What Would Settle It

Two kinds of evidence could settle who wrote the Shakespeare works. The more direct is a contemporary record of the writing: a manuscript, a letter about the work, a payment for it, or a reference by someone in a position to know. A record of that kind counts whether or not it names the writer, provided it can be tied to one, and it would still need authenticating. It has to record the writing itself. A sonnet, dedication or allusion that has to be interpreted before it points to anyone is a source being read, and it counts as circumstantial evidence.

The other is circumstantial evidence that discriminates. Courts convict on it and historians attribute anonymous works on it. In this question, where correspondences can be found for every candidate, it would have to meet five conditions:

1. Primary sources. The facts a case uses must be documented in primary sources from the period, as defined on the Two Frameworks page, not taken from later tradition or secondary accounts; reasoning comes in when the facts are put together. A match between a passage in the works and such a source counts only if a reader can check it without first accepting the reading: shared wording, or a detail found in that source and not, so far as anyone has shown, in anything else then available. A match that depends on interpreting the passage (reading Polonius as a portrait of Burghley, say) is a reading, and counts as nothing more.
2. Base rates. A fact counts for a candidate only as far as it doesn’t also fit many other people. A classical education or years at court fits hundreds, so it narrows the field very little. Each fact therefore comes with a rough estimate of how common it was among literate English adults of the period (1558 to 1623), the same group for every candidate. Nobody has to list that group. The estimate comes from records that already exist: for attendance at an Inn of Court, the admission registers, which ran to hundreds a year; for a peerage, the roll of peers. A rough figure with its working shown is enough, and anyone can challenge it. The group is defined by time and literacy alone, never by rank, education, occupation, travel or any other feature a case relies on. Where records of the writing narrow the field, the group is the people those records leave open (see condition 5). Where the evidence allows only a range, the case uses the end that makes the fact more common, so a fact counts only if it still narrows the field on the cautious figure.
3. Shared causes counted once. Facts that come from one cause count once. For someone of noble birth, a classical education, French, time at court, a knowledge of falconry and heraldry, and travel abroad commonly followed from that birth, and where they do they count as one. Where a candidate’s documents show that one of them came from something else, such as a profession, an office or an embassy, it counts on its own. This site’s own convergence table applies the same rule.
4. A control. Before a method is trusted to find the author, it should be tested on people nobody proposes as the author, to see whether it picks them out too. The advocate writes down the rules first: the features the case looks for, what counts as a match, and how many matches identify someone. Someone else then chooses a few well-documented contemporaries for the test, and the same rules are applied to them. If no one independent will choose them, this site will, from people nobody has proposed as the author, and will publish its choice before the rules are applied. If the rules find matches for them as readily as for the candidate, they aren’t identifying anyone. Ideally the person applying the rules doesn’t know which name is the candidate, since anyone who knows the intended answer tends, without meaning to, to read matches generously for that person; where that isn’t possible, someone with no stake in the result applies them and records a reason for each judgement. Choosing the controls and applying the rules are judgements for a person: an AI model can do neither, since whoever writes the prompt decides what the model is shown and how the question is put, and a model shown a case’s documents will often recognise the candidate. Every judgement is published, and the rules can’t be changed once the controls are chosen.
5. Singling out. Ruling rivals out is not enough, since a list of rivals ruled out still leaves everyone who was not on the list. The facts have to point to the candidate themselves. Taken together, with shared causes counted once, they should be rare enough that, on the base rates in condition 2, you wouldn’t expect even one other person in the group to fit them all. If the working suggests there could be others, the case hasn’t singled anyone out.
   
   When records narrow the field. Records of the writing sometimes name more than one possible writer, or conflict: two men each claim the same essay, say. Neither record settles the question alone, but together they limit who the writer can have been. The group is then the people the records name, and a circumstantial case has only to single out one of them. A record narrows the field only if it would count as a record of the writing in its own right: tied to the particular work, authenticated, and made while the person it names was alive, by someone in a position to know. It must also identify a person. A pen name identifies no one, and where the question is who stood behind a name, the name narrows nothing. The rule holds in every direction: a name on a title page cannot by itself narrow the field to the man from Stratford, and a pen name, initials or an anagram cannot narrow it to anyone else.

Why these five

The five follow from one principle. Evidence points to an author to the extent that it is much more likely if that person wrote the works than if someone else did. Each condition protects one part of that comparison, and each blocks a way attributions are known to have gone wrong.

