Henry VIII and The Two Noble Kinsmen
from First Principles

1. The Problem

Henry VIII — or All Is True, as the play was known in its own time — is one of the longest-running authorship debates in Shakespeare studies. The play was printed in the First Folio of 1623 without attribution to any collaborator. It was accepted as entirely Shakespeare's until 1850, when James Spedding argued from internal verse evidence that substantial portions must be by a different hand. That hand, he proposed, was John Fletcher.

Spedding's argument has proven remarkably durable. It was reinforced by Cyrus Hoy's linguistic-marker work in the early 1960s, extended by Jonathan Hope's sociolinguistic analysis in the 1990s, and forcefully advocated by Brian Vickers in 2002. More recent computational work — especially Petr Plecháč's rolling attribution study published in Digital Scholarship in the Humanities in 2021 — broadly supports collaboration while introducing some refinements to the scene-level map.

Yet the question has also never been fully settled. A determined minority of scholars has maintained Shakespeare's sole authorship. Editors have generally been cautious about exact internal boundaries. And the methods used to argue for collaboration have shifted substantially over time: from verse metrics, to linguistic markers, to function-word profiles, to rolling computational attribution.

This project does not try to settle the full history of the question in one stroke. It does something more focused: it begins from the text itself rather than from a previously published map, it validates every method before interpreting results, and it asks what survives when suspicious scenes are divided into their local dramatic sub-units.

Why this is a hard problem

Four features of Henry VIII make the authorship question genuinely difficult, independent of any particular method:

1. No external documentary evidence. There is no contemporary document directly naming Fletcher as co-author. The collaboration hypothesis rests entirely on internal evidence — style, vocabulary, verse patterns, and linguistic habit.
2. The play contains multiple dramatic registers. Ceremonial pageantry, legal accusation, comic crowd scenes, intimate prophecy, and political counsel all appear in the same play. Different dramatic registers can themselves generate different stylometric profiles, independent of authorship.
3. The First Folio text may have been mediated. Compositorial practice, scribal transcription, and editorial normalization can affect the very features — spelling, contraction, function-word choice — that authorship methods depend on.
4. Methods disagree on units of attribution. Some scholars work at the scene level. Others work at the rolling-window level. Others work scene-by-scene. The scale at which you measure can determine what you find.

2. The History of Scholarship

Understanding the current project requires understanding what the existing scholarship has and has not established. The following is an expanded review of the major positions, with particular attention to their methodologies and their limitations.

Recent Literature (2020–2026)

Beyond the core texts above, the following recent scholarship is relevant to the present project:

• Plecháč (2021, DSH): The published version of the arXiv preprint (1911.05652). Full citation: "Relative contributions of Shakespeare and Fletcher in Henry VIII: An analysis based on most frequent words and most frequent rhythmic patterns." Digital Scholarship in the Humanities 36.2 (2021): 430–438. This is the most directly comparable computational study. The current project's rolling attribution result is considerably more conservative than Plecháč's — the key difference being the reference corpus and feature-family choice.
• Plecháč replication data (GitHub: versotym/henry8): Plecháč made all data and code publicly available. The present project did not use his code, but the availability allows direct methodological comparison. His method combines vocabulary frequency and verse rhythm; the current project uses character trigrams, function words, and linguistic markers, then adds CWFP as a content-word layer.
• Sharpe (2023, Shakespeare & Collaborative Writing, OUP): A systematic methodological and historical account of Shakespeare's collaborative practices. Includes Henry VIII and The Two Noble Kinsmen as central cases. Key contribution: framing collaboration attribution as a problem of convergence across evidence types rather than any single test — exactly the multi-method approach this project uses.
• Stanford Global Shakespeare Encyclopedia entry on Henry VIII: A useful recent summary of the scholarly consensus, noting that while most scholars now accept collaboration, "scholars using different methods have attempted to discriminate between Shakespeare's and Fletcher's parts without arriving at consensus."
• Analytics India Magazine (2019–2020) / TechXplore coverage of ML approaches: Popular-science coverage of machine-learning attribution claiming Fletcher wrote roughly half the play. These reports often overstate the certainty of rolling attribution results. The present project's validation-first design is partly a response to overconfident ML claims in this space.
• 2021 Stylometric reanalysis (Craig/Kinney critique): Subsequent scholarship has identified methodological problems with the Zeta test as implemented in the 2009 Craig/Kinney volume. This critique is relevant because it shows how sensitive attribution conclusions can be to feature selection and distance metric — a core concern of the present validation framework.

