Voynich Null Problem: Do Some Signs Mean Nothing?

Some Voynich signs may mean nothing.

That sentence sounds reckless until we remember that historical cryptographers sometimes designed symbols to do exactly that.

A null is a ciphertext element deliberately inserted without plaintext value.

It can distract a codebreaker.

Flatten frequencies.

Separate groups.

Fill awkward space.

Or simply create uncertainty about which visible marks matter.

Historical cipher instructions from early modern Europe explicitly discuss null characters, and surviving keys show that nulls were part of real cryptographic practice from the fifteenth century onward.

So the Voynich null hypothesis is not fantasy.

It is also one of the easiest hypotheses to abuse.

A proposed reading almost works.

One glyph is inconvenient.

Call it null.

The next word contains two extra signs.

Call those null too.

Soon the decoder can throw away whatever the theory cannot explain.

That is not cryptanalysis.

It is selective deletion.

A real null hypothesis must tell us which signs carry no plaintext information before we know which deletions make the proposed translation look better.

That is the Null Problem.

Quick Read

Direct answer: nulls are historically plausible in a fifteenth-century cipher, but no Voynich glyph, token family or positional class has been securely demonstrated to be semantically null. Any proposed null system must be defined independently, improve a coherent underlying model across unseen text, preserve fixed values elsewhere, and explain why the supposedly meaningless forms have the manuscript-wide positional, scribal and document-role distributions that they do.

Historical nomenclators and diplomatic ciphers often contained nulls.
Nulls are deliberately valueless in plaintext terms, but they can have procedural functions in ciphertext.
Historical nulls could be single signs, numbers, groups or even meaningful-looking words.
A 2022 HistoCrypt corpus study reported nulls as extremely common in surviving fifteenth-century cipher keys.
Later cipher instructions explicitly advised where nulls should or should not be placed.
Therefore “a fifteenth-century cipher could use nulls” is historically sound.
“This Voynich sign is null” remains a separate claim requiring evidence.
Gallows characters are tempting null candidates because of their positional behaviour, but paragraph and line architecture gives them strong structural roles that a null model must explain.
Rare glyphs are tempting nulls because they contribute little to frequency totals, but rarity alone does not imply meaninglessness.
Repeated filler-like endings may reflect line fitting, abbreviation, morphology or encoding rather than nullity.
A true null can be meaningless in plaintext while still being placed under a rule.
Removing candidate nulls should reveal stronger reusable structure rather than merely create one attractive translation.
The same null rule must work across Currier regimes, scribes, labels and running text where the mechanism is claimed to apply.
A null hypothesis should be penalised for every degree of freedom it adds.
No accepted Voynich solution currently provides a fixed, predictive null inventory.

What “Null” Means in Historical Cryptography

The word sounds like “mistake” or “noise”.

Historically, a null could be intentional.

The sender and intended receiver know that a particular symbol contributes no plaintext content.

The interceptor does not.

That asymmetry is useful.

Frequency analysis depends on assuming that visible ciphertext signs correspond somehow to source-language units.

Insert meaningless material and the codebreaker now has to solve two problems:

What do the meaningful signs represent?
Which signs should be ignored entirely?

That is why nulls appear in historical key instructions.

They complicate inference cheaply.

Nulls Were Real Enough to Receive Rules

Historical cipher manuals and key instructions did not merely list nulls.

They sometimes told encipherers how to use them.

One modern study of early modern cipher-key instructions traces guidance from the fifteenth century onward concerning null characters, homophones and nomenclator entries.

Later instructions can be surprisingly operational.

Do not put nulls only at obvious boundaries.

Distribute them.

Use enough to trouble the decipherer.

This tells us something important for Voynich.

A null can be semantically empty while still having a non-random production distribution.

So “this sign occurs systematically” does not automatically disqualify nullity.

But systematic placement creates a mechanism that must itself be explained.

Null Is Not the Same as Filler

Suppose a scribe reaches the right margin with too little content to fill the line.

They add a conventional flourish or extend a word form.

That may have no lexical meaning.

But its purpose is typographic rather than cryptographic.

This is a filler or line-fitting behaviour.

A cryptographic null is inserted to alter the ciphertext relation to plaintext.

The two can look identical on the page.

Their predictions differ.

Line filler should correlate strongly with available physical space.
Cryptographic nulls may be distributed by key rule independent of margin geometry.
A filler may occur mostly near line ends.
A null designed for cryptographic confusion may deliberately avoid such predictability.

Voynich line effects therefore matter directly to the null question.

Null Is Not the Same as Decoration

An enlarged paragraph-initial form may be decorative.

A flourish can identify a section boundary.

A rare stroke can be palaeographic emphasis.