• Condition 1: the Collier forgeries. In 1852 John Payne Collier produced a copy of the Second Folio (1632) with manuscript corrections he said were by an “old corrector”, and published them as Notes and Emendations to the Text of Shakespeare (1853). In 1859 Sir Frederic Madden and Nicholas Hamilton pronounced the corrections “incontestably forgeries of modern date”; the old-looking hand was in sepia paint laid over erased pencil notes in a modern hand closely resembling Collier’s. He had also interpolated Shakespeare’s name into a genuine letter at Dulwich. In each case the conclusion came first and the document was made to carry it.
• Conditions 2 and 3: Sally Clark. At her 1999 trial an expert put the chance of two cot deaths in one family at 1 in 73 million, by squaring the figure for one. The Royal Statistical Society’s statement of 23 October 2001 said the figure “has no statistical basis”, that it “would only be valid if SIDS cases arose independently within families”, and that reading it as the chance the deaths were accidental was “a serious error of logic known as the Prosecutor’s Fallacy”. The fallacy is to mistake how rare the evidence would be if the accused were innocent for how likely it is that the accused is innocent. In authorship terms, a fact that fits one Elizabethan in a thousand does not make it a thousand to one that the candidate wrote the works, because in a population that large it still fits a great many people. Her convictions were overturned at her second appeal, in January 2003.
• Condition 4: the Bible code. Witztum, Rips and Rosenberg (Statistical Science, 1994) claimed Genesis encoded information about later rabbis. McKay, Bar-Natan, Bar-Hillel and Kalai (Statistical Science, 1999) found comparable “significance levels” in a Hebrew translation of War and Peace, and traced the original result to “wiggle room” in how names, dates and methods were chosen.
• Condition 5: the Funeral Elegy. Donald Foster attributed “A Funeral Elegy” by “W.S.” to Shakespeare in 1989. In 2002 he withdrew the attribution in favour of John Ford. On David Kathman’s account, Foster had “assumed that the initials W.S. were in fact those of the author, and thus didn’t investigate Ford and other non-W.S. candidates fully enough”.

The conditions can be met. Twelve of the Federalist essays, published under the pen name Publius, were claimed by both Alexander Hamilton and James Madison: Hamilton in a list he left before his fatal duel with Aaron Burr, Madison in the information he gave for the 1818 Gideon edition. The pen name narrows nothing, but each list names a person in a position to know, so the group for those essays is Hamilton and Madison. Mosteller and Wallace (Journal of the American Statistical Association, 1963) counted small function words, avoiding words such as “war” whose rate depends on the subject, and took each man’s rates from his undisputed essays: “upon”, for example, runs at about 3 per thousand words in Hamilton and about 1/6 per thousand in Madison (condition 2). Anyone can check counts of that kind without first accepting a reading (condition 1). They treated the words as independent, measured how far they were not, and cut their odds by about 9 per cent to allow for it (condition 3). Before turning to the disputed essays, they tested the method on known Hamilton and Madison essays that had played no part in choosing the words, and it separated the two men without an error (condition 4). It gave all twelve disputed essays to Madison, No. 55 with the weakest odds, about 80 to 1, singling him out from the only other writer the records leave open (condition 5). Scholars now generally accept the attribution.

For Shakspere, a case that met all five would be strong enough for this site to treat the question as settled, and it would say so. For anyone else, it would identify the author, as a contemporary record naming the writer would, but the question would not be settled until Shakspere’s part was explained (below): why his name was on the title pages, and why he left no literary paper trail. None has yet met the five conditions. More parallels of the kind that fit the rivals too add length without adding discrimination.

Statements inside the works

A first-person statement in a disputed work cannot by itself identify the work’s author. The speaker of a poem may be a persona, and who the author was is the point in dispute. A statement can still count when it points to something outside the work, and a reader can check the link without first accepting the reading; it is then circumstantial evidence and has to meet the five conditions like any other. Forensic physical evidence that is consistent with a candidacy but stops short of recording authorship is in the same position.

Spenser used the persona of Colin Clout in The Shepheardes Calender, Colin Clouts Come Home Againe and The Faerie Queene, a transparent self-portrait documented from outside sources, yet no one argues that Colin Clout’s statements prove Spenser’s authorship. We know Spenser wrote the poems from evidence outside them.

* * * * *

The same standard for every candidate

An enquiry into a question of fact should be able to say in advance what would change its mind, and this page is that statement for the authorship question. It also answers the charge that the question is a conspiracy theory, one that absorbs any counter-evidence: a question with stated conditions for settling it can be settled. On why that matters, see On Knowing.

