Method & Standards

Reading Without a Reference Point

How to judge a historical case when you cannot check the facts yourself

About This Page

Here is a predicament almost every reader of the authorship debate will recognise. You read a long, carefully footnoted reconstruction (a chain of names, dates, letters and motives that ends by identifying the author) and by the last page you are convinced. Some weeks later you read an equally careful piece that reaches a different conclusion, and you are convinced by that one too. You have no specialist training in Elizabethan history to fall back on, no way to check whether the archival reference on page nine says what the writer says it says. Absent that background, how are you supposed to tell a sound case from a merely persuasive one?

The reassuring part of the answer is that you do not need the specialist background to make real progress. Most of what separates a strong historical argument from a weak one is structural, and structure can be judged from the armchair. This page is a small set of tests you can run on any reconstruction (for any candidate, including the orthodox one) without leaving your chair or verifying a single fact. It is the companion to The Self-Check: that page is for testing a conviction you already hold; this one is for testing a case someone else is putting in front of you.

1. Start by distrusting the feeling of fit

The sensation these pieces produce (everything clicking into place, each detail confirming the last) feels like the argument working. It is not. Internal coherence is the cheapest thing in the world to manufacture: any competent writer can build a story in which every fact supports every other, and a clever writer can do it for two incompatible conclusions on consecutive weekends. That is precisely the situation you found yourself in.

A physicist has the right instinct here already. Give a curve enough free parameters and it will pass through every data point you own, fitting the past perfectly while predicting the future not at all. A historical narrative with enough latitude to invent motives, assume undocumented meetings and re-date events on demand is that over-fitted curve. It fits because it was shaped to fit. Coherence is a necessary condition for a theory to be taken seriously, and almost worthless as a way of telling a true theory from a false one, because the false ones are coherent too. The feeling of fit is the first thing to set aside, not the thing to trust.

The test: notice when a piece is making you feel that it all hangs together, and treat that feeling as a signal to slow down rather than a reason to assent. The question is never whether the story is coherent. It is whether anything in it could have come out otherwise.

2. The one question that does the real work

If you take a single tool from this page, take this one. For each striking fact a case offers, do not ask “does this fit the theory?” Almost everything fits; that is the problem above. Ask instead: is this fact more to be expected if the theory is true than if the ordinary account is true? Only the gap between those two expectations is evidence. A detail equally at home under both accounts carries no weight in favour of either, however vivid, ingenious or often-repeated it is.

In your own language, this is a likelihood ratio, and it is the whole of the matter. The evidential weight of an observation is how much more probable it is under one hypothesis than under its rival. When those probabilities are equal the ratio is one, and a ratio of one is silence (zero information) no matter how loudly the fact is presented. Vividness is not weight. A well-lit coincidence is still a coincidence.

Two illustrations, cutting in opposite directions so the tool stays honest. Suppose a case leans on the wide classical learning displayed in the works. Reading in Ovid and Seneca is certainly expected if candidate A wrote them, but it is just as expected if candidate B did, or the traditional author did, because every serious candidate was steeped in the classics. The observation is equally probable all round; its likelihood ratio is one; it discriminates nothing, and the weight the author places on it is borrowed, not earned. Now suppose instead a case turns on a dated document that would exist under one account and simply could not exist under the other. That observation is far more probable under the one than the other, and it does real work. Same reader, same armchair, opposite verdicts, arrived at by asking one question.

The test: for every “look how this fits” moment, ask what the ordinary explanation would predict for the same fact. If it predicts the fact just as comfortably, cross the item out. It is decoration, not evidence, and a case is only as strong as the items that survive this crossing-out. This is the reader’s form of the point made about stacked correspondences in Logic Errors.

3. Underline the verbs

This one is mechanical, needs no historical knowledge whatever, and is the fastest way to see the skeleton of an argument. Read with a pen and mark every verb that carries the weight of a step. Two vocabularies will separate on the page. One is the language of record: the register lists, the letter states, the parish entry records, the accounts show. The other is the language of inference supplied by the author: I conclude, I surmise, it is reasonable to suppose, he would have, someone must have, probably, no doubt, surely.

Neither vocabulary is illegitimate. History cannot be written without inference, and a good argument earns its inferential steps. What you are checking is where the load sits. Follow the spine of the case (the chain that carries you from the documents to the conclusion) and see which vocabulary its load-bearing links are written in. If the documented verbs decorate the edges while every joint that actually holds the structure up is a “would have” or an “I surmise,” you are looking at a tower of speculation wearing a thin coat of fact. The coat is real. It is not what is holding the tower up.

The test: trace the argument from its evidence to its conclusion and count how many of the connecting steps are documented and how many are supplied. A case whose every critical junction is an inference has told you a story about the evidence; it has not been compelled by it.

