Voynich Bench Characters: ch, sh, Ligatures and Gallows Frames

The Voynich Manuscript contains a shape that behaves less like a letter and more like architecture.

Two low curved elements face one another.

A horizontal-looking span appears between them.

The result resembles a bench.

Researchers therefore call the family the bench characters.

In EVA, the two most familiar forms are written ch and sh.

Once again, the Roman letters are modern analytical labels.

They do not tell us that the medieval writer intended a c followed by an h.

That is exactly why benches are so important.

They force us to ask what a character is.

Is Voynich ch one grapheme whose shape happens to be decomposable?

Two graphemes written together?

A ligature?

A frame with an internal slot?

An abbreviation?

A component of a larger code unit?

The question becomes more dramatic when a gallows form appears inside the bench.

EVA then produces strings such as ckh, cth, cph and cfh.

To a computer, those look like three-character sequences.

To the eye, they look like one integrated compound.

The manuscript therefore gives us a visual grammar before it gives us a language.

The bench problem is not whether ch “means” something. It is whether the writing system builds complex units by inserting components into reusable graphic frames.

If it does, Voynichese is compositional at a level deeper than the ordinary EVA character stream.

Quick Read

One-sentence answer: the Voynich bench forms conventionally transliterated ch and sh are highly productive compound-looking structures that occupy restricted positions in tokens, participate in dense families, and can host gallows forms inside pedestalled variants; their behaviour makes them central to any model of Voynich grapheme composition, while leaving unresolved whether they are individual letters, ligatures, multi-part graphemes, abbreviations or functional frames.

“Bench” is a visual nickname, not a decoded function.
EVA ch and sh are analytical transliterations of complex-looking visible forms.
The Roman sequence ch should not automatically be interpreted as two phonetic letters.
Synthetic alphabets can represent the same visible form as one unit.
The bench family occurs extremely frequently in ordinary Voynich tokens.
Jorge Stolfi’s descriptive word grammar places ch/sh-like forms in a recurring “mantle” layer around other token components.
Stolfi also noted distributional relationships between the bench family and repeated e/ee structures, suggesting a broader structural class rather than a simple alphabetic pairing.
Bench forms can be interrupted by or integrated with gallows characters, creating pedestalled gallows.
This reuse makes “frame plus inserted component” a serious graphical model.
Simple benches and gallows-bearing benches must be explained together.
Character segmentation affects entropy because EVA’s component representation creates highly predictable internal transitions.
Historical ligatures and abbreviations show that compound written forms were normal manuscript technology, but no external sign has been proven the direct source of Voynich ch or sh.
A successful decipherment should predict when ch, sh, ee-like and pedestalled forms substitute for or exclude one another.

The bench family is therefore not a small palaeographic curiosity.

It may reveal the level at which Voynich characters are assembled.

First: ch Is Not Necessarily c + h

EVA was designed to be analytical.

When a visible sign can be described as recurring components, EVA often exposes those components in the transliteration.

That is incredibly useful for comparison.

It can also create a false ontology.

A reader sees ch in a text file and assumes:

first character c, then character h.

But a synthetic transliteration may treat the same visible shape as one compound unit.

The manuscript itself sits underneath both descriptions.

The correct question is physical and functional.

Does the writer produce two independently meaningful signs?

Or one complex sign whose strokes can be decomposed visually?

This distinction is fundamental because all later statistics depend on the answer.

Why “Bench” Is a Better Nickname Than a Translation

The bench nickname describes visual geometry.

That is safer than calling the form a syllable, ligature or abbreviation before the function is established.

Even so, “bench” can influence interpretation.

A bench sounds as though something sits on it.

Then pedestalled gallows seem naturally “inserted” into a frame.

That may be correct.

Or the compound may simply be one traditional sign whose parts are not independently meaningful.

The nickname should therefore remain visual:

a low compound-looking form with reusable left/right elements.

Anything beyond that needs distributional or palaeographic support.

ch and sh Form a Natural Visible Pair

Voynich ch and sh share the same broad bench architecture while differing in an internal or opening stroke feature.

This makes them an obvious contrast pair.

