.3The issue of 11 October
The third item opens by observing that influenza now covers practically the whole country, that the eastern states still send the largest numbers, and that the West's numbers are climbing fast.
This is the first of the four issues to report public measures. The Washington entry says that several named towns, Seattle first among them, have shut their schools and forbidden public gatherings. A sentence saying that schools and theatres have not been closed follows the New York entry in the transcription, though the transcription's header assigns it to Maryland; which state it concerns is left open here. A third entry reports a change in the machinery of knowledge rather than of control: Massachusetts made influenza notifiable on 4 October, and its reports now seem to show cases falling in the east while the west of the state sees more.
Pennsylvania's entry locates the heaviest prevalence in Philadelphia's industrial districts and their surroundings, in the anthracite country of the Lehigh Valley, and in patches of the soft-coal fields. Again there is no count and no measure. Missouri's entry names St Louis, with Kansas City and St Charles, as places from which cases have been reported, and says nothing of what St Louis had done.
The two cities that the timing thesis sets side by side thus appear in the same issue, each as a place with reported cases, neither with any mention of the measures a later study would treat as decisive. The one closing the issue clearly reports, in Washington, is given without comment.
4.4The issue of 18 October
The fourth item opens with the sharpest statement of the four. Reports since the last issue, it says, show influenza spreading fast and present in every state, with cases rising in nearly every community apart from a few in the east, where they seem to be falling.
The entries of this issue are fuller and more often cumulative, and several qualify their own totals. California's figures for deaths in the first half of October are said to be incomplete. Arkansas reports that many rural cases receive no medical attendance and so go unreported. Virginia sets an estimate of its cases beside a smaller number actually reported by the same date. Minnesota's health officer sees no cause for alarm across the state as a whole. The marks of incompleteness are the reports' own: cases unseen by any physician, and reported counts that fall short of officials' estimates. These are this article's observations about the 1918 entries; Lipsitch and colleagues write of a related problem, that mild cases are missed and that reporting differs between places and over time (Lipsitch and others 2015).
Pennsylvania's entry now gives two statewide totals of registered deaths up to and including 14 October, one for influenza and one for pneumonia, and adds that the greater number of them occurred in Philadelphia. It is the first of the four Pennsylvania entries to give a figure, and the figure is statewide, cumulative, split by registered cause, and dated 14 October, before the issue of 18 October went to press. It says nothing of the parade or the order. Missouri's entry gives cases and deaths for St Louis up to 16 October, and those reported on that day.
The last issue, then, gives a later reader something that a comparison of cities can begin to use: counts with a date and a scope. It also shows how far those counts were from the series a comparison needs. Pennsylvania's figure is for the state, not the city; it separates influenza from pneumonia by registered cause; and it is a running total to one day, not a rate over a fixed period. The record's phrase for what the later study did, fixing the observation period and the denominator of the mortality rate, names exactly what these entries lack.
4.5What the reports knew and what they did not say
Sorted by the four questions, the reports of autumn 1918 present as known what had arrived by telegraph before each issue went to press, on the word of state health officers, the Service's own officers and, in some entries, local reports at second hand. They present as unclear the completeness of their own information, from the opening of the first issue to the repeated qualifications of the last; they say nothing at all about the nature of the disease or the value of any measure. What they presume is the worth of the running account itself: the Surgeon General asked for telegrams because a national picture, even an incomplete one, was worth having. They do not say what it was for. They neither advise nor recommend, and they record the closings in Washington without comment.
And they name no point at which any action would be looked at again, because they are not documents of action. In the terms of Neustadt and Fineberg's recommendation, there is no decision here whose reconsideration could be planned. The decisions were the cities'; the federal journal heard about the disease and, in one issue, about closings in one state.
The federal record of the hour kept the signs and not the orders, with honest marks of their gaps, in a form that changed over four weeks from a list of places to a list of counts. Read as Rosenberg's progressive revelation (Rosenberg 1992), the sequence is plain on the page: a first issue in which Pennsylvania is quiet, a second in which Philadelphia is named as epidemic, a third that places the disease in the city's industrial districts, a fourth that places most of the state's registered deaths there. Recognition did not wait for counts. The journal named Philadelphia epidemic on 4 October and gave no figure for Pennsylvania until 18 October.
