The King's Book
Days after Charles I was beheaded, a book of his private prayers and reflections made him a martyr. After the Restoration a bishop said he had written it. Whose words were they?
Eikon Basilike ("The King's Image") appeared within days of the execution on 30 January 1649. It ran through dozens of editions in a year and shaped how the King was remembered. Parliament commissioned John Milton to answer it. After 1660, Bishop John Gauden privately claimed he was the author. Royalists produced sworn testimony that the King wrote it himself. The question has never been settled.
This open question asks whether the book's writing habits sit closer to Charles's own private letters or to Gauden's published work.
To the best of our knowledge as of 2026-10-01. First result (C-07, proposed, not reviewed): the book as printed does not carry Charles's measured writing habits. It is consistent with Gauden but does not single him out.
The question
Open Question question: Chapter by chapter, does the first issue of Eikon Basilike (February 1649) match the writing habits of Charles I, of John Gauden, or of neither, when measured against Jeremy Taylor as a control?
Chapter-level results matter because the strongest middle position is that Gauden worked from the King's own papers. If so, some chapters may read like Charles and others like Gauden.
Claims
Evidential weight (cool) and adoption weight (warm) are shown separately. anchor_checked: no means we have read only secondary descriptions.
Primary text, read directly: Eikonoklastes tells readers to compare "Pammela's Prayer" in "the 3d. Book of Sir Philips Arcadia p. 248" with "the first Prayer of his Majestie, deliverd to Dr. Juxon."
Universal.
Not disputed by any source we found, but not yet checked by us. Both texts are available; the comparison is our positive control.
Generally accepted.
His letters pressing the claim survive and are cited across the literature.
Universal. Whether the claim is true is C-04.
Testimony on both sides, from interested parties. No measured test of the text found.
Divided. Many modern accounts lean toward Gauden working from royal papers.
Rests on later royalist testimony about the printer William Dugard. Not yet examined by us.
Repeated as "apparently" true on Wikipedia's Dugard page; disputed by others.
Absence from three web searches only. Weak.
Not a claim anyone holds.
Our own measured result (F10), on clean transcriptions, with the Charles profile shown to work across letters and formal argument (F8). Limited by a small sample of Charles's formal prose and by genre (G-05, G-06). Not reviewed.
New today.
Depth: what this does not show
It does not show Gauden wrote the book. It also cannot separate "someone else wrote it" from "Charles drafted it and someone else rewrote it", since heavy editing would also remove his habits.
Not shown. The book sits between Gauden's own works and other clergy (F10), and Gauden's profile attracts most learned 1640s prose (F9).
Our first run (F6) would have suggested this. We are recording that it did not survive controls.
Positions on authorship
| Position | Held by (examples) | Evidential basis | Adoption |
|---|---|---|---|
| Charles | William Levett (sworn statement); Richard Hollingworth (1692); Thomas Wagstaffe (1711) | Eyewitness testimony of the King writing at Carisbrooke. Testimony from a loyal servant, given decades later. | Low to moderate |
| Gauden | Gauden himself (1661); his widow's narrative | His own claim, made while seeking a richer bishopric. Self-interested, but specific. | Moderate |
| Composite | Many modern accounts | Gauden composed or polished from the King's papers; Jeremy Taylor sometimes named as reviser. Fits both bodies of testimony. | Moderate; common today |
Divergence note. Every position rests on testimony from people with a stake in the answer. None rests on a test of the text itself.
Feasibility check, 2026-10-01
sources
Clean, free texts for every candidate
The Text Creation Partnership has free, hand-typed transcriptions of all the key texts:
- Eikon Basilike, first issue (A38258): 60,359 words
- Charles's captured private letters, The Kings Cabinet Opened (1645, A31932): 24,215 words, including Parliament's notes, which must be removed
- Gauden: 14 free works, for example Three Sermons (1642): 35,309; The Love of Truth and Peace (1641): 12,196; The Religious & Loyal Protestation (January 1649): 4,094; Hiera Dakrya (1659): 377,574
- Jeremy Taylor: 45 free works, for example An Apology for Set Forms of Liturgy (1649): 30,162
- Milton, Eikonoklastes (A50898): 71,704
- Sidney, Arcadia (1590, A12229)
→ Effect: go. Hand-typed text avoids the scan-noise problem we had to work around in Q-001.
sources
The first issue has no prayers
The transcribed issue (Grismond for Royston, "within a few days of the death of the King") contains no Pamela prayer; the prayers were added in later editions.