The key questions are:

Do they substitute in the same token positions?
Do they occur with similar endings?
Do they differ by Currier regime?
Does one favour particular document roles?
Does the graphical difference behave like one feature changing while the rest of the unit remains stable?

If yes, ch and sh may be members of one graphemic family.

That still leaves many interpretations.

Two related phonemes.

Two grammatical states.

Two code values.

Two graphical variants.

The visible pair gives us a family.

It does not give us the feature name.

Stolfi Put Benches in the Mantle

Jorge Stolfi’s descriptive grammar is useful because it asks how Voynich tokens are assembled without requiring translation first.

In his core–mantle–crust model, ch and sh are prominent members of a “mantle” layer that can occur around a gallows-containing core or form tokens even when no gallows core is present.

This is not an accepted linguistic grammar.

It is a descriptive compression of visible word structure.

Its value is that it reveals non-random architecture.

Bench forms do not behave like freely scattered characters.

They occupy recurring structural positions.

A successful decipherment should eventually tell us what underlying linguistic or encoding relationship produced the mantle-like distribution.

The ee Relationship Is a Clue That the Bench May Be a Structural Class

One of Stolfi’s interesting observations is that repeated e-like forms can behave statistically in ways reminiscent of the ch/sh bench family.

This is odd if we assume every EVA character is an ordinary letter.

Why should a two-e sequence occupy a role similar to one bench form?

Several explanations are possible.

ee and ch/sh may encode related linguistic material.
They may occupy the same structural slot but represent different classes.
The bench may itself be composed from e/c-like minim structures.
The similarity may arise from a generation template rather than semantics.

The deeper lesson is that grapheme identity may not align with EVA string length.

Two computer characters can behave like one unit.

One complex visible unit can behave like a combination.

The bench family sits directly in that ambiguity.

The Bench Can Open to Receive a Gallows

This is the feature that makes benches structurally extraordinary.

A gallows form can appear inside the bench architecture.

EVA represents the resulting compounds analytically as forms such as:

ckh;
cth;
cph;
cfh.

The gallows article owned the tall signs.

This article owns the frame relationship.

The same bench-like shell accepts several tall members.

That substitution strongly suggests some kind of compositional architecture at the visible level.

But visible composition does not guarantee semantic composition.

A calligrapher can build several unrelated signs from one common flourish.

A ligature system can reuse frames.

A code can reuse a shell around different payload values.

The frame is real.

Its information content remains the question.

Frame + Payload Is a Serious Model

Suppose the bench is a stable frame.

Suppose the inserted gallows is a variable payload.

Then ckh, cth, cph and cfh have a relational architecture:

FRAME + VARIABLE

This kind of construction is common in information systems.

A grammatical template can hold one changing stem.

A cipher symbol can hold a state plus value.

A compound character can combine semantic or phonetic components.

A decorative writing tradition can reuse one ornamental shell.

The model becomes testable if the bench contributes the same effect across different inserted gallows.

If ckh and cth differ in exactly the way k and t differ outside the frame, compositionality strengthens.

If the framed forms have completely unrelated distributions, the whole compound may be the more appropriate unit.

The Alternative: Each Pedestalled Form Is One Grapheme

Synthetic transliteration systems make this alternative easier to imagine.

Perhaps ckh is not c+k+h at all.

It is one complex sign.

The apparent frame and inserted gallows are merely how that sign is drawn.

If so, EVA creates internal dependencies that the original script did not possess as sequential character transitions.

This could partly explain low entropy.

c strongly predicts one tall element.

The tall element strongly predicts h.

But if the entire form is one grapheme, those internal predictions are representation artefacts.

The one-grapheme model therefore makes entropy a tool for testing segmentation—but palaeography must decide whether the physical strokes support it.

The Bench Is One of the Reasons EVA Must Not Be Read Aloud

The sequence ch feels familiar in English.

It has a sound.

sh has another.

This familiarity is accidental.

If EVA had called the same visible forms X1 and X2, no one would instinctively pronounce them.

The manuscript would not have changed.

This makes bench forms an excellent demonstration of transliteration contamination.

EVA is superb for typing.

It is dangerous when the mnemonic label becomes a phonetic hypothesis without evidence.