4.6The record beside the reports
Skeptron's record of Philadelphia gives the sequence in two sentences that stand both in its scenario and in its answer on the concrete facts:
Civilian patient reports are given as September 17, the large-scale parade as September 28, and the introduction of school closures, assembly restrictions and the like as October 3. The research compares cities by fixing the observation period and denominator of the mortality rate.
The verb is precise. The dates are given, that is, given by the source, which the record names as the study of 2007 by Hatchett and colleagues. The record's first answer says the same of its own dating:
The public citation of the source currently cited is a 2007 research announcement.
The three dates, then, are a modern study's reconstruction, carried into a modern summary. They are not in the reports of 1918 read above, which mention neither the parade nor the order and do not name Philadelphia until the issue of 4 October. The record makes no claim that they are. It also states the gap between the first two of its dates and refuses to read a cause from it:
11 days. The preceding first civilian patient report and the later Philadelphia parade and delayed assembly restrictions are not merged into the same action. The point of occurrence, the observation conditions and the version of the causal investigation are separated for comparison. The precedence of the periods alone does not establish a single cause.
And it answers its own boundary question with a refusal:
None. Distinguishes the observational research comparison from a causal assertion about a single event. Does not use it as a modern individual treatment guideline, and verifies the policy timing and the scope of the data at the time.
The question it answers is this:
Can the pre- and post-difference of one city determine with certainty that a single event was the cause of all deaths?
The answer separates two things that the debate over 1918 sometimes runs together: the comparison across cities, which is observational, and a causal statement about one city and one event, such as the parade. The record does not say the parade caused deaths, and it does not say the late order did. It says the evidence of a single city's before and after cannot carry that claim.
Three things follow from setting the record beside the reports. First, the record holds the dates the timing debate needs and the reports of the hour lack. Those dates are a later study's, and the record's title, with its word "delayed", already takes that study's point of view. Delay is a relation between an act and a moment fixed afterwards as the start of the outbreak. In the weekly reports no such moment exists, only a quiet Pennsylvania entry followed by an epidemic one.
Second, the record's caution about cause matches the caution of the study it draws on. Hatchett and colleagues compare cities by timing and do not isolate the effect of any one measure (Hatchett, Mecher and Lipsitch 2007). The record adds that a single city's sequence cannot settle a single cause. Neither assigns a cause to one city's deaths.
Third, the record keeps apart the times the reports blur. The reports print what was received up to press time, which is neither the time of the event nor that of the count. The record separates the first civilian reports, the parade, the order and the source, and that separation lets a later reader see that the journal's silence about Philadelphia on 27 September is a fact about the journal's sources, not about the city.
4.7Crosby's city and the journal's city
Crosby's Philadelphia is a city overwhelmed: hospitals, doctors, nurses and the means of burial all failed under the load (Crosby 1989). None of that is in the federal journal, which gives the city one sentence calling the disease epidemic, one placing it in the industrial districts and one assigning it most of the state's registered deaths. The journal was a digest of telegrams about prevalence and was not built to carry the state of hospitals or cemeteries. Its form bears lightly on Crosby's question about why the epidemic left so little mark on American memory: at the national level, Philadelphia's epidemic was recorded exactly as Vermont's or Nebraska's was, a line in an alphabetical list.
4.8What this case adds
To the timing debate, the case adds a reading of what the federal record of the hour held: the signs, marked as incomplete, and not the orders. The dates that make the comparison of cities possible are not in the four weekly reports; they enter the platform record from a study of 2007. The reports show instead how recognition of Philadelphia's epidemic reached the federal level, as a change in a one-line entry, before any figure was given. To the precaution debate, the case adds early signs collected without any statement of what they were to inform or of any condition for action. The reports are neither precautionary nor late; they are not decisions. The decisions were the cities', on information the digest does not show.