4. Count the moving parts

A rich reconstruction often needs a great many things to be true at once: a hidden identity here, an undocumented journey there, a name read as a pun, a date quietly shifted, a coincidence of initials, a second secret to explain the first. Each of these is a free parameter, a dial the author was allowed to turn to make the story fit. The more dials a case needs, the weaker it is, even as it grows more absorbing to read, because a machine with enough adjustable parts can be tuned to produce any output at all.

You can literally count them. Every time the argument requires you to grant a fresh supposition (not a documented fact, but a “let us suppose” the rest of the case then leans on), add one to the tally. A conclusion reached with two or three such grants is in a different class from one that needs a dozen interlocking ones, each holding the next in place. Anagrams, hidden initials and decoded patterns belong in this count too, and they are the cheapest parameters of all: a long text will yield almost any short name you go looking for, as the quantitative test of cipher methods on this site demonstrates.

The test: tally the number of independent things the case needs you to accept without documentation for its conclusion to stand. Prefer the account that needs the fewest. Elaboration is not depth; frequently it is the sound of a theory paying for its own survival.

5. Follow the footnotes to their source

You cannot travel to the archive, but you can still tell a great deal about the citations without reading them. Two properties matter, and both are visible from the armchair. The first is whether a reference is in principle checkable: a named document with a shelfmark, a dated register entry, a specific page (something that could be produced and examined by anyone who disagreed). The second is whether the references are independent of the case they support, or whether they circle back to the same small group of advocates writing in the same venues.

A tower of footnotes that all lead, eventually, to earlier work by the same author and their allies is a closed loop, not corroboration. It can give an argument the texture of deep scholarship while adding no independent support at all, because every strand traces back to the same origin. Genuine corroboration comes from sources that had no stake in the conclusion and no contact with one another and still point the same way. When you scan the notes, you are not verifying the history; you are checking whether the history could be verified by someone who wanted to, and whether the support comes from outside the circle of believers or only from within it.

The test: ask of the citations, could a determined sceptic look this up, and do these sources include any that had no interest in the answer? Footnotes that are neither checkable nor independent add authority to the eye without adding evidence to the argument.

6. Ask what would have changed the author’s mind

A healthy argument can tell you what evidence would have counted against it. Watch how a case treats the things it lacks. If the absence of the letters you would expect, the silence of the contemporaries who should have spoken, and the missing manuscript are each folded in as proof of how well the secret was kept, then no possible finding could have told against the theory: every gap has been converted into confirmation. A structure that absorbs every outcome as support for itself is not being tested by the evidence. It is immune to it.

This is the reader’s version of a point the site makes at length about the ad hoc defence and the argument from silence. The productive question is not “could the evidence have been suppressed?” (in any historical inquiry it could) but “is there any finding this author would have accepted as telling against the theory?” If you cannot name one, and the piece never names one either, the coherence you admired in it is the coherence of a system that cannot lose. Full treatment: the ad hoc defence and the argument from silence.

The test: try to state, in one sentence, the discovery that would have forced this author to give up the theory. If no such discovery exists (if every absence becomes concealment and every presence becomes reach) you are reading an argument built so that nothing can refute it, which is a different thing from an argument that nothing has refuted.

7. Note the venue, once

Where a piece appears is worth a single, measured thought, no more. An argument published by a society founded to advance the very conclusion it reaches, reviewed by readers who already share it, has not yet passed through the one process that kills most clever-but-wrong ideas: sustained scrutiny by people who would be glad to find it wrong. That does not make it false, and holding the venue against the argument would be its own error; the content stands or falls on the evidence, not on the masthead. What the venue tells you is only how much adversarial pressure the case has already survived, which bears on how much of your own you should apply.

The test: let the venue set your level of scrutiny, never your verdict. A friendly forum is a reason to test the argument harder yourself, not a reason to dismiss it, and never a substitute for the six questions above.

What this leaves you with

None of these tests tells you who wrote the works. That is not their job, and it is not a failing that they decline to. What they give the non-specialist reader is the thing the predicament at the top of this page was missing: a way to grade a case on something other than how good it feels to read. A reconstruction that survives them (whose striking facts discriminate rather than merely fit, whose spine is documented rather than surmised, which needs few grants rather than many, whose sources can be checked and are independent, and which could in principle have been proved wrong) has earned a measure of your assent. One that convinces only because none of these questions was asked of it has charmed you, which feels identical from the inside and is not the same thing.

The tests apply with exactly equal force to the orthodox account and to every alternative. A vividly written biography that reads the author’s inner life out of the plays and back onto a documented man is running the same move as the most elaborate pen-name reconstruction, and it answers to the same questions. That symmetry is the point. For the underlying standard this site applies to every candidate once the reading is done, see The Two Frameworks; for the same audit turned on a conviction you already hold, see The Self-Check; and for the named forms these errors take, see Logic Errors in the Authorship Debate.