The safest mental reading is:

bench-A and bench-B.

Meaning should arrive later.

Could ch and sh Be Related Sounds?

Natural writing systems often contain graphically related signs for related sounds.

A small added stroke can distinguish one consonant from another.

Diacritics can modify a base letter.

Therefore a phonological contrast is possible.

But a sound hypothesis must recover phonotactics.

If ch and sh represent related phonemes, their neighbouring grapheme distributions should be compatible with one language’s sound structure.

Pedestalled forms must also make phonetic sense.

And the resulting pronunciation system must explain the manuscript’s unusually strong character constraints.

Visual resemblance alone does not grant a sound.

Could Benches Be Abbreviations?

Medieval scribes routinely used compound signs and abbreviation marks.

Cappelli’s classic repertory shows just how visually varied abbreviation technology could be.

Some historical forms remind modern readers of parts of Voynich writing.

This establishes historical possibility at the level of manuscript practice.

It does not provide a direct expansion for ch or sh.

A real abbreviation model should show that bench forms expand into stable recurring sequences.

Perhaps ch represents one common syllable or ending.

Perhaps sh represents a related variant.

The same expansion should work across unrelated visual sections and unseen pages.

Without that, “abbreviation-like” remains a capability comparison.

Could Benches Be Code Frames?

A code or cipher interpretation is especially natural for the frame-plus-payload architecture.

Imagine the bench identifies one category.

The inserted gallows chooses one value inside that category.

Or the bench marks a verbose group while the internal sign identifies one plaintext element.

This could generate a limited family of legal compounds and strong conditional entropy.

The hypothesis becomes serious only when one encoding rule generates:

simple ch/sh forms;
pedestalled gallows;
their token positions;
their Currier differences;
their relationships to word families.

A cipher should explain the whole bench ecosystem.

Not merely give one compound an attractive plaintext value.

Could Benches Be Generated Templates?

A constrained token generator can create benches very easily.

Define a mantle slot.

Choose ch, sh, ee or nothing.

Optionally insert a gallows.

Add a legal ending.

The resulting corpus naturally has:

slot-like word structure;
dense families;
highly predictable character sequences;
rare illegal combinations.

This explains surface form elegantly.

As always, the burden moves upward.

Why does the generator change by Currier regime?

Why do labels have another register?

Why do token distributions align with visual sections?

Generation can explain architecture without yet explaining information.

Benches Are Not Uniform Across Currier Regimes

The relative use of ch/sh-containing families changes across A/B and newer distributional classifications.

This matters because a bench model cannot be purely static.

If ch and sh are phonetic, perhaps the underlying vocabulary or dialect changes.

If they are grammatical, perhaps grammatical distributions change by register.

If they are code components, perhaps the encoding table or state changes.

If they are scribal, perhaps writers favour different graphic variants.

The bench family therefore gives Currier theories another independent burden.

An A↔B transformation should predict bench-family shifts rather than merely relabel them.

Benches Interact With q-Series Word Families

Voynich word structure is layered.

q-like opening material, o/a-like round elements, gallows cores, bench families and endings can combine under strong restrictions.

This means the bench cannot be deciphered in isolation.

If a q-bearing token and its non-q relative both contain the same bench, a grammatical-prefix theory predicts the bench’s underlying role should remain stable.

If changing ch to sh creates another near-neighbour, the difference should have a stable function if the system is morphological.

The whole token architecture creates a lattice of substitutions.

A real decipherment should make that lattice meaningful rather than assigning each full token independently.

Benches Matter for the Space Problem

If ch or sh is one grapheme, its internal apparent boundary should never be treated like a space-delimited boundary.

That sounds obvious.

But the larger principle matters.

Writing contains several scales of separation.

stroke separation;
grapheme separation;
compound separation;
token spaces;
line breaks.

The Voynich bench demonstrates why visual gaps and joins have to be classified by level.

A small join inside a complex sign may be meaningful palaeographically but irrelevant lexically.

A large gap may mark a token boundary but not a word boundary.

Segmentation is hierarchical.

Benches Matter for Syntax Too

If ch/sh families correspond to grammatical or lexical classes, tokens containing them may have preferred neighbours.

