Disunion Feed · The verification record

What this page is

Every passage on this site is put through a checking program before it ships. The program splits the quoted excerpt at each [...] mark, fetches the cited document live, and requires every fragment to appear in it. A spot-check is not enough: a spot-check passes happily while a sentence quietly dropped from the middle turns a real quote into a fabricated one. That is a mistake we made early, and it is why the check works the way it does.

It is not a character-for-character comparison, so it is worth being exact about where it gives ground. It tries the strictest comparison first and loosens in steps: treating curly and straight quotation marks alike, then ignoring capitalisation entirely, then ignoring whitespace, and at its loosest also dropping hyphens, dashes and apostrophes — which scanning software inserts and breaks constantly. It reports which step succeeded, and we read that. Apart from differences those steps erase, it does not allow a word to be missing, added or swapped — and at its loosest step what they erase is a pair like well and we'll. That is a real limit, not a rounding.

This page is that program's result together with our reading of every case where it objected. The program finds disagreements; it cannot tell you whether a disagreement is a broken scan or a bad quotation, and a person has to say which. Those readings are ours, and each one is written out below so you can disagree with it. The page exists because the alternative is asking you to take our word for the whole thing rather than for the parts we actually judged.

Everything below is the result of one run of that program on 2026-08-21. It attempted all 134 sources and completed a comparison on 131 of them; the other 3 are reported as unknowns below rather than folded into either result. It is a dated reading, not a standing fact: these are 20 independent archives on the public internet, and the numbers move between runs when one of them is slow, down, or refuses us.

• 114matched their source
• 17differed, each explained below
• 1our checker cannot read at all
• 2could not be reached that day

Those add to 134. 3 of those passages have no comparison result at all, and we keep them out of both other counts on purpose. We do not know whether the unreadable one differs, because our checker cannot read an image of a page. We do not know whether the 2 we could not reach differ either — one archive returned nothing and one refused the plain request our checker makes, so no comparison happened in those cases. Calling an unknown a known result is the kind of rounding this page exists to avoid, and filing any of them under the matches would be worse.

The 17 that differ, the 1 we cannot check, and the 2 we could not reach

Each one is listed below with what the difference is. They fall into four kinds, and the kinds matter more than the count: some are our own edits, some are damage in the scanned source, one is a passage our checker cannot read, and some were simply unreachable the day we ran the check. Only the first two are disagreements we have actually weighed. Several of the entries record mistakes of ours beyond the text itself — a link that pointed at the wrong page, a checker that matched the wrong article. Those are noted where they happened.

3 are changes we made

These are ours, so they go first. Judge their size yourself.

14 are damage in the archived text, where we judge our card correct

Newspapers from the 1850s reach us through photographs of microfilm, and then through software or scholars turning those photographs into text. Both stages make characteristic mistakes — a capital I read as the digit 1, an x for a g, a page number dropped into the middle of a sentence. In each case below the archived text is broken in a way we can name, and we judge the card correct. Note the word judge: for most of these the corruption is obvious enough that we do not think it is a close call, but it is our reading and the next paragraph of each entry tells you what it rests on.

1 cannot be checked by our checker at all

It sits apart because it is a different kind of gap: not a disagreement we have weighed, but a check we were never able to run.

2 could not be reached on 2026-08-21

These are not judgements about the passages, and not all of them are the archive's doing — one common cause is that an archive refuses the plain request our checker makes, which is a limit of our method rather than a fault of theirs. Each row below states the cause our checker recorded. We report them rather than re-using an older result, because a number carried forward from a run we are not showing you is exactly the kind of thing this page is supposed to make visible. A later run may well move them into one of the groups above — and often does: a passage in this group on one run is frequently a match on the next.

Every passage, with its source

Every row links to the passage on this site and to the best copy of its source we could reach, and says whether our checker matched it.

One thing to understand about those links, because it changes what "go and check" means. Most of what survives of the 1850s press online is not a photograph of the page — it is a transcription made by a library or a scholarly project, and that is what most of these links open. Photographs of the original pages exist for only a handful. Where a card rests on a secondary transcription rather than a first-hand printing, the card itself says so; the weakest citation in the corpus is Mahala Doyle's letter to John Brown, whose original manuscript we have never been able to reach. A transcription agreeing with our card is real evidence and it is not the same as the page itself, and we would rather you knew which one you were looking at.

Kansas-Nebraska, 1854 (11 passages)

The Sumner caning, 1856 (19 passages)

Dred Scott, 1857 (28 passages)

Harpers Ferry, 1859 (22 passages)

The secession winter, 1860–61 (54 passages)

All source material is pre-1930 and in the public domain. Transcription hosts include the Valley of the Shadow project (University of Richmond / New American History), the Secession Era Editorials Project (Furman University), the Frederick Douglass Papers (Yale), Fair Use Repository, the Internet Archive, Canadiana, and the Library of Congress.

The explained differences on this page and in the checker come from one shared file, so those two cannot quietly acquire different lists from each other. That guarantee covers the explanations and nothing more: a passage the checker could not reach has no entry in that file at all, which is why the unreached ones above are described from the run itself rather than from our notes. Everything else here is written by hand and carries the risks of that: the explanations are ours, they reflect our judgment rather than the checker's output, and they can be wrong. The full run against all 134 sources is something we do deliberately rather than on a schedule, because it fetches from 20 archives — so an archive could change under us without this page knowing. If you find something here that does not match what you see at the source, we would genuinely like to be told.