Voynich Reading Direction: Left-to-Right, Circular and Radial Text

Which way does Voynich read?

The shortest answer is reassuring.

Most ordinary running text is written from left to right.

Lines proceed from top to bottom.

The left margin is usually the stable one.

That looks familiar.

Then the manuscript starts drawing circles.

Text bends around circumferences.

Words point along radii.

Labels rotate to follow diagrams.

A foldout makes the page itself larger than one normal orientation.

The Rosettes sheet contains writing that gradually changes orientation around a full circle: left-to-right on one side of the diagram, top-to-bottom above it, right-to-left on the opposite side, and bottom-to-top below.

Now “reading direction” stops being one question.

It becomes several.

Which way was the pen moving?
Which way should the reader traverse the sequence?
Where does a circle begin?
Does a radial string read inward or outward?
Does page rotation change the intended sequence?
Is a modern transcription preserving geometry or flattening it?

Those questions matter because sequence statistics are directional.

A bigram AB is not BA.

A prefix becomes a suffix if a string is reversed.

The first token on a ring changes if the ring is rotated.

A radial label can attach to centre or perimeter depending on which way it is read.

Before deciphering sequence, we have to know which sequence we actually have.

Direction is not a cosmetic property of Voynich transcription. In circular and radial text, direction is part of the data.

Quick Read

One-sentence answer: ordinary Voynich paragraph text is overwhelmingly written left-to-right and top-to-bottom, but diagrammatic writing uses multiple geometric orientations; the current IVTFF reference inventory distinguishes 84 circular loci—83 clockwise and one anti-clockwise—and 142 radial loci—75 inward and 67 outward—so any serious sequence analysis must preserve direction, start point, wrap-around adjacency and page orientation rather than flatten every locus into one assumed left-to-right line.

Standard running text is visibly left-to-right and top-to-bottom.
This observation concerns the direction of principal prose writing, not the identity of the language.
Left-to-right writing is compatible with many European, Mediterranean and adapted scripts and therefore does not establish provenance.
Voynich diagrams contain circular writing along circumferences.
Current IVTFF reference data classify 84 circular writing loci.
83 are recorded as clockwise.
One is recorded as anti-clockwise.
The single anti-clockwise circular locus occurs in the lower-left small circle on the Rosettes sheet.
Current reference data also classify 142 radial text loci.
75 run from outside toward the centre.
67 run from the centre outward.
Radial direction therefore cannot be assumed from one global convention.
A circular sequence has no inherent linear “first” token unless a start point is physically indicated.
Some Rosettes circles have visible starting points; one has two.
Modern linear transcription necessarily cuts circular text somewhere and can create an artificial boundary.
Wrap-around neighbours must remain adjacent in analysis.
Changing start point rotates a circular sequence without changing the manuscript.
Reversing direction changes token adjacency and can create false matches or destroy real ones.
Reading direction should be distinguished from glyph orientation and from the writer’s body/page position.
The geometry should be recovered before semantic order is claimed.

The correct global answer is therefore:

Voynich prose has a dominant direction; Voynich information geometry has several.

Ordinary Prose Gives Us the Strongest Directional Anchor

The main text is the least ambiguous place to begin.

Lines begin on the left.

Characters are formed consistently within that orientation.

Paragraphs proceed downward.

Right margins are rougher than left margins.

Illustration obstacles alter the available rightward extent.

All of this supports left-to-right production very strongly.

This is an observation of writing behaviour.

It is not a translation.

It does not tell us whether the script encodes Latin, a Romance language, another language, a cipher or structured notation.

Many systems can be written left-to-right.

Direction narrows the physical script behaviour more than it narrows semantic identity.

Writing Direction Is Not the Same as Language Direction

A cipher can reverse plaintext.

A transliteration can reorder units.

A script adapted from another tradition can retain or change writing direction.

Therefore seeing left-to-right Voynich writing does not prove the underlying language, if any, is normally written left-to-right.

For example, an encoder could take source material from a right-to-left language and output a left-to-right cipher.

A shorthand system could be designed independently of its source script.

The visible direction belongs first to the Voynich writing system.

Any inference to source language requires another step.

script direction constrains mechanism; it does not identify language by itself.

The Manuscript Uses Circular Writing Extensively

Astronomical, zodiac and cosmological-looking pages frequently integrate text into circles.

This is not one exceptional decorative flourish.

It is a recurring document technology.

The modern IVTFF reference system therefore treats circular writing as its own locus type rather than forcing it into ordinary paragraph text.