If they are purely graphical, neighbour preferences should be inherited from the underlying units rather than the bench shape itself.

If they are generated templates, adjacency may be weak even while internal token structure is strong.

This creates a useful bridge to the fourth article in this batch.

Voynich has very strong within-token constraints but surprisingly weak whole-token word-order repetition.

Perhaps the relevant “syntax” operates partly between sub-token classes rather than full word identities.

Bench-containing classes are one candidate level at which that hidden order can be tested.

The Bench Frame Makes Negative Evidence Powerful

If the frame can host k, t, p and f, why are some theoretically possible combinations rare or absent?

This is where a compositional grammar earns its value.

A free frame-plus-payload system predicts every payload should be equally possible unless another rule intervenes.

If actual frequencies differ radically, the system has selection constraints.

Those constraints can reveal hidden categories.

Phonotactics.

Morphology.

Code legality.

Scribal convention.

The rare combinations may therefore be more informative than the common ones.

They tell us where visible compositionality stops being free.

One External Lookalike Cannot Solve the Bench

The bench forms have often been compared with medieval Latin abbreviation signs, alchemical marks, ligatures and early cipher characters.

This comparison space is useful.

It tells us what scribes of the period could do graphically.

Complex compound signs were historically normal.

Abbreviation technology was rich.

Secret-writing alphabets could use unfamiliar signs.

But if one external sign looks like ch, that does not make its known Latin expansion the Voynich value.

Visual transmission must be demonstrated historically and functionally.

Resemblance opens a candidate file.

It does not close the case.

What Survives the Bench Work

ch and sh are highly productive recurring visible structures.
They form a natural graphical family.
They occupy restricted positions inside tokens.
They participate in word-family structure rather than appearing freely.
They can form compound-looking units with gallows characters.
The same bench architecture can host several gallows types.
Analytical and synthetic transcriptions disagree about the appropriate unit level.
That disagreement materially affects character counts and entropy.
Stolfi’s descriptive grammar shows bench forms participating in a repeatable mantle-like structural layer.
ee-like structures can show distributional relationships with bench forms, suggesting the deeper class may not align simply with Roman-letter count.
A successful model must explain ordinary benches and pedestalled forms as parts of one writing architecture.

What Does Not Survive as Established Knowledge

EVA ch is proven to be two medieval letters.
EVA sh is proven to be two medieval letters.
ch and sh have known phonetic values.
They are proven Latin abbreviations.
They are proven cipher groups.
The bench is proven to be a grammatical frame.
The gallows inside a bench is proven to be a payload variable.
Pedestalled forms are proven single graphemes.
Pedestalled forms are proven multi-grapheme sequences.
One external manuscript lookalike identifies their function.

The graphic composition survives.

The semantic composition remains unknown.

A Better Bench Analysis

Treat ch/sh as visible forms before phonetic strings.
Compare analytical and synthetic transcriptions.
Map the stroke construction palaeographically.
Compare ch and sh contexts directly.
Compare benches with ee-like structural alternatives.
Analyse simple benches and pedestalled benches together.
Test whether the bench contributes the same effect across inserted gallows.
Condition distributions on Currier regime and hand.
Measure how segmentation changes entropy and word families.
Require any semantic or encoding value to predict unseen compounds.

What Would Count as a Real Bench Breakthrough?

Imagine a palaeographic study shows that ch and sh are consistently written as single continuous ligatures, while the same bench skeleton opens predictably when a gallows is inserted.

A decipherment then demonstrates that the bench contributes one stable grammatical or phonological value across all compounds.

The inserted gallows contributes another stable value.

Unseen combinations decode correctly.

That would validate deep compositionality.

Or imagine a synthetic model treating each bench and pedestalled form as one grapheme dramatically simplifies the sequence, aligns conditional entropy with a known historical writing system, and improves held-out translation.

That would support whole-unit graphemes.

Either result would be a major advance because it answers a question deeper than vocabulary:

what are the actual building blocks of Voynich writing?

Primary School: The Frame and the Shape Inside It

Draw a box with a star inside.

Then draw the same box with a triangle inside.

Ask a child what stays the same and what changes.

The frame stays.

