References and Reproducibility

Where everything in this site came from, and how to verify it.

Every claim on this site comes from a particular row in a particular CSV file in a particular research repository. The purpose of this short closing chapter is to make that traceable. If you have read this far you may have wondered, at any of several points, “how do we know that?” The answer in every case is: we counted it, we logged the count, we wrote the count to a file, and the file is in the repository. This chapter says where.

I’ve written this as a brief reading and reference essay rather than a numbered bibliography. The literal bibliography is in the scholarly paper: §7 of paper.md carries the full citation list in academic form. What follows here is the layperson’s version: what we used, why, and where it lives.

Primary sources: the two corpora

The Founders’ writings are taken from Founders Online, the open-access archive maintained by the National Archives at founders.archives.gov. It collects in one place the established scholarly editions of six Founders: The Papers of John Adams, The Papers of Benjamin Franklin, The Papers of Alexander Hamilton, The Papers of Thomas Jefferson, The Papers of James Madison, and The Papers of George Washington, plus their Retirement papers and family correspondence series. We downloaded the full text for all six, deduplicated against editorial duplications, and ingested into a local SQLite database. The result is 68,287 individual documents comprising approximately 23.7 million words of writing, spanning roughly 1722 (Franklin’s Silence Dogood essays) to 1836 (Madison’s death).

Without Founders Online, this project would have been impossible at any reasonable scale. The archive is one of the most consequential open-text infrastructure projects of the last twenty years. A reader interested in any individual passage cited on this site can find the same passage at Founders Online with its full editorial apparatus.

The Shakespeare corpus is taken from Project Gutenberg’s public-domain editions of the thirty-six canonical plays plus the Sonnets and A Lover’s Complaint. We used Project Gutenberg’s plain-text exports, stripped of the standard Gutenberg license boilerplate (a separate normalization step, verified by a small audit script that confirms zero remaining boilerplate sentinel strings). The full Shakespeare corpus, after cleaning, is 891,092 words across 38 documents. The 1623 First Folio orthography is partially preserved. We did not modernize spelling because doing so would lose the very variants (hath, doth, thou) that several of the case studies depend on.

Methodological framework: the books behind the methods

The methodology rests on one open-access textbook: Anatol Stefanowitsch’s Corpus Linguistics: A Guide to the Methodology (Language Science Press, 2020). It is the most readable and most thorough recent introduction to the corpus-linguistic toolkit, and it is free to download from the publisher. Every statistical test on this site traces to a chapter in Stefanowitsch: log-likelihood (G) keyword analysis from Ch. 7 and Ch. 10, sample-size correction for vocabulary measures from Ch. 9, differential collocate analysis from Ch. 7.1.3, Configural Frequency Analysis from Ch. 6.6, target-domain metaphor extraction from Ch. 11.

The structure of the scholarly paper (one chapter per case study, each with research question, construct, operationalization, hypothesis, test, and effect size) follows the template laid out in Stefan Th. Gries and Magali Paquot’s Practical Handbook of Corpus Linguistics (Springer, 2020). Where Stefanowitsch gives the statistical machinery, Gries and Paquot give the rhetorical machinery: how a corpus paper is supposed to be organized so that a reader can follow it.

Two further reference works inform the more interpretive parts of the project. The first is Jonathan Hope’s Shakespeare and Language: Reason, Eloquence and Artifice in the Renaissance (Bloomsbury, 2010), which gave us our working sense of what Shakespeare’s early-modern English does: what it can do that the Founders’ late-eighteenth-century English cannot, and vice versa. The second is Kevin J. Hayes’s The Road to Monticello: The Life and Mind of Thomas Jefferson (Oxford, 2008), and Hayes’s broader body of scholarship on the reading lives of the Founders. Where the corpus tells us what the Founders wrote, Hayes tells us what they read, and the two pictures triangulate each other.

