Chapter 3. Data and methods

3.1The primary texts

Two texts are read. The first is the account of the outbreak near Golden Square in John Snow's On the Mode of Communication of Cholera, second edition, much enlarged (London: John Churchill, 1855), printed pages 38 to 55, in the transcription of Project Gutenberg (Snow 1855). The passage runs from the opening of the account to its last paragraph before the account of Deptford. It includes Snow's report to the Board of Guardians, his Table I of attacks and deaths by day from 19 August to 30 September, and the paragraphs on the fall of the attacks before the handle was removed. Gutenberg's transcribers say they silently corrected obvious typographical errors; the text is not marked as uncorrected, and it is quoted here at length and word for word, with Snow's spellings ("connexion", "diarrhœa") and his punctuation. Omissions are marked by an ellipsis between two verbatim pieces. The comparison of the two south London water companies, which both the reading note on Snow's book and the biography single out as his stronger evidence, lies outside this passage and is not read here.

The second is Louis Pasteur's paper Méthode pour prévenir la rage après morsure, read at the Académie des sciences on Monday, 26 October 1885 and printed in the Comptes rendus, volume 101, pages 765 to 772 (Pasteur 1885). The only text of it available to this study is the Internet Archive's machine reading of the printed volume, kept uncorrected: the author's own name is misread in the title line, letters are confused throughout, and the table of the inoculations given to Meister is garbled. The rules under which this article was written forbid quoting a source text marked as uncorrected machine reading, since a quotation copied from it would pass off the machine's errors as Pasteur's words. Pasteur is therefore followed in close paraphrase, in the order of his paper and in this article's own English, and no sentence of his appears in quotation marks. Every detail attributed to him below is in that paper; where the garbled table would be needed, the article relies on Pasteur's prose, which states the totals.

The asymmetry is a limit of this article, and it should be stated plainly. Snow can be heard in his own words; Pasteur cannot. A reader who wants to weigh a phrase of Pasteur's must go to the page images.

3.2The platform records

Beside each text stands a record written by Skeptron, a studio of the RISCON family that publishes questions about the past drawn from public records. Each record gives a dated event, a short scenario, and five questions with Skeptron's answers: the date of the record, a concrete fact, the boundary of what can be judged, a calendar check, and the condition of a later comparison. The two records used here are titled, in English translation, Number of outbreak cases before and after removal of the Broad Street pump handle (Skeptron's key HQ-HC2-SNOW1854) and Pasteur's Meister Case and the Scope of Clinical Evidence (HQ-HC6-PASTEUR1885), both dated 2026-09-14. This article refers to them by the parent study's short labels, The Broad Street Pump and Pasteur and Joseph Meister.

These records are modern summaries written for the platform. They are not archival documents and they are not nineteenth-century evidence of anything. Neither is a synthetic scene: both are public-record questions built on public sources that the records name. For Broad Street the named source is the UCLA John Snow Archive's page on the outbreak in Snow's 1855 book; for Meister it is a newsletter of the Institut Pasteur dated 2015, not Pasteur's paper of 1885. Skeptron writes in Korean, and the records are quoted here in an English translation made once for the parent study, copied from that translation character for character. Where a record and a primary text differ, both are given.

Each record also carries sentences that recur, with names and dates changed, across Skeptron's history cases. They are Skeptron's method, its house rules for reading any past event, and not findings about these two. An example, from the Meister record:

This article reads such sentences as the platform's rules and does not treat them as evidence about 1854 or 1885. The fourth question of each record, a weekday computed from the calendar, says nothing about the event and is not used.

3.3The study and its ledger

This article is drawn from a longer study, Before the Count, which reads twenty-three of Skeptron's medical and public-health cases closely and codes eighty-nine of its history cases on the same questions. The coding was done twice, independently, by one AI model run twice. Each coding gave a value for each question, the words in the record that supported it, and a reason; a value counts as settled only where the two codings agree. The questions were registered, with the study's hypotheses and predictions, before any case was coded. They ask whether a record shows a public act on disease, and of what kind; whether the act was taken before its cause, or its own effect, was established; whether the record names a gap in time between an event and its notice, count or settled cause; whether it gives a count and how it qualifies it; whether it names who bore the harm or cost while the act waited; and whether, apart from Skeptron's recurring sentences, it names a later finding that would change how the act should be read.

In plain words, the ledger reads both records alike on five of these questions. Both codings found a public act in each; both found that each act came before its cause or its effect was established; both found a gap in time named; both found a later finding named that would change the reading; and both found that neither record names who bore the cost while the act waited. They read the two differently on two points. They classed the Broad Street act as a restriction and the Meister act as a treatment, and they found that the Broad Street record gives its count as settled while the Meister record qualifies its count by scope. The ledger's values are given in the Appendix in its own terms. They are readings of the records, not of Snow's and Pasteur's texts, and where the texts say more the article follows the texts.

3.4How the reports are read

Each report is read for three things, in this order. First, who had to be persuaded, and in what setting: an authority, a colleague, an academy, a reader. Second, what the author put forward as evidence at the moment of the act, as distinct from what he added afterwards. Third, whom the author names as a witness, assistant or guarantor, and at what point in the argument. The first two questions belong to the evidence side of the debate; the third belongs to the alliance side. Reading both in the same document, rather than choosing one, is the method's whole point.