The centre changes.

Now ask whether the frame and centre might each carry information.

The child has discovered the bench-plus-gallows question without needing any cryptography.

Lower Secondary: One Sign or a Formula?

Give students four symbols that share one frame and have four different centres.

Ask them to build two models.

Each complete symbol is unrelated.
Frame and centre each contribute information.

What observation would distinguish them?

If the same centre has the same effect outside and inside the frame, compositionality strengthens.

This is the core logic of pedestalled-gallows analysis.

Upper Secondary: Change the Alphabet, Change the Statistics

Represent the same four compounds in two ways.

Model A gives each compound one symbol.

Model B writes frame-left + centre + frame-right.

Calculate alphabet size and adjacent-pair frequencies.

The numbers change even though the source drawings do not.

The exercise shows why entropy cannot be interpreted independently of grapheme ontology.

JC and Adult Readers: Compositionality Must Be Identifiable

At a higher level, the bench problem is one of factorisation.

Does the probability or function of a compound factor into contributions from frame and inserted element?

If yes, a compositional model can explain several surface forms with fewer parameters.

If no, treating each compound as an atomic grapheme may be more economical.

The important comparison is predictive.

Which model generalises better to rare or withheld forms?

That is how “looks compositional” becomes evidence of actual compositionality.

A Parent and Teacher Guide

Do not pronounce EVA ch or sh.
Separate visible components from assumed letters.
Notice the reusable bench frame.
Compare simple and gallows-bearing forms.
Ask whether frame and centre contribute independent effects.
Show how one-versus-many character models change statistics.
Require the final interpretation to predict rare combinations.

The transferable lesson is fundamental:

an object can look made of parts without those parts necessarily being independent units.

Reader Checklist: Before You Explain ch or sh

Are you reading EVA ch/sh as familiar phonetic digraphs?
Which visible form is the transcription representing?
Is the alphabet analytical or synthetic?
Does palaeography support one unit or multiple components?
Do ch and sh substitute in comparable contexts?
How do they relate to ee-like structures?
How does the bench change when a gallows is inserted?
Does the inserted gallows retain its ordinary contextual effect?
Does Currier regime change bench distribution?
Does proposed hand change it?
How does segmentation alter entropy?
Does the model predict unseen or rare compounds?

Frequently Asked Questions

What are Voynich bench characters?

They are common compound-looking forms conventionally transliterated in EVA as ch and sh because their visible geometry resembles a low bench-like structure.

Are ch and sh two letters each?

Not established. EVA is analytical and exposes visible components. Synthetic transcription systems can treat the forms as single units.

Do they sound like English ch and sh?

Unknown. The Roman letters are transcription labels, not decoded sounds.

How are benches related to gallows?

The bench architecture can combine with gallows forms, producing pedestalled compounds such as those transliterated ckh, cth, cph and cfh.

What did Stolfi mean by the mantle?

In his descriptive core–mantle–crust grammar, ch/sh and related structures occupy a recurring layer around token cores. It is a structural model, not a deciphered grammar.

Could benches be abbreviations?

Possibly. Medieval manuscripts used complex abbreviation signs and ligatures. No accepted expansion has been demonstrated for Voynich benches.

Why do benches matter for entropy?

Because an analytical transcription can turn one compound visible unit into several highly predictable sequential characters. Treating the same form synthetically changes the entropy calculation.

What is the strongest current conclusion?

The bench family is a highly structured, reusable graphical architecture that participates in token composition and gallows compounds. Its true graphemic and semantic decomposition remains unresolved.

Research and Further Reading

The Final Idea

The bench characters are important because they look like the manuscript showing us how it builds things.

A stable frame.

A related alternate frame.

A gallows inserted inside.

Recurring positions around token cores.

That architecture is hard to dismiss as random penmanship.

But architecture is not yet semantics.

A cathedral arch has parts before we know the prayers spoken beneath it.

The bench may likewise tell us how Voynich writing is constructed before it tells us what that construction says.

If the final decipherment is correct, it should make the bench look inevitable rather than merely familiar.

Voynich | Everything eduKate Knows and Tested | The Bench Characters: ch, sh and the Frame Around the Gallows