3. Data, Texts, and the Reference Corpus

The database

All analysis rests on a local SQLite database of 527 early modern English plays. The database stores spoken-token sequences, lemmas (via MorphAdorner), speaker information, act/scene structure, and bibliographic metadata. Lemmatization allows comparison of lexical habits independently of surface spelling variation, which matters significantly for a play like Henry VIII where compositor practice may have standardized or altered original spelling.

The target play is Henry VIII (database play ID 502, 36th play in the First Folio subset). A diplomatic Folger transcription was separately consulted for the original-spelling investigation.

The secure reference corpus

Author-comparison analysis uses a deliberately narrow set of secure reference plays. These are plays whose single authorship is not seriously disputed and which provide enough text for reliable window-level comparison. The corpus is not claimed to represent the entirety of either author's output — it represents plays secure enough to use as a calibration standard.

The corpus has 16 Shakespeare plays and 6 Fletcher plays. The asymmetry matters: the method will be more comfortable recognizing Shakespeare than Fletcher, and this is confirmed in validation. All conclusions are read against this known asymmetry.

Six governing assumptions

1. No previously published Henry VIII section map functions as the answer key.
2. Large scenes are not presumed homogeneous. If a scene changes local dramatic function, it may need manual subdivision.
3. No method is interpreted before being validated on secure single-author material.
4. No single feature family decides the case alone.
5. TNK can be used as a definitely mixed Shakespeare–Fletcher play, but not as a secure internal partition map.
6. Any chunk that weakens substantially under finer subdivision is treated cautiously, even if it looked persuasive at coarser scale.

4. How Henry VIII Was Manually Chunked

Why scene level was not enough

The project did not begin with the 46-chunk map. It began with scene-level exploration. That first pass immediately revealed a problem: many of Henry VIII's scenes are too long and too internally varied to be treated as single authorial units. A single scene in this play may contain a ceremonial opening, a legal accusation, a private emotional reaction, a packet exchange, and a formal conclusion — five dramatically distinct phases within one scene boundary. Attributing that kind of scene to one author is not just potentially wrong; it is not even a well-formed question.

The methodological response was manual chunking: inserting boundaries only where the dramatic action itself makes a local break intelligible, in the way a literary critic would notice a change even without any stylometric agenda.

Principles of chunking

A new chunk boundary was drawn when one or more of the following occurred:

• A speaker bloc changed (e.g., from public noble exchange to private royal reflection)
• The dramatic function changed (e.g., from ceremonial prelude to legal accusation)
• The local rhetorical mode changed (e.g., from banter to proclamation, or from report to prophecy)
• A scene contained an embedded unit that a literary critic would obviously discuss separately — even without any stylometric question

This means the chunk is best understood as a local dramatic unit, not a mechanical token slice. The result was a 46-chunk map across the play's 16 scenes.

Representative chunking examples

Figure 1. The 46-chunk map of Henry VIII. Colors indicate the working judgment that emerged from the first-principles ledger: Shakespeare-like (teal), mixed/transitional (amber), Fletcher-like (red), or too short to assess (grey). The strong Shakespearean backbone running through Acts 1–3 is immediately visible, as are the narrow Fletcher-like pockets in 1.3 and 5.3.

5. How The Two Noble Kinsmen Was (and Was Not) Used

Earlier drafts of this project were tempted to use The Two Noble Kinsmen (TNK) as a validation case: if a method can recover the familiar internal Shakespeare/Fletcher partition of TNK, then it can be trusted on Henry VIII. This is a tempting shortcut, but it is methodologically circular.

The standard TNK partition (Acts 1, 2, and 5 mostly Shakespeare; Acts 3 and 4 mostly Fletcher) is itself disputed and internally derived. Using it as a validation target means assuming precisely what you are supposed to test. If your method recovers the disputed TNK map, all you have shown is that your method agrees with a prior disputed conclusion — not that either is correct.

This is a narrower use of TNK, but a cleaner one. Under this rule, TNK is useful for the following question: Do Henry VIII and TNK both show mixed internal texture clearly above the single-author baseline? That question can be answered without assuming we know which TNK scenes belong to which author.

6. The Methods, Explained

Method 1: The Refined Stack

The main local attribution method is the "refined stack," combining three feature families. Each Henry VIII chunk is represented in three different ways, and compared against rolling windows of the same length drawn from secure Shakespeare and Fletcher reference plays.