If such material carries no sound value, it is tempting to call it null.

But “no phonetic value” is not the same as “no information”.

A paragraph marker can carry document information.

A capital letter can carry boundary information beyond its lexical identity.

A rubric can organise retrieval.

Therefore a sign can be non-lexical without being null in the broader information system.

Before calling a Voynich sign meaningless, ask whether it is carrying structure instead of plaintext.

The Gallows Are an Obvious Temptation

Gallows characters are visually dramatic.

They are also highly positional.

Some occur preferentially at line and paragraph beginnings.

Pedestalled variants complicate the character inventory further.

A cipher theorist may ask whether gallows are null padding designed to disrupt frequency analysis.

It is a legitimate hypothesis.

It also inherits difficult evidence.

If gallows are purely cryptographic nulls, why are they so tightly associated with document boundaries?

Why do paragraph-initial words “grow” gallows-like features?

Why do different gallows forms have different distributions?

A historical null can be positional, but the position rule now needs an operational reason.

Removing the gallows should also improve a stable underlying decoding across many contexts.

No accepted result has demonstrated that.

A Paragraph Marker Can Be Null to Plaintext and Still Not Be “Nothing”

Suppose a gallows-like form means:

new entry begins here.

It may contribute nothing to the sentence translated into English.

Yet it carries document function.

This distinction is particularly important because Voynich has weakly understood paratext.

The manuscript may integrate heading, entry and control information into ordinary-looking glyphs.

A decoder that strips such signs as meaningless could accidentally erase the document grammar needed to understand the book.

Rare Glyphs Are Another Temptation

A one-off sign contributes almost nothing to stable frequency analysis.

Perhaps it is meaningless.

Perhaps.

But rarity has many causes.

allograph;
ligature;
scribal error;
damaged common glyph;
special abbreviation;
foreign proper name;
numeral or notation;
rare but meaningful code;
null.

A null theory does not win simply because a sign is inconveniently rare.

It must explain the rarity under a larger rule.

Whole Tokens Could Be Null Too

Historical nullity was not restricted to single characters.

Some cipher traditions used larger meaningless units.

Early modern instructions even contain examples of ordinary meaningful-looking words designated as superfluous or null in coded correspondence.

This opens a more radical Voynich possibility.

Perhaps one visible token is not a word or code at all.

It is padding.

Again, flexibility becomes dangerous.

A solver must identify the null-token class from an independent rule.

Frequency.

Position.

Key family.

Or another reproducible mechanism.

Declaring every untranslated token “padding” is not a code.

Nulls Can Explain High Surface Diversity

Add optional null material to a base ciphertext unit and one underlying plaintext form acquires many visible variants.

Base form:

ABCDE

With optional nulls:

XABCDE;
ABYCDE;
ABCDEX;
XABYCDE.

The visible vocabulary expands.

Word families become dense.

Exact repetition decreases.

This makes null insertion relevant to several Voynich anomalies.

But optional nulls also make a model dangerously powerful.

If any sign can be ignored anywhere, any surface string can conceal many plaintexts.

A serious null system must sharply limit which additions are legal.

Nulls Could Affect Word-Length Statistics

Optional padding changes visible token length.

If null insertion itself follows bounded positional rules, it can help generate a narrow or near-binomial word-length distribution.

This makes the Null Problem directly connected to the Word-Length Problem.

A good model should not merely say “nulls explain the lengths”.

It should derive the observed length distribution from the legal null positions and insertion probabilities.

Then test the result on held-out text.

Nulls Could Affect Entropy

Meaningless material sounds as though it should make text more random.

Not necessarily.

If null placement follows a rule, the inserted material can be highly predictable.

A null prefix always appears in one slot.

A terminal filler uses one of three signs.

Surface entropy can therefore rise at one level and fall at another.

The Entropy article matters because a null mechanism must reproduce the manuscript’s unusually constrained character transitions rather than simply inject arbitrary noise.

Nulls Must Respect Currier A and B

If one null inventory is manuscript-wide, its signs should remain null across Currier regimes unless the key itself changes.

If the null inventory changes between A and B, the theory now requires a state transition.

That may be legitimate.

Historical ciphers could change keys.

But the change should explain observed distributional discontinuities rather than merely follow them after the fact.

Any null system therefore inherits the Drift and Currier problems.

Nulls Must Respect Scribal Hands

One scribe may prefer one null form.

Another may rotate among several.

That is historically plausible if scribes share a key but differ in habits.

However, if every proposed hand receives its own ad hoc list of nulls, the model becomes hard to falsify.

The correct workflow is to derive the null rule independently and then ask whether hand-specific preferences explain residual variation.

Labels Are a Hard Null Test

Labels are short.

Many are unique.