That distinction is intellectually important.

Circular text has:

angular position;
clockwise or anti-clockwise direction;
possible explicit start points;
wrap-around adjacency;
relationship to a centre or sector.

Ordinary prose has none of these in the same way.

Flattening both into identical strings throws away source structure.

Current Reference Data Record 83 Clockwise Circular Loci

The IVTFF inventory currently records 84 circular writing loci.

Of those, 83 are classified as clockwise.

This is a strong manuscript-level asymmetry.

It suggests clockwise circular writing is the normal diagrammatic convention used by the manuscript.

But the meaning of “clockwise” needs care.

The characters are oriented along the ring so that the reader follows the text around the circumference in that direction.

This does not tell us that the underlying concept itself proceeds clockwise.

A circular table can be written clockwise while representing an unordered set.

A calendar can follow a cycle.

A label ring can merely place names around objects.

Writing direction is not automatically semantic direction.

The Single Anti-Clockwise Case Is More Valuable Than Its Rarity Suggests

One circular locus in the current inventory runs anti-clockwise.

It occurs in the small lower-left circle on the Rosettes sheet.

This exception matters because it disproves a universal mechanical rule:

Voynich circular writing is not physically incapable of running anti-clockwise.

The maker could reverse direction.

Therefore clockwise dominance is a convention or functional preference rather than a motor impossibility.

That creates a useful question.

Why is this locus different?

Does it have a different document role?
Does its vocabulary differ?
Is direction caused by available geometry?
Does it mirror another nearby element?
Is it a scribal/layout exception rather than semantic signal?

The rarity makes interpretation difficult.

It also makes the case a valuable stress test for any theory claiming that direction carries meaning.

Radial Writing Splits Almost Evenly Between Inward and Outward

Radial text creates a different directional system.

The current IVTFF inventory records 142 radial loci.

75 are classified as inward: outside toward the centre.

67 are outward: centre toward the perimeter.

Unlike the overwhelming clockwise bias in circular text, radial direction is almost balanced.

This tells us immediately that no universal “all radial text reads outwards” rule is safe.

Direction is selected locally.

The local choice may reflect:

which object the text labels;
available writing space;
reader ergonomics;
diagram topology;
semantic orientation;
scribal habit.

A model should test these alternatives rather than assigning inward/outward semantic values by intuition.

Inward Writing Does Not Automatically Mean Input

Direction metaphors are seductive.

Inward = input.

Outward = output.

Clockwise = forward.

Anti-clockwise = reverse.

These analogies are useful for generating hypotheses.

They are not decoded functions.

A radial label may be written inward simply because its outer endpoint gives the writer room to begin.

Another may be outward because the centre is less crowded.

To become semantic, direction should correlate with independent variables.

Do inward strings use one textual class?

Do outward strings use another?

Does the difference recur across unrelated diagrams?

If not, geometry remains the simpler explanation.

A Circle Has Direction but No Natural Beginning

This is the central mathematical problem.

A line has two ends.

A circle has none.

If text runs clockwise around a ring, we know neighbour order once we choose one point.

We still do not know which token is first.

Some Voynich circles visibly indicate starting points.

The Rosettes description notes that starting points are usually indicated for its large circles, and one circle appears to have two.

Elsewhere the start can be more ambiguous.

Modern transcription must choose one.

That choice needs metadata.

Otherwise the first token in the file acquires a privilege the parchment may not grant it.

Rotation Can Manufacture Matches

Suppose two circular sequences contain the same twenty tokens but are rotated by five positions.

Write them linearly from arbitrary start points.

They look different.

Rotate one sequence until the tokens align.

Now they look identical.

If the rotation was chosen after seeing the desired match, the comparison is overfit.

Circular comparisons should therefore either:

use a physically evidenced start marker;
test all rotations and correct for the search freedom;
use rotation-invariant features.

This applies directly to f57v, zodiac rings and cosmological diagrams.

A circular match is not meaningful until the rotation freedom has been accounted for.

Reversal Can Manufacture Matches Too

Direction doubles the search space.

A sequence can be compared clockwise.

Or anti-clockwise.

If both are allowed after seeing the target, chance matches become easier.

This is especially dangerous in proposed cipher wheels or zodiac mappings.

Try every start point.

Try both directions.

Try multiple segmentation rules.

One impressive alignment will eventually appear.

A serious model should pre-specify direction from the handwriting or diagram before semantic comparison.

Direction is another parameter that must not be free.

