New Gorhambury·Baconian Ciphers

One-Day Report · Biliteral Cipher · September 8, 2026

Today's claim

Every lower-case italic w in the thirty-six plays — 3,080 of them by the Bodleian’s own transcription — was found on the photographs and measured against both of Riverbank’s stencils. In the fount the plays are set in there is one w, not two, and no third shape either. The bi-formed w exists on four pages of a larger italic — the case the preliminary leaves are set in, and the case Riverbank cut its alphabets from. Alongside it: Friedman’s name is on those plates twenty-two times, as the draughtsman and never as the author of a claim; and Riverbank’s boast that a fourteen-year-old can read the cipher turns out to be scored against Riverbank’s own answer.

findings may be superseded by later reports

The short version

In 1899 Elizabeth Wells Gallup said the First Folio carried a hidden message in the shapes of its italic letters: two founts so alike that no reader notices, alternating to spell a binary code Bacon had described in 1623. Every test this project has run since the second of September has come back the same way — the rules Riverbank published for telling the two founts apart do not hold on the page Riverbank chose — with a single exception. The lower-case w separates, and separates cleanly, on a feature none of their rules names: it comes in two heights.

So the w was the letter worth settling, and it could be settled exhaustively. The Bodleian’s transcription of the Folio records every letter of every page and whether it was set in roman or italic. That turns a vague question into a countable one: there are 3,080 lower-case italic w’s in the thirty-six plays, on 624 pages. Over the seventh and eighth of September the project went and found them — not a sample, not an estimate — on the Bodleian’s own 600-dpi photographs, and measured every one against both of the w stencils cut from Riverbank’s 1916 alphabet plates.

Today’s result in one sentence. 2,960 of the 3,080 were found and measured. In the text italic — the fount the plays are actually set in — they are one population, not two: 34.7 px at 600 dpi (1.47 mm), standard deviation 2.1 px, the same 34–35 px mode in every one of the thirty-six plays, and Hartigan’s dip test returns p ≈ 0.98 against bimodality. Riverbank’s two stencils fit that fount only after being shrunk to 74 per cent, and once shrunk, 2,865 of 2,877 are the a-form. There is no second w in the text italic of the First Folio, and no third shape either.
Where the two forms really live. Two w sorts do exist in the book — on four pages. The Prologues to Troilus and Cressida and Henry VIII and the Epilogues to 2 Henry IV and The Tempest are set in a larger italic, and on two of them the case mixes a short sort (43–51 px, Riverbank’s a-form at full size) with a tall one (54–59 px, the b-form). That is the same large-italic case the preliminary leaves are set in — the Digges and I.M. poems — and it is the case Riverbank’s alphabet plates were cut from. The bi-formed w is a property of one case of type, not of the book.
Histogram of the height of every italic w in the First Folio, with the two Riverbank stencil sizes marked, and sample glyphs from each fount
drag sideways on a phone · tap or click for full size
Every italic w in the thirty-six plays, measured. The blue bars are the text italic: one hump, 34–35 px, in every play. The red bars are the large italic of the four Prologue and Epilogue pages, where the two sorts sit far above it. The gold lines are the sizes of Riverbank’s two stencils, which belong to the red population and not to the blue.
Each italic w registered separately to the a-form and b-form stencils, plotted as fit against fit
drag sideways on a phone · tap or click for full size
The same glyphs, registered to each stencil independently with scale, rotation and position free. A letter that matched neither would fall near the origin; a third, unrecorded shape would show as its own cloud. Neither happens. Only 42 glyphs in the whole census fit either stencil worse than 0.55, and those are broken or ink-filled sorts.
Table of the census by play: pages, w's recorded in the Bodleian transcription, w's found, percentage, mean height, standard deviation
drag sideways on a phone · tap or click for full size
The census play by play, with the shortfalls shown rather than smoothed away. What matters is the height column: 34–35 px in the Comedies, the Histories and the Tragedies alike. Whatever else changed as the Folio was set, the w of the text italic did not.
Twenty-two hand signatures reading W.F.Friedman, cropped from the corners of the Riverbank plates
drag sideways on a phone · tap or click for full size
Read off the plates at 300 dpi, where no machine transcription sees them: twenty-two hand signatures, “W.F.Friedman” with a flourish and a small ’16, in the corners of every Alphabet and Classifier plate in The Keys and all nine plates of the I.M. Poem series in Fundamental Principles. Each one differs — signed plate by plate on the originals, then photographed. In the whole Riverbank biliteral literature, Friedman’s name appears only here: as the draughtsman.

Questions

What is the biliteral cipher?

Sir Francis Bacon described it in 1623. Every letter becomes a group of five symbols, each symbol either an a or a b — five binary digits, three hundred years before anyone called them that. AAAAA is A, AAAAB is B, and so on.

