Sources and methodology

Each document is imported from a named digital edition, extracted by a script rather than retyped by hand, and compared against a second independent copy where one could be found. Original spelling and punctuation are kept. Markup and site navigation are stripped, and a few texts need further mechanical handling as well: rejoining a word the printer broke across a line, or removing a transcriber’s modern-spelling glosses. Every document’s own page lists what was done to it under Text handling.

Paragraph numbers used for linking and bookmarking (¶ 1, ¶ 2, and so on) are this edition’s own navigation, not standard scholarly citations. Use the citation button on any document page for a conventional reference.

The Constitution, the Declaration, the Articles of Confederation, the Bill of Rights, and all 85 Federalist papers have each been checked against an independent second witness. Of the 44 Anti-Federalist documents, 40 have now been checked the same way. Four have not: Brutus XVI, Cato IV, Cato V, and Luther Martin’s Genuine Information. In each case no independent second copy meeting the same standard could be found: an available comparison page turned out to be an excerpt or a mismatched document, or, for the one long address, would need to be reassembled from several separately dated pieces first. None of the four is known or suspected to contain an error; the gap is that the comparison has not been run, not a defect the comparison found.

The three founding-era writings on slavery come from three different institutions: a keyed transcription of Benjamin Rush’s 1773 pamphlet from the Text Creation Partnership, the Massachusetts Historical Society’s transcription of the 1788 petition manuscript, and the National Archives’ own transcription of the 1790 memorial it holds. All three have been compared against a second copy, and the three comparisons are not equally strong.

The 1790 memorial has a witness of the usual kind: a second modern transcription of the same original, made by another institution. The two agree closely, differing mainly at the opening, where the witness begins after the address line. For the other two, the best second copy found was not a transcription of the same thing. Rush’s is a machine reading of a scan of a different edition, uncorrected. The petition’s is a reprint that regularized the petitioners’ own spelling throughout. Both comparisons found the same content in the same order, six of six checked passages for Rush and eight of eight for the petition, which is what a comparison of that kind can establish. Neither confirms the wording word for word, and the low overall agreement they report reflects the difference in edition, spelling, and scan quality rather than a fault found in the text here.

Being checked against an independent second copy is a different claim from knowing exactly what a source was itself copied from. Five of the six Anti-Federalist families used here (Brutus, Federal Farmer, Cato, the Pennsylvania Minority, and Luther Martin’s Genuine Information) come from digital editions that do not state their own immediate copy-text: the particular paper or manuscript the transcriber worked from. George Mason’s Objections is the exception, with a documented manuscript copy-text cited on its own page. The open copy-text question for the other five is disclosed on each affected document’s own page and is not settled by an independent-witness check, which tests whether two separately produced transcriptions agree with each other, not what either one was originally copied from.

An independent-witness check is a mechanical comparison, not a historian’s reading. It establishes that two separately produced transcriptions agree with each other. It does not certify that either one is faithful to the original in every particular. No text here has had the further human review a line-by-line scholarly collation against the manuscript or the first printing would require.