Character trigrams (char): overlapping three-character sequences (e.g., hon, ono, nou from honour). These capture recurrent spelling shapes and some lexical habits without becoming a simple bag of full words. They are particularly useful because they are somewhat insensitive to morphological inflection.

Function-word profile (fw): high-frequency words like and, of, with, for. In authorship work, function words are prized because they are less topic-dependent than content words — a playwright uses the and that habitually, regardless of whether the scene is about love or politics.

Marker profile (marker): a small targeted set of authorship markers with a known prior in Shakespeare/Fletcher attribution — especially pronoun preferences, auxiliary verb choices, and contraction habits (see Method 4 below for detail).

For each feature family, the chunk is compared to reference windows using cosine distance. The best-matching windows on each side are identified. The method then asks: on average, is this chunk nearer to Shakespeare windows or Fletcher windows? If two of three feature families agree on one author, that is the chunk's reading. If they split or the margin is small, the chunk is Mixed.

Method 2: Rolling Attribution

The rolling-attribution pass asks the same authorship question continuously rather than chunk-by-chunk. A fixed-size window (400 or 700 tokens) is moved through the entire play one step at a time. Each window is compared to the secure reference corpus. The result is a continuous Fletcher-proxy trace — a number between 0 and 1 at each position, where higher values indicate closer proximity to the Fletcher reference plays.

This method is the direct local equivalent of Plecháč's rolling attribution. The key difference is the feature stack (trigrams + markers + function words here, versus most-frequent words + most-frequent rhythmic types in Plecháč) and the reference corpus (16 Shakespeare / 6 Fletcher here vs. a different selection in Plecháč).

The rolling method has two advantages over chunk-based analysis: it does not depend on chunk boundary placement, and it can reveal whether apparent Fletcher pressure forms a genuine continuous block or a brief local spike.

Method 3: Verse and Rhyme

The verse-and-rhyme layer reconstructs line-by-line structure from the token sequence and measures several lineation-sensitive features: feminine-ending tendencies, strong end-stopping, shared lines between speakers, and adjacent rhyme density. This layer matters because Spedding's original argument was already lineation-sensitive, and because several scholars (including Plecháč) have found verse patterns diagnostic.

However, in this project's validation, verse features perform only moderately on the secure reference corpus. Feminine endings achieve about 0.727 accuracy for the Shakespeare/Fletcher distinction. Shared-line and end-stop claims are weaker still. The verse layer is therefore used as a secondary check rather than a primary decision criterion.

Method 4: Classic Marker Audit

Rather than simply assuming that the traditional Fletcher markers work, the project tested them against the secure corpus. The question was: which traditional markers actually separate secure Shakespeare plays from secure Fletcher plays in the local reference data?

The strongest markers are 'em/them and has/hath, each reaching 0.955 accuracy. The ye marker is strong in principle but complicated by its sensitivity to database normalization — some lemmatization pipelines collapse ye and you, which would suppress the signal. Results using ye are read with extra caution.

Crucially, even strong markers behave unevenly across Henry VIII: they are often local within a scene rather than scene-wide. This is one of the most important empirical results in the entire project — it undermines the traditional assumption that a scene with several Fletcher marker hits is a uniformly Fletcherian scene.

Method 5: Original-Spelling Check

Because spelling arguments recur in the Henry VIII tradition (Hoy's work relied partly on ye and 'em in original spelling), the project returned to a diplomatic Folger transcription of the play. This layer was handled with caution. Ordinary Folio spellings — heere, beleeue, honour — are as likely to reflect compositor practice as authorial habit. More local colloquial forms — ye, 'em, ha's — are more informative but still potentially affected by scribe or compositor.

The main value of this layer is negative: spelling evidence can intensify a pre-existing suspicion, but it does not by itself settle a chunk.

Method 6: Content-Word Frequency Profile (CWFP)

The CWFP method asks a fundamentally different question from the refined stack. It ignores function words and linguistic markers entirely, and compares texts only through recurring content lemmas — words like honour, grace, blood, power, lord, love. The vocabulary is built from ~2,567 content lemmas that appear in at least 50% of reference plays.

CWFP is useful because it provides an independent signal. If a chunk shows Fletcherward pressure in both the refined stack and CWFP, that cross-method convergence is stronger evidence than either alone. CWFP is more vulnerable to subject matter (a politically intense scene will use different content words than a comic one), but it is less vulnerable to the marker-transmission problems that affect function-word approaches.