The same standard applies to every candidate. Two conditions must be met simultaneously to settle the question. The first is evidence that settles who wrote the works: a primary source document connecting a named individual to writing under the Shakespeare name, or circumstantial evidence that meets the five conditions above. The second, required for any alternative candidate, is an account of Shakspere of Stratford: why his name appeared on the title pages, what his role was, and why he left no primary source evidence of a literary career of any kind. A theory that satisfies the first condition without addressing the second has not settled the question; it has redistributed it. The frameworks this site applies are tools for assessing the state of play in the absence of such evidence; they record how much documented evidence each candidate has, but they do not say who most likely wrote the works, and they cannot substitute for meeting both conditions.

Stated as an asymmetry: the authorship question and the Stratford anomaly collapse into a single problem only if the document names Shakspere himself, the attribution closes and the missing literary record is supplied in the same stroke. A document naming any other candidate closes the attribution and leaves the anomaly standing, sharpened: an unexplained absence becomes an unexplained presence. One discovery would end the enquiry; the other would end only half of it. This distinction is stated in full under One Unresolved Condition on the Two Frameworks page.

What a strong convergence result would look like: a strong convergence result would be one in which the lines met all five conditions above. The convergence table meets conditions 1 and 3 but not 2, 4 or 5, which is why no total on it is evidence of authorship.

No candidate yet meets these conditions, Shakspere included. The analytical findings are on the What the Evidence Returns page.

* * * * *

The Stratford silence

The silence is a structural problem that any alternative theory must address. The name ‘Shakespeare’ on the title pages was associated with a man apparently embedded in the professional theatre who left no literary trace whatsoever, no letters, no record of books owned, no payments for writing, no literary mourning at death.

Either evidence connects him to the writing (a document, or a circumstantial case that meets the five conditions), which would close the question in his favour, or a documented comparable pattern is established showing that the silence is not anomalous for a writer with his specific profile. Without one of these two outcomes, the silence remains an open evidentiary problem regardless of how well another candidate scores.

The existence of hundreds of Shakespeare biographies does not satisfy this condition. Most biographies of Shakspere make one of two moves: they use the works as biographical evidence, inferring travel, legal training, or court access from what the plays contain, and then offer the resulting profile as support for the attribution. That reasoning is circular. Or they argue that the primary source silence is normal for writers of the period. Diana Price’s survey of twenty-four of Shakespeare's literary contemporaries shows it is not: every one of them left primary source evidence of a literary career in at least one of Price’s ten categories. Shakspere did not; he is the only one of the twenty-five for whom no such evidence exists. The biographies have not closed the evidentiary gap; they have largely worked around it.

* * * * *

A concealed author

Many alternative cases assume the author wrote under a pen name and hid his part. That does not put him beyond this page’s standard. A concealed author would still have lived a documented life, and circumstantial evidence drawn from that life can identify him if it meets the five conditions; the Federalist essays were published under a pen name too.

What concealment cannot do is count in its own favour. If missing documents are exactly what a pseudonym would predict, no missing document can ever count against the claim, and the claim is safe from any evidence. A hypothesis protected that way cannot be tested.

The comparison group cuts the ground from under the objection at its root. Price’s twenty-four contemporaries left primary source evidence of a literary career in at least one of her ten categories, despite also sometimes publishing anonymously or under initials. Pseudonymity and anonymity were common in the period, yet the documents survived anyway. The absence for Shakspere is not explained by pseudonymity as a general condition of Elizabethan literary life. It is anomalous within a comparison group subject to identical conditions.

* * * * *

For Shakspere of Stratford

A record of the writing

Any one of the following primary source documents, if authenticated, would settle the question for Shakspere of Stratford.

A manuscript (any manuscript) of a Shakespeare play or poem in Shakspere’s own hand. This is the single most direct possible evidence. No such manuscript has been found for any Shakespeare work, in any hand, and after four centuries of motivated searching by scholars on both sides, finding one now is as close to impossible as historical research allows. It is listed here not as a research goal but as the theoretical gold standard, the thing that would settle the question beyond argument if it existed. The more tractable evidence types follow.

A letter in Shakspere’s hand discussing a play or poem he was writing, referencing a character, a plot, a patron, a performance. His surviving handwriting, if the Hand D attribution in Sir Thomas More is accepted, amounts to three pages of a collaborative manuscript. No letter of any kind in his hand has been found.

A payment record connecting him specifically to the act of writing a named work, not performing, not shareholding, not property transaction. A payment from a printer, a patron, or a theatre company for a specific play or poem, with his name attached as author.

A contemporary reference, produced during his lifetime, explicitly naming the man from Stratford as the author of specific plays, from someone in a position to know. Not a reference to the name on a title page, and not posthumous testimony. A document produced while he was alive to confirm or deny it.

The circumstantial case

The five conditions apply to the traditional attribution as they do to every other. The orthodox case begins with an identification: every reference to the author Shakespeare (the title pages, Meres’s list, the praise of contemporaries) is treated as a reference to William Shakspere of Stratford. No primary source from his lifetime makes that identification. The lifetime records of the Stratford man concern property, business and the acting company; the lifetime references to the author name no place, family or property that would tie him to Stratford (The Name Identification Assumption).