→ Effect: the authorship test runs on uncontaminated text. The prayer question (C-05) becomes a separate, later dig.
method
Length is no longer the problem
60,000 words is far above the roughly 2,500–5,000-word floor that sank the word-frequency test in Q-001. Even single chapters (about 2,000 words each) are borderline, so chapters will be grouped where needed.
→ Effect: the classic word-frequency method (Burrows' Delta) is usable, alongside phrase matching.
confound
The genre problem from Q-001 applies again
Charles's corpus is private letters; Gauden's is sermons and treatises; the Eikon is devotional meditation. A test could end up sorting by genre, not author.
→ Effect: Taylor's devotional works become the genre control. If the Eikon simply looks "devotional", it will match Taylor as well as Gauden, and we will see that.
sources
Most of Charles's public writing is unsafe as a baseline
The TCP lists over 600 items under Charles I, but most are proclamations and declarations drafted by ministers such as Edward Hyde.
→ Effect: only his private letters count as Charles's own words.
Result, 2026-10-01 (finding F10)
Method: Burrows' Delta, a standard word-frequency comparison, on the 150 most common words after spelling normalisation. The question it answers: how close is each 2,000-word passage to an author's profile? Profiles never include the work being tested.
| Passages | n | Distance to Charles (median) | Distance to Gauden (median) |
|---|---|---|---|
| Charles's own writing, held out | 12 | 0.465 (0.45–0.51) | 0.744 |
| Gauden's own works, held out | 25 | 0.805 | 0.524 (0.49–0.58) |
| Eight other 1640s–50s clergy | 31 | 0.724 | 0.632 (0.59–0.71) |
| Eikon Basilike | 29 | 0.702 (0.67–0.72) | 0.591 (0.56–0.65) |
Lower is closer. Ranges in brackets are the middle half of passages.
What this shows. The book's passages are as far from Charles as other clergymen's writing is, and the ranges do not overlap with Charles's own. On Gauden, the book sits between his own works and other clergy, so it fits him without singling him out. The result fits "Gauden wrote it" and "Charles drafted, someone else rewrote it". It counts against the book being Charles's own unedited composition.
first run
A first run pointed to Gauden. It did not survive controls.
19 of 28 chapters came out nearest Gauden, many in all 200 resampled runs. But 4 chapters came out nearest Milton, who cannot have written them. That was the warning sign.
→ Effect: not reported as a finding about authorship. Kept on record because a headline built on it would have been wrong.
confound
The first Charles profile detected letters, not Charles
Queen Henrietta Maria's letters from the same 1645 collection came out strongly "Charles" (0.606 vs 0.85–0.89 for the others). Taylor's letter came out "Gauden".
→ Effect: added Charles's own formal argument (Newcastle papers, 1646) and Henderson's replies from the same pamphlet.
calibration
Charles is recognisable across genres
In the same pamphlet, Henderson's held-out passages came out Henderson 7 of 7. Charles's formal passages came out nearest a Charles profile 5 of 5, never Gauden. None of 31 passages from other clergy came out as Charles.
→ Effect: the Charles profile measures Charles, not just "letters". This is what makes C-07 meaningful.
confound
Gauden's profile attracts most learned clergy prose
17 of 31 passages from eight clergy not in the model came out nearest Gauden. Reynolds, Fuller, Hall and Baxter did so every time.
→ Effect: "nearest Gauden" means "a learned clergyman wrote this". Switched from "which profile is nearest" to the distance comparison above.
chapters
No chapter of 1,500 words or more falls inside Charles's own range
Each chapter was compared with passages of the same length from Charles's own writing and from the eight other clergy. In the main check, no chapter sat as close to Charles as his own passages do (0 of 28); every chapter sat where other clergy sit.
A fairer check uses only his formal papers against a profile built from his letters. At 1,500 words, 0 of 19 chapters fall inside his range. At 1,000 words the ranges overlap (8 of 26), so short chapters cannot be decided.
→ Effect: the composite theory cannot be supported for the longer chapters. For the short ones (about 1,000 words: chapters 1, 3, 7, 18, 22, 23, 25) the test cannot tell. G-07 partly closed.
positive control
The phrase matcher finds Sidney's prayer unaided
The 1687 "Prayer in time of Captivity" shares 73 six-word phrases with Arcadia, against 0–1 with Gauden, Taylor, Milton or the book's own chapters. The 1687 text changed Pamela's opening, "O all-seeing Light, and eternal Life of all things", to "O Powerful and eternal God", but kept the rest.