Method 7: Genre and Mode Contamination Test

The most important methodological question in the entire project is this: could the Fletcherward pressure in suspect Henry VIII chunks simply reflect their dramatic mode rather than their authorship? Court satire might look Fletcherward. Comic crowd scenes might look Fletcherward. Processional narration might look Fletcherward — regardless of who wrote the passage.

The contamination test directly addresses this objection. For each surviving pressure chunk, the project found exact-length windows from secure Shakespeare plays that matched the Henry VIII chunk as closely as possible in dramatic geometry: same approximate speaker count, same turn structure, same concentration of speech among a small number of speakers. These matched Shakespeare windows were then put through the refined stack and CWFP.

If the matched Shakespeare controls also drift Fletcherward, the contamination explanation is supported and the Henry VIII pressure is inconclusive. If the Henry VIII chunks are consistently more Fletcherward than their matched Shakespeare controls, the contamination defense weakens.

7. Validation Before Interpretation

The asymmetry in method performance is the most important single fact about the project's evidentiary standard. It must be understood before any Henry VIII result is interpreted.

Refined stack validation

Leave-one-play-out validation on the secure corpus gives the following results:

CWFP validation

Figure 2. Validation accuracy on the secure reference corpus for both main attribution methods. Both methods recover Shakespeare more reliably than Fletcher. This asymmetry means that a conservative Henry VIII result — one where relatively few chunks appear Fletcherian — cannot be taken at face value as evidence against collaboration. The method is easier on Shakespeare. Any positive Fletcher finding therefore carries more weight than an absence of Fletcher findings.

8. The Main Henry VIII Findings

The first-principles analysis produces a result that is neither the broad traditional Fletcher map nor an all-Shakespeare reading. The play has a genuine Shakespearean backbone — visible across multiple methods and across the manual chunking, the rolling attribution trace, and the content-word profile. But a handful of locally persistent Fletcherward pockets survive the full battery of tests.

The Shakespearean backbone

The Wolsey material — the political rise and fall across Acts 1–3 — remains consistently and strongly Shakespearean across the refined stack, rolling attribution, and content-word profile. Scenes like 1.1, 1.2, 2.4, and the entire 3.2 complex (Wolsey's fall and its aftermath) are among the most securely Shakespearean regions in any attribution analysis.

The legal-ceremonial scenes — particularly the trial of Katherine in 2.4 and the Cranmer vindication in Act 5 — are also strongly Shakespearean. The play's political-historical architecture appears to be Shakespeare's in its most fundamental design.

What looked Fletcherian and then weakened

Before local subdivision, several parent chunks looked suspicious. What happened to each under finer resolution is the key story of the project:

Figure 3. What happens when the main parent pressure chunks are divided into local sub-units. In most cases, the local sub-units are predominantly Shakespearean even when the parent chunk looked suspicious. This does not mean the pressure was imaginary — it means the old scene-scale labels were too coarse. The most important residue after subdivision is in 1.3a (partly) and 5.3a (strongly).

Figure 4. The classic Spedding/Hoy Fletcher scenes shown alongside the chunk-level working map from this study. The key observation: traditional scene-level assignments often cover regions that the chunk-level analysis shows to be internally divided or largely Shakespearean. The traditional map is not simply wrong — its pressure points are often real — but it is too coarse about the internal composition of those scenes.

Figure 5. The whole-play method matrix: for each chunk, the working ledger judgment, the refined stack result, the CWFP result, and the rolling attribution result at 400 and 700 tokens. The most important pattern: the four methods broadly agree on the Shakespeare-like backbone. Where they disagree most is in the middle of the play (Acts 2–3) and in the problem zones of 1.3 and 5.3.

Rolling Attribution: The Continuous View

The rolling attribution pass approaches Henry VIII from the opposite direction from manual chunking: it makes no assumptions about where dramatic units begin and end, and simply asks, window by window, whether each 400- or 700-token stretch is closer to the Shakespeare or Fletcher reference corpus.

The result is striking. The continuous trace is overwhelmingly Mixed, with only a very small number of decisively Fletcher-classified windows:

The rolling trace does not produce a broad alternating Shakespeare/Fletcher map. The play is not neatly divisible into large Shakespeare blocks and large Fletcher blocks. Instead, most windows sit in the mixed zone, with a small number of strong Shakespeare windows in the Wolsey and trial material, and a handful of Fletcher windows concentrated in 5.3a and, more weakly, 1.3a.