They avoid some prose-heavy constructions, especially q-like openings.

If labels are names or identifiers, cryptographic nulls may be especially costly because each short string contains little redundant material.

A manuscript-wide null mechanism should predict whether labels use the same padding system.

If they do not, why not?

Document role can be the answer.

But the rule should be explicit before values are assigned.

The Edge Problem Creates a New Test

Recent 2026 work reports stronger coupling between token-edge glyphs than between exact token identities.

Nulls could interact with this in several ways.

If edge glyphs are null padding, the edge coupling may reflect padding rules rather than syntax.
If nulls are internal, edge coupling should survive their removal.
If uncertain spaces split larger units, some supposed nulls may actually be components displaced by transcription.

A null theory should therefore be tested against boundary strength rather than only conventional word tokens.

A Null Theory Must Improve Something Specific

Removing signs always reduces data.

Reduced data can look simpler automatically.

That is not enough.

A valid null hypothesis should improve an independently meaningful target.

source-language phonotactics;
stable word families;
cross-word grammar;
known crib alignment;
historically plausible ciphertext reconstruction;
prediction of unseen text.

“The entropy got lower after I deleted 20% of the symbols” proves little by itself.

Deletion tends to simplify.

The question is whether it reveals the correct hidden system.

The Null Inventory Must Be Frozen

Suppose a candidate set of null signs is discovered on Currier B herbal pages.

Freeze it.

Do not add another sign when a pharmaceutical page becomes inconvenient.

Do not restore a “null” when a zodiac label needs its value.

Apply the rule unchanged to unseen bifolia.

The more the null inventory stabilises, the more scientific weight it can carry.

The more it grows to rescue each example, the less it explains.

Nulls Must Have a Production Cost

Every meaningless sign still has to be written.

Ink.

Time.

Memory.

Key complexity.

The Voynich Manuscript is long.

A theory claiming that a large fraction of its writing is padding should explain why the historical producer accepted that cost.

For diplomatic secrecy, the benefit is obvious.

For a private technical manual, perhaps less so.

Purpose and production economics become part of the mechanism test.

Nulls and the Almost-Correctionless Manuscript

Null insertion complicates writing.

If the scribe must remember both plaintext encoding and padding rules, we might expect more execution errors.

Voynich shows remarkably few conspicuous corrections.

This does not rule out nulls.

Historical cipher clerks could be trained.

But a proposed null system should be hand-executable enough to match the observed fluency.

A theoretically elegant null schedule requiring constant calculation would be a poor production fit.

Historical Controls Matter More Than Imagined Ciphers

The strongest way to test nulls is constructive.

Take historically documented cipher keys containing nulls.

Encrypt period-appropriate plaintext.

Measure the output.

How do word lengths change?

Character entropy?

Exact repetition?

Edge coupling?

Local word families?

This is better than inventing a custom Voynich null cipher whose parameters were selected because they reproduce Voynich.

The Control Problem remains the parent discipline.

Nulls Do Not Rescue a Translation From Every Failure

This rule should be ruthless.

If a translation requires a sign to be null because that sign prevents one word from matching, then the same sign must be null everywhere the rule says it is null.

If the translation later needs that sign to represent a vowel, the theory has contradicted itself.

Contextual nullity is possible in sophisticated systems.

But context must be defined mechanically.

“It is null when I need it to be” is not context.

What Survives the Null Work

Nulls are historically plausible in fifteenth-century and later cipher practice.
Historical nulls could be single signs or larger units.
Nulls can be deliberately placed under rules while carrying no plaintext value.
Voynich has several classes—gallows, rare signs, positional additions and filler-like forms—that can reasonably be tested for nullity.
No such class has been accepted as a fixed null inventory.
Paragraph and line positioning makes some candidates structurally informative even if they eventually prove non-lexical.
A null system could contribute to surface diversity, word-length variation and reduced exact repetition.
A null hypothesis must be evaluated with realistic historical controls.
Candidate nulls must be frozen before unseen testing.
Removing nulls must reveal stable transferable structure rather than one locally convenient reading.

What Does Not Survive as Established Knowledge

Gallows characters are proven nulls.
Rare glyphs are meaningless filler.
Line-final additions are proven cryptographic padding.
Any untranslated character may be discarded.
Voynich uses the same null practice as a particular surviving diplomatic key.
A lower entropy after deletion proves correct null identification.
A null hypothesis by itself proves the manuscript is ciphertext.
Null insertion explains the whole manuscript.

Historical plausibility survives.

The Voynich null inventory does not.