→ Effect: C-02 can move to established once reviewed.
Run files: README · corpora.py · charles1646.py · delta.py · genre_check.py · genre_matched.py · attractor_check.py · verification.py · pamela_control.py · chapters.py · chapters_fair.py · outputs: F6 F7 F8 F9 F10 F11 F12 F12-fair
Test design DT-B1
- Positive control first. The phrase matcher from Q-001 must find Pamela's prayer in Arcadia on its own.
- Build corpora. Charles: the Cabinet letters with Parliament's notes removed. Gauden: his secure 1640s–1650s works, leaving out disputed attributions. Taylor: devotional and polemical works of the 1640s.
- Calibrate. Hold out 2,000-word samples of each author and check the method assigns them correctly. The decision bar is set from this calibration, not guessed in advance (lesson from Q-001, F10).
- Score the Eikon chapter by chapter, with resampled uncertainty ranges.
- Publish code, corpora lists and every result.
Mostly Charles
Supports the royalist account. A ghostwriter skilled at imitating the King is not excluded.
Mostly Gauden
Supports Gauden's claim. Gauden may still have worked from royal papers.
Mixed by chapter
Fits the composite view, and would show which chapters to look at.
Conspiracy overlay
The claim: "Parliament's propagandists planted a plagiarised prayer in the King's Book to discredit it."
Documented
The prayer matches Sidney's Arcadia, and Milton used that to attack the book (C-01, C-02).
Unproven
Someone on Milton's side arranged for it to be inserted (C-05).
A real borrowing that was used against the King does not show who put it there. That needs evidence about the printing itself: which editions carry the prayer, and when.
Open gaps
Excavation log
Append-only.
Open Question Q-002 opened from the Q-001 queue (A-08). Chosen because its text is long enough for style tests, which Q-001's French letters were not.
Found clean transcriptions for every candidate in the EEBO-TCP corpus. Confirmed the first issue has no prayers.
Length cleared. Genre confound carried over from Q-001; Taylor added as genre control. Charles's public declarations excluded as ministerial drafts.
Read Milton's charge directly in Eikonoklastes; anchored.
DT-B1 planned, with the decision bar to be set from calibration rather than in advance.
First run: 19/28 chapters nearest Gauden; 4 nearest Milton. Treated as a warning, not a finding.
Genre check: the Queen's letters scored as "Charles". First Charles profile invalid on its own.
Added Newcastle papers (Charles vs Henderson, same pamphlet). Charles recognisable across genres.
Eight control clergy: 17/31 passages nearest Gauden. "Nearest" method abandoned for distance comparison.
Distance comparison: book far from Charles, between Gauden and other clergy. C-07 filed as proposed; C-08 logged as not shown.
Chapter-level test: no chapter of 1,500+ words inside Charles's range; short chapters undecidable. G-07 partly closed.
Positive control passed: captivity prayer matches Arcadia (73 six-word phrases).
Sources
- Text Creation Partnership (EEBO-TCP), catalogue and transcriptions. github.com/textcreationpartnership primary_text · free
- Eikon Basilike, first issue, 1648 [1649]. TCP A38258 · Gutenberg primary_text
- Charles I and Alexander Henderson, The Papers which passed at New-Castle (1649). TCP A78958 primary_text
- Bruce, J. (ed.), Charles I in 1646 (Camden Society, 1856). Internet Archive primary_text
- The Kings Cabinet Opened (1645). TCP A31932 primary_text
- Milton, Eikonoklastes (1649). TCP A50898 primary_text
- Sidney, The Countesse of Pembrokes Arcadia (1590). TCP A12229 primary_text
- Gauden, Three Sermons (1642), TCP A42498; The Religious & Loyal Protestation (1649), TCP A42492 primary_text
- "Eikon Basilike." Wikipedia synthetic
- University of Maryland Libraries, "Authorship Debate." lib.guides.umd.edu synthetic
- "William Dugard." Wikipedia synthetic · adoption signal
- "John Gauden and the Authorship of the Eikon Basilike." The Library, s5-XX(2). Oxford Academic interpretive · not yet read