The Public Health Service's reports of autumn 1918 knew what state officers had telegraphed before each issue went to press, and said it was incomplete. They named Philadelphia first as a place of epidemic and later as the place of most of Pennsylvania's registered deaths. They did not record the parade, the order or any measure in Philadelphia, and named nothing that would change any action. The report of February 2020 knew a set of dated, attributed reports and counts; it said what was unclear and what it believed by analogy; it named its assumption and admitted that its measures might not prevent what they were meant to slow. It promised revision as knowledge arrived and named no point at which revision would occur. In the next week's issue, one of its own dates was corrected.
The contribution is a reading of the documents of the hour for two debates. To the debate over the timing of measures in 1918, it adds that the dates on which comparisons of cities rest were not part of the federal running account, and that the platform's summary of Philadelphia takes them, by its own statement, from a study of 2007. To the debate over precaution, it adds a report that acted on signs by naming purposes rather than a principle, acknowledged ignorance as well as uncertainty, and left the costs of its measures unstated. To the discussion of international declarations, it adds the report's dated record of the emergency of 30 January beside the platform's record of 11 March, two acts of different kinds that an argument about warning has to tell apart.
The limits are those of the evidence: a paraphrased text of 1918 without its figures; a federal digest rather than the city's record; a report of 2020 that covers only its first weeks and shows the effect of nothing; and platform records built each on a single later source, coded in the parent study by one model run twice.
Three lines of further work follow. The 1918 issues could be read from the page images, their counts given, and set beside Philadelphia's own records of the same weeks, so that the dates the later studies use can be traced to documents of the hour. The report of February 2020 could be read beside the agency's later reports of that year, to ask whether the points of reconsideration it left open were ever named. And other platform records built on a single later study could be tested for how far their separations of time reproduce that study's dates rather than the documents' own. Each would test what two cases can only show: that the record of the hour and the record of the later argument keep different times, and that judging an act in its own time needs both.
Appendix: the ledger's values for the two cases
The values below are the parent study's, as its two codings gave them, for the platform records of the two cases. They are readings of the records, not of the reports of 1918 and 2020. "Unsettled" means the two codings differed.
| Case | act | act kind | act before proof | gap named | count | bearer named | revision in case |
|---|---|---|---|---|---|---|---|
| The Philadelphia Parade | yes | restriction-of-movement-work-or-assembly | yes | named | no-count | named | named |
| The Pandemic Characterisation | yes | notice-declaration-or-certification | yes | named | unsettled | not-named | named |
Agreement between the two codings across all 89 cases, for these cells: act 75 of 89 (84%); act kind 74 of 89 (83%); act before proof 69 of 89 (78%); gap named 76 of 89 (85%); count 78 of 89 (88%); bearer named 70 of 89 (79%); revision in case 67 of 89 (75%).
Among the 23 cases the parent study reads closely: act kind was restriction-of-movement-work-or-assembly for 5 and notice-declaration-or-certification for 6; bearer named was not-named for 13, named for 5, unsettled for 5.
The parent study's own fifth hypothesis (E5: in most cases that show a public act, the case's own text does not name who bore the harm or the cost while the act waited) was read as unclear on the strict reading (supports 17, against 6, unclear 49, not applicable 17; share 74%) and as supported on the lenient reading (supports 21, against 9, unclear 20, not applicable 39; share 70%). Its fourth (E4: in most cases that show a public act, the case's own text, apart from the studio's recurring sentences, does not name what later finding would change how the act should be read) was read as unclear strict (supports 9, against 14, unclear 49, not applicable 17; share 39%) and against lenient (supports 10, against 18, unclear 22, not applicable 39; share 36%).
The parent study's test of whether a record lets a later reader judge an act in its own time (a public act; a named gap between the event and its record; who bore the wait named; what would change the reading named) passed in 1 of 36 cases (3%); among the 23 cases read closely, 1. On the parent study's case-by-case table, the case that passes is The Philadelphia Parade, the only one of the 23 rows with all four. The Pandemic Characterisation fails on bearer named. The result is not robust to further readers: where two of three readings agree, the three comparisons give 2 of 20, 3 of 37 and 3 of 36; where every family that read a case gives the same value, 0 of 21; where a majority of families agrees, 2 of 28. The record that passes takes its dates, as Section 4.6 shows, from a study of 2007.