The orthodox case then points to documents from after his death: the First Folio, whose verses call the author “Sweet Swan of Avon” and refer to “thy Stratford Moniment”, and the monument at Holy Trinity. Neither names William Shakspere of Stratford as the writer of the plays. “Avon” can point to Stratford or, as Camden records, to Hampton Court, and Digges’s line and the monument associate the name with Stratford; the orthodox case reads that association as the identification, which is the same assumption, made after his death. It joins these to the company records and to the will’s interlined bequest to his “fellows” Heminges, Burbage and Condell (see The Case for Shakspere and the linking documents), which connect the Stratford man with the acting company.

Measured against the five conditions, each document is sound, but the link that matters is supplied by the identification instead of being shown in the lifetime record (condition 1). The case points to one man only once the identification is granted (condition 5). Company membership fits every sharer in the company, so it cannot make the identification on its own (condition 2). The posthumous witnesses are not independent of one another: Digges’s line points to the monument, and the Folio’s verses and its address to the readers come from one publishing enterprise (condition 3). Condition 4 has no direct counterpart for testimony of this kind; the nearest test would be how often posthumous identifications in the period proved wrong, and this site has not measured that.

Against all this stands a lifetime record that, alone among the twenty-five writers Price examined, holds no trace of a writing life (The Documentary Problem). No primary source connects him with the act of writing, and a single one would settle the question at once.

For any alternative candidate

The standard is identical. The same type of evidence that would settle it for Shakspere would identify any alternative candidate as the author, and the question would then be settled once Shakspere’s part was explained (above). This symmetry is the governing principle.

A record of the writing

A manuscript of any Shakespeare play or poem in their hand. The same caveat applies: after four centuries of searching, none exists, and none is likely to be found. It remains the theoretical gold standard. A letter in their hand acknowledging writing under the Shakespeare name, or discussing the works as their own. A will, diary entry, or private correspondence explicitly attributing the works to them, produced by someone with direct knowledge. A primary source connecting them to the publication of a Shakespeare work, a contract, a payment, printer’s correspondence identifying them as the author.

The Audley End forensic attribution, annotations in books at Audley End House attributed by forensic analysis to de Vere’s hand, with correspondences to the Roman plays, is the closest any candidate has come to this standard. It is primary source evidence attributed to a candidate’s own hand, connected to the Shakespeare canon by textual analysis. It records his reading, and falls short of a manuscript or a statement of authorship. Full analysis of the Audley End finding.

Thomas North’s annotated manuscripts, identified by Dennis McCarthy and used to argue that North wrote source plays later revised into the Shakespeare canon, represent a different kind of primary source argument. If the source-text fingerprint methodology is validated by independent scholars, it would constitute strong evidence for North’s involvement, though the question of revision versus authorship would remain open.

The circumstantial case

The five conditions for circumstantial evidence apply to every alternative candidate in the same way as to the traditional attribution. None of the cases assessed on this site reports base rates or a control for its evidence, and on this site’s assessments none yet meets all five; each candidate’s page sets out what the case rests on.

* * * * *

What would not settle it

Correspondences without base rates. A correspondence is consistent with a hypothesis without confirming it. Until each comes with an estimate of how many other people it would fit (condition 2), a catalogue of them, however large, does not add up to a probability. The correspondence problem is examined in full on the Advocates page.

Stylometric studies that take the canon as the measure of Shakespeare’s style. These presuppose the attribution they claim to test: the training corpus used to characterise “Shakespeare’s style” is defined by the traditional attribution. A study of that kind cannot independently confirm the attribution without circularity. What it can do, usefully, is identify collaborative passages within a work already attributed to Shakespeare. A study that compares a candidate’s own undisputed writing with the works presupposes nothing about who wrote them; it is circumstantial evidence like any other and has to meet the five conditions, as the Federalist study did (above).

Cryptographic claims. No Baconian cipher, acrostic, or numerical code embedded in the Folio has been verified by an independent cryptographer. The methodology is not replicable and the conclusions are not falsifiable by the normal standards of cryptanalysis.

Arguments from possibility. It is not impossible that Shakspere made undocumented trips to Italy, acquired undocumented legal training, maintained undocumented literary friendships. Possibility establishes only that a hypothesis has not been disproved; it says nothing about probability.

Arguments from tradition. Four centuries of attribution is a sociological fact, not an evidentiary one. Traditions can be wrong, and the question of whether this one is supported by primary source documentation is precisely what this site examines.