A Better Null Analysis

Define a candidate null class without using the desired plaintext.
State whether nullity is character-, position-, token- or context-dependent.
Separate cryptographic nulls from layout filler and paratext.
Measure the candidate distribution across Currier, hand, line position and document role.
Freeze the rule.
Remove or neutralise candidate nulls on unseen text.
Measure improvement in independently meaningful targets.
Compare against matched historical null-containing ciphers.
Charge the theory for every additional context rule.
Preserve failures where the null rule makes the text worse.

What Would Count as a Real Null Breakthrough?

Imagine a palaeographic and statistical study identifies one visible sign class that behaves independently of lexical families but under a compact positional rule.

The candidate class is frozen.

When removed from held-out bifolia, several independent things improve at once.

word families collapse into stable underlying forms;
source-side phonotactics become more coherent;
exact phrase recurrence increases to plausible levels;
Currier differences simplify under one key-state model;
a known or independently discovered crib aligns without further deletions.

Then a historically plausible cipher using the same null rule reproduces the surface statistics from meaningful period text.

That would be a genuine null breakthrough.

A null becomes evidence when deleting it makes many things predictable that were not used to decide the sign was null.

Primary School: The Extra Symbols

Write a simple secret message where every third symbol is meaningless.

Tell one child the rule.

Do not tell another.

The same meaningless symbols that are easy for the intended reader to ignore become a real obstacle for the outsider.

Then let the outsider choose any symbols they want to delete.

They can make many fake messages.

The exercise teaches the difference between a fixed null rule and convenient deletion.

Secondary School: Null, Filler or Marker?

Give students a synthetic document with three non-lexical elements:

random cryptographic nulls;
line-end fillers;
paragraph markers.

All three carry no ordinary word meaning.

But their distributions differ.

Ask students to identify the mechanism from placement before reading any plaintext.

JC and Adult Readers: Nulls as Latent Deletion Variables

At a higher level, null detection is a latent-variable problem.

Each observed sign has a hidden state:

information-bearing;
null;
structural/non-lexical.

The danger is identifiability.

A model allowed to label arbitrary signs null can always increase apparent regularity.

Therefore null complexity must be penalised and validated out of sample.

The strongest model is the smallest deletion rule that improves several independent properties simultaneously.

A Parent and Teacher Guide

Teach that historical nulls are real.
Separate “no plaintext value” from “no structural information”.
Distinguish nulls, fillers, decoration and paragraph markers.
Never identify nulls by deleting whatever blocks a translation.
Freeze candidate null rules before validation.
Use historical cipher controls.
Demand improvement on unseen text.
Keep failed deletions visible.

The transferable reasoning lesson is:

removing information can always make a pattern look cleaner; the scientific question is whether the deletion rule was justified before the cleanliness appeared.

Reader Checklist: Before You Call a Voynich Sign Null

What independent property identifies the candidate?
Is the candidate a sign, token, position or context?
Could it instead be filler?
Could it be paratext?
Could it be an allograph or compound component?
How does it distribute by line position?
By paragraph position?
By Currier regime?
By proposed hand?
By document role?
What historical cipher control uses comparable nulls?
Does removal improve source-side structure?
Does the improvement survive held-out bifolia?
Does the theory ever restore a supposedly null sign when convenient?
How much model freedom does the null rule add?

Frequently Asked Questions

Did medieval ciphers use meaningless symbols?

Yes. Historical cipher keys and instructions document nulls—symbols or other units deliberately given no plaintext value—to complicate cryptanalysis.

Are there proven nulls in Voynich?

No accepted Voynich null inventory exists.

Could gallows be nulls?

They can be tested as candidates, but their strong paragraph and line-position behaviour means a null interpretation must explain why meaningless padding is tied so closely to document structure.

Can a whole word be null?

Historically, larger null units and meaningless-looking additions existed. A Voynich token-level null hypothesis remains possible but requires a fixed selection rule.

Would deleting nulls decipher Voynich?

Only if the null rule is independently justified and the remaining material develops stable, predictive language or code structure. Deletion alone is not decipherment.

Does the existence of historical nulls prove Voynich is a cipher?

No. It establishes historical plausibility for one mechanism class, not evidence that the mechanism is actually present in Voynich.

Research and Further Reading

The Final Idea

The null hypothesis deserves respect because history makes it real.

Fifteenth-century cryptography could contain deliberately meaningless material.

But history gives us something else too.

Rules.

Nulls belonged to keys.

They were selected.

Placed.

Used by convention.

They were not permission for the intended reader to delete whichever symbols made the message difficult.

Voynich deserves the same standard.

If some Voynich signs really mean nothing, the breakthrough will come when a fixed rule tells us exactly which ones—and the rest of the manuscript becomes more predictive before we are allowed to change that rule again.

Voynich | Everything eduKate Knows and Tested | The Null Problem: Are Some Voynich Signs Supposed to Mean Nothing?