This is the most direct local answer to the Plecháč-style rolling question. The current result is considerably more conservative than Plecháč's. The key methodological differences: the current project's reference corpus is more carefully validated, and the feature stack (trigrams + markers + function words) is different from Plecháč's most-frequent-word plus verse-rhythm combination. Whether this difference reflects a genuine methodological improvement, or reflects the different feature families responding to different aspects of style, is itself an open research question.

Figure 6. The rolling attribution trace across Henry VIII. Each point represents one overlapping window; higher values indicate closer proximity to the Fletcher reference plays. The strong Shakespearean windows in Acts 1–3 (the Wolsey material) are clearly visible. The Fletcher spikes are narrow and localized — not the broad continuous Fletcher blocks that scene-level tradition implies.

9. The TNK Coarse Benchmark

Under the stricter evidentiary rule, TNK is used only to ask: do Henry VIII and TNK both show coarse mixed internal texture, clearly above the single-author baseline? The answer is yes for both plays — but in different ways.

TNK has a higher mixed share — more of its windows sit in the middle zone — but a lower switch rate and lower entropy. This means TNK looks steadily mixed, while Henry VIII looks more volatile: it swings more sharply between Shakespeareward and Fletcherward readings. Both plays are well above the single-author baseline in internal texture complexity.

This difference between the two plays is itself informative. If the method simply flattened all plays into undifferentiated ambiguity, Henry VIII and TNK would look identical. They do not. Henry VIII is more volatile; TNK is more uniformly mixed. That distinction is consistent with Henry VIII having a strong Shakespearean backbone with a smaller number of concentrated Fletcherward pockets, while TNK has more evenly interleaved authorial contributions.

Figure 7. Rolling Fletcher-proxy traces for Henry VIII and TNK at 700-token windows. TNK (right) is more steadily mixed; Henry VIII (left) fluctuates more sharply. The narrow Fletcherward spikes in Henry VIII contrast with TNK's more uniform mixed texture. Note that the current project does not use this comparison to locate Fletcher in TNK — it uses it only to ask whether both plays are recognizably mixed relative to single-author plays.

Figure 8. Whole-play mixed-texture metrics for Henry VIII, TNK, and the secure single-author reference plays. Both Henry VIII and TNK sit clearly above the single-author plays in mixed-share and several variance metrics. The difference between them illustrates that the method is sensitive to the character of the mixing, not just its presence.

10. CWFP Results

The CWFP pass provides a new kind of evidence because it is based on common content words rather than function words or markers. A chunk that shows Fletcherward pressure in CWFP cannot be dismissed by arguing that the refined stack merely reflects marker frequency or function-word habit.

Summary of CWFP classifications

CWFP does not restore the traditional large Fletcher share. Only one chunk receives a clear Fletcher classification — 5.4b (Cranmer's first prophecy movement), whose nearest CWFP neighbor is Fletcher's Valentinian. But the margin pattern is more informative than the classification count: many chunks that CWFP classifies as Mixed are slightly closer to the Fletcher reference than to Shakespeare, and these subtle margins converge with other methods in consistent ways.

Most Fletcher-leaning by CWFP

Most Shakespeare-leaning by CWFP

Figure 9. The surviving pressure chunks under the two main attribution methods. Negative margin means the chunk is more Fletcherward. The overlaid points show the number of validated Fletcherward markers in the same chunk. The convergence of negative CWFP margin, negative refined margin, and high marker count in the same chunks is the strongest available evidence for locally persistent Fletcherward pressure.

CWFP–Marker Crosswalk

Cross-method convergence is the strongest form of evidence available. The following chunks show both high validated marker hits (≥3) and a Fletcherward CWFP margin:

Note that several of these chunks — 2.2b, 3.2d, 4.1a, 4.1c — are classified as Shakespeare-like by the refined stack despite having high marker counts and Fletcherward CWFP neighbors. This is an important interpretive point: it suggests that the marker and content-word signals in these chunks may be picking up dramatic mode (ceremonial narration, counselor banter, public accusation) rather than authorial identity. The genre/mode contamination test was designed precisely to investigate this possibility.

11. The Genre/Mode Contamination Test: Results

This is the methodological crux of the project. For each surviving pressure chunk, the test found 20 exact-length windows from secure Shakespeare plays with the closest available dramatic geometry (speaker count, turn structure, speaker concentration). Those confirmed-Shakespeare windows were then run through the refined stack and CWFP. If they also drift Fletcherward, the Henry VIII pressure could be dismissed as a mode effect. If they do not, the contamination defense weakens.