The two symbols need not be letters. They can be two founts of type so similar that a reader notices nothing. Print an innocent passage in an alternating mixture of the two, and the passage carries a second message that only someone who can tell the founts apart can read. Bacon called this writing “omnia per omnia” — anything by means of anything.

What was done on these two days, in plain terms?

Every lower-case italic w in the thirty-six plays was found on the photographs and measured. The Bodleian publishes both a machine-readable transcription of the Folio and 600-dpi images of every page, so it is possible to know in advance how many there should be (3,080), to go and locate each one, and to report honestly how many were missed.

Each glyph found was then registered twice — once against the a-form w stencil and once against the b-form — with size, angle and position all free to move. That is the test Walter asked for in July: not a single number computed from a glyph, but the two shapes laid over it at full resolution, so that a letter belonging to neither would be visible as such.

What did you find?

One w, everywhere. The text italic of the plays uses a single sort: 34.7 px tall on average, spread of about two pixels, and the same peak in all thirty-six plays. Statistically there is no sign of two overlapping populations — the dip test, which asks exactly that question, comes back as negative as it can.

The two forms Riverbank drew are real, but they belong to a different case of type: the larger italic used for four Prologue and Epilogue pages and for the preliminary poems. In that case, and only there, a short w and a tall w sit side by side on the same line.

Does this disprove Gallup’s decipherment?

No, and it is worth being exact about why. This is a census of one letter. Gallup claimed the cipher across many letters and many pages, and a negative result for w in the text italic does not by itself settle the others.

What it does is remove the one piece of positive evidence this project had generated. Since the second of September the w was the sole letter that separated into two groups under measurement. The obvious reading was that it might be the surviving trace of a two-fount setting. The census says otherwise: the separation was real, but it lives on four pages of a different fount, and those are the pages Riverbank’s own exemplars were cut from. The finding is a necessary condition failing, not a message being read.

If there is no second w, what were Riverbank’s two stencils cut from?

From the large italic. Both stencils match that case at its own size — the a-form at scale 1.0, the b-form at scale 1.0 — and they fit the text italic only after being reduced to about three-quarters. A printer’s case that has acquired two w sorts of different height is an ordinary thing; it is what happens when sorts are replenished from another founder’s stock.

That is the whole difficulty with the exemplars, and it was already recorded on 5 September under a different heading. The forms were defined on the pages where the difference exists, and then applied to pages where it does not.

Whose claim is this — Friedman’s or Fabyan’s?

Fabyan’s. William F. Friedman went on to become the most consequential cryptologist the United States has produced, and his name is often attached to the Riverbank biliteral work as though he had vouched for it. He did not. His name occurs in these four booklets in exactly one place: hand-signed, twenty-two times, in the corner of a plate, the way a draughtsman signs a drawing.

The claims are signed by George Fabyan, who owned Riverbank. Friedman drew the alphabets. He later spent much of his life arguing that the Shakespeare ciphers were not there.

Riverbank said a fourteen-year-old could learn to read the cipher. Is that not evidence?

It is evidence of something, but not of the cipher. The passage is in the Foreword to Fundamental Principles of the Baconian Ciphers (1916) — signed Fabyan, not Friedman. It says the principles are “really a simple matter”, that “young people of fourteen years and upwards… note readily differences in type”, and that one student at Riverbank marked 940 letters in three hours with only eight errors. The same page recommends the cipher for kindergarten teaching “in blocks, beads, weaving, or colors”, and for the blind, where it would require learning “two differences… in place of twenty-six”.

Two things follow. The first is that sorting two designed founts is genuinely easy, which nobody disputes — and Fabyan concedes in the very next sentence that the Elizabethan volumes are “a more difficult matter” because there “the two forms of type are not so clearly distinguishable”. The second is the important one: eight errors against what? The student was scored against Riverbank’s own answer. There is no independent key. An exercise like that measures agreement with the house reading, not recovery of a message — which is the same circularity already recorded here for the exemplars, arriving now from the pedagogy side.

The pamphlet that would carry the strong form of this claim, Ciphers for the Little Folks (Riverbank, 1916), is not in hand. Until it is, that part of the record is unexamined, and is listed on the Reference tab as a gap rather than passed over.

What would change your mind?

Stated before the census was run, and unchanged by it: a second w population in the text italic, separated on the images without reference to Gallup’s marks, holding across plays set by different compositors, and decoding to language when read five letters at a time. The census tested the first condition on the whole book and it failed. If the same census on a, e and o — which is next — found two populations where w shows one, that would be a genuine surprise and would be reported as one.

Who is doing this, and why?

Walter Wilkinson, in Worland, Wyoming, working from free public sources: the Bodleian’s images and transcription, Library of Congress scans of the Riverbank booklets, and open-source tools. Nothing here depends on access anyone could withdraw.

The wider project is New Gorhambury, and it includes Experiment One, which argues a conclusion. These reports deliberately do not. They record what was tested, what was found, and what failed, on the day it happened — including, as here, results that cut against the Baconian case.