Methodological inheritance: the project’s nearest model

The closest single methodological model for the project’s structure is Lincoln Mullen’s America’s Public Bible: A Commentary (Stanford University Press, 2023), which traces the circulation of Biblical quotations across nineteenth-century American newspapers and presents the findings as a digital commentary on a parchment-styled website at americaspublicbible.supdigital.org. The shape of this site (long-form essays, per-finding case studies, an interactive explorer layer over the same underlying corpus, parchment design language) is directly modelled on what Mullen built. None of the substantive findings here come from America’s Public Bible; the inheritance is of structure and form.

The research infrastructure

Everything on this site (every claim, every passage, every chart, every catalogue row) rests on a Python analysis pipeline that lives in a separate research repository. The pipeline does five things in order: it ingests Founders Online and Project Gutenberg into a SQLite database with one row per document; it normalizes spelling and tokenization across the two corpora; it runs eight analysis chapters (keyword analysis, vocabulary richness, differential collocates, personal reference, archaic-form survival, metaphor pattern, multivariate CFA, n-gram coinage) producing eighty-five separate CSV outputs; it runs three influence-rank measurements over the per-case results (Influence-1, Influence-2, Influence-3); and it emits a single integrated scholarly paper in Markdown.

Six of the case-study CSVs power most of what the site renders. The catalogue explorer is built from catalogue_direct_quotes.csv and catalogue_named_references.csv. The Honour Test draws from the per-target collocate files in tables/cs3_*. The Composite Ranking pulls from influence3_composite.csv. The eleven-method convergence matrix comes from six_method_convergence.csv (the file kept its original name as later methods were added). And the timeline draws from the date column of both catalogue files. Every claim on the site can be traced from the prose to a row in one of these CSVs.

The scholarly paper carries the full statistical apparatus in paper.md: G-statistics, Bonferroni-corrected p-values, effect sizes on Stefanowitsch’s verbal scale, methodological caveats, a methods appendix, and the full bibliography in academic form. The narrative paper carries the literary discussion without the statistical apparatus. The summary paper carries the result table without either the apparatus or the narrative. All three are downloadable from the papers page.

The site itself

This site is built in Next.js 14 with Tailwind CSS, Recharts for the interactive bar charts, and a few hand-rolled SVG components for the timeline and the per-passage rate plots. The page templates and the design tokens (parchment palette, EB Garamond + IM Fell DW Pica typography, the iron-gall ink and folio-red accent colours) are open in the public site repository at github.com/willimj3/shakespeare-in-the-republic. The site is statically built and deployed to GitHub Pages on every push to main. Any reader who wants to extend it (new case studies, new explorers, new findings against the same corpus) can fork the repo and rebuild.

Images and attribution

Every historical image on this site is in the public domain or used under permissive Creative Commons terms. Full attribution (image holder, year of original creation, and licence terms) is on the credits page. The principal sources are the Bodleian Library’s digital edition of the 1623 First Folio (Macbeth, Othello, Julius Caesar pages), the National Archives image library (Adams diary manuscript, Washington orderly book), the National Portrait Gallery (Trumbull, Stuart portraits) and Wikimedia Commons. Where we have lightly composited or cropped an image for layout reasons, the modification is noted in the credits row.

Acknowledgments

Several debts. To Anthropic, for Claude Code, the AI coding agent that drafted the Python pipeline, wrote the bulk of the essays and case studies, and built the interactive site you’re reading. The terms of that collaboration are described in detail at About this project. To Lincoln Mullen, for the digital-humanities model the site is built on. To Anatol Stefanowitsch, for an open-access textbook so generous and so clear that an independent researcher with no formal corpus-linguistics training could read it and produce work this faithful to the discipline’s standards. To the editors of the Founders’ papers projects at Yale, Princeton, the University of Chicago, the University of Virginia, and the Massachusetts Historical Society, whose decades of patient editorial work make the corpus exist. To the staff at the National Archives who maintain Founders Online. To the Project Gutenberg volunteers who digitized Shakespeare. To the Bodleian Library and the National Portrait Gallery for public-domain image releases.

Return to the essay index, read one of the case studies, or open the interactive explorer.