The percentile column is the key number. A percentile of 0.000 means the Henry VIII chunk is more Fletcherward than every single one of the 20 matched Shakespeare controls. Most of the live pressure chunks achieve this. This is a strong result: it means the contamination objection cannot explain away the pressure in these specific locations.

Figure 10. Percentile of each live Henry VIII pressure chunk among its matched Shakespeare controls. Lower percentile means the Henry VIII chunk is more Fletcherward than most of the Shakespeare controls matched to it in dramatic geometry. The near-zero percentiles for 1.3a, 5.3a, 3.1b, 3.2d, 4.1c, and 2.2b are the most important numbers in this chart.

12. Comparing the Current Map to the Traditional Scene Partition

The following table shows each of the sixteen scenes in Henry VIII, indicating whether traditional scholarship assigned it to Fletcher, how many validated marker hits the current analysis found, and how the current chunk-level analysis distributes the scene's local units.

The most striking pattern: among the nine scenes that Spedding assigned to Fletcher, the current chunk-level analysis finds only two scenes that contain any F-like chunks (1.3 and 5.3). The other seven scenes that Spedding called Fletcher territory contain chunks that resolve as Shakespearean or Mixed under the current analysis. This does not mean Spedding was entirely wrong — his pressure instincts were often correct at the scene level. It means the Fletcher signal within those scenes is often local rather than scene-wide.

13. Conclusions

The findings can be stated in a way that is both cautious and affirmative. They do not require choosing between "Spedding was right" and "the play is all Shakespeare." A more specific and better-supported picture is available.

What would change the picture

The following findings would substantially revise the current conclusions:

• A larger, better-balanced Fletcher reference corpus (currently only 6 plays) that improves Fletcher recovery in validation — this would allow stronger positive Fletcher claims
• A validated spelling investigation on the diplomatic Folger text that shows systematic colloquial contraction patterns aligning with the pressure chunks — this would strengthen the marker evidence
• A Plecháč-style replication that uses the same feature stack as the current project but finds broader Fletcher blocks — this would indicate the difference is in the reference corpus rather than the features
• Newly discovered documentary evidence linking Fletcher to the play — this would remove the inference problem entirely, though it would not settle the internal map

Appendix: Complete Chunk-by-Chunk Method Ledger

The following table gives all 46 chunks with their scene assignment, token count, short label, and judgments from the four main analysis layers: the refined stack, CWFP, rolling attribution at 400 and 700 tokens, validated marker hit count, and the working judgment that synthesizes across methods.

Works Cited and Comparison Points

• Spedding, James. "Who Wrote Shakespeare's Henry VIII?" Gentleman's Magazine (1850). Princeton archive.
• Hoy, Cyrus. "The Shares of Fletcher and His Collaborators in the Beaumont and Fletcher Canon." Studies in Bibliography 8–15 (1956–1962).
• Hope, Jonathan. The Authorship of Shakespeare's Plays: A Socio-linguistic Study. Cambridge: Cambridge UP, 1994. Cambridge chapter.
• Vickers, Brian. Shakespeare, Co-Author: A Historical Study of Five Collaborative Plays. Oxford: OUP, 2002. OUP.
• Craig, Hugh and Arthur F. Kinney. Shakespeare, Computers, and the Mystery of Authorship. Cambridge: Cambridge UP, 2009.
• Plecháč, Petr. "Relative contributions of Shakespeare and Fletcher in Henry VIII: An Analysis Based on Most Frequent Words and Most Frequent Rhythmic Patterns." Digital Scholarship in the Humanities 36.2 (2021): 430–438. arXiv preprint · Published version · Replication data.
• Sharpe, Will, ed. King Henry VIII; or All is True. New Oxford Shakespeare. Oxford: OUP, 2016.
• Sharpe, Will. Shakespeare & Collaborative Writing. Oxford Shakespeare Topics. Oxford: OUP, 2023.
• Margeson, John, intro. Henry VIII. Cambridge: Cambridge UP. Cambridge appendix.
• Folger Shakespeare Library. Henry VIII Appendix on Authorship. Folger.
• Stanford Global Shakespeare Encyclopedia, "Henry VIII" entry. Stanford.

Henry VIII Authorship Project · March 2026 · Codex research workspace
Database: early_modern_plays.db · 527 plays · Target: play ID 502 · Chunks: 46