Wrap-Around Adjacency Must Survive Transcription

Linear text files have beginnings and endings.

A circular locus does not.

If a ring is transcribed:

A B C D E

the file makes E look terminal.

Physically, E may be adjacent to A.

A bigram study that ignores E→A destroys one true adjacency and creates a false boundary.

For circular sequence analysis, the data structure should therefore be cyclic.

This is a small technical choice with large consequences.

Syntax, repetition and character-transition measurements all depend on it.

The Rosettes Sheet Demonstrates Full-Body Rotation

The Rosettes sheet is a spectacular example because writing orientation changes continuously around the expanded field.

On the left side of the page, writing can run left-to-right.

Near the top, it turns toward top-to-bottom.

On the right, it becomes right-to-left.

Along the lower region, it can be bottom-to-top.

This does not mean the script suddenly changes its underlying language direction four times.

The page itself is being used as a circular writing surface.

Zandbergen notes that the orientation is compatible with a right-handed writer working around a page positioned to the writer’s right.

That is a production inference, not a decoded semantic flow.

The example is valuable because it separates glyph orientation from conceptual direction.

Page Rotation Can Be Part of the Historical Reading Experience

A modern screen automatically keeps “up” fixed.

A manuscript reader can rotate the codex or their body.

Foldouts make this even easier or more necessary.

A text region that looks upside-down in one display orientation may be comfortable after the sheet is turned.

Therefore “upside-down” should not automatically mean reverse or hidden.

It may simply reflect how a large diagram is physically handled.

Any theory assigning semantic meaning to orientation must first eliminate reader ergonomics.

The physical object is interactive.

The scan is not.

Direction Should Be Inferred From Glyph Formation When Possible

A writing direction is strongest when the characters themselves support it.

Stroke entry and exit.

Ligature direction.

Connections between characters.

Pen-angle behaviour.

Spacing after completed forms.

These palaeographic cues can distinguish the writer’s direction from a modern reader’s arbitrary traversal.

This is especially important for isolated circular sequences.

A diagram may be symmetric enough to permit both directions geometrically.

The pen strokes can still reveal how it was actually written.

Production order comes before interpretive order.

Writing Order Can Differ From Reading Order

This is a subtle but important distinction.

A scribe may write labels in whichever order is physically convenient.

A reader may later interpret them according to diagram order.

A circle may be written clockwise but consulted by jumping to one relevant sector rather than reading the full ring sequentially.

A table can be filled row-by-row but read column-by-column.

Therefore palaeographic pen direction is not always the same as semantic traversal.

A decipherment that assumes “written clockwise, therefore meant to be read as one clockwise sentence” adds an extra claim.

The claim needs evidence from textual dependence.

Do neighbouring ring tokens behave as sequence neighbours?

If not, the ring may be an index rather than prose.

A Label May Have Orientation but No Reading Sequence

Single labels complicate the idea of direction further.

A label has internal character order.

But multiple labels around a diagram need not form one continuous sentence.

Their order may be:

spatial;
categorical;
ordinal;
irrelevant.

Reading all labels clockwise simply because they lie around a circle can create a sequence the manuscript never intended.

This matters for zodiac labels and astronomical diagrams.

The geometry tells us where each label belongs.

It does not automatically tell us that Label 12 grammatically follows Label 11.

Spatial adjacency and syntactic adjacency must remain separate.

Direction Changes Character Statistics

The q-series article showed that q-like forms are overwhelmingly token-initial.

Minim strings tend toward token endings.

Gallows have positional preferences.

Reverse the text and all of these become opposite-position effects.

This gives direction a strong internal consistency test.

The chosen orientation should make the same character families behave coherently across prose, circular and radial loci where the same text system applies.

If one proposed direction makes q behave like a prefix-like gate while the reverse makes it an inexplicable suffix in every comparable context, the former gains support.

Direction is therefore not arbitrary once manuscript-wide positional grammar is considered.

Direction Changes Syntax Statistics

Syntax is directional.

Class A followed by Class B is not the same as Class B followed by Class A.

Reddy and Knight’s boundary effects depend on previous-token ending predicting next-token beginning.

Reverse the sequence and the conditioning direction changes.

For circular text, this means the clockwise/anti-clockwise choice can be tested by how well established token-edge dependencies generalise.

If one direction preserves the same final→initial relations seen in prose and the reverse does not, the first direction gains structural support.

This is stronger than deciding direction from visual convenience alone.

Known prose behaviour can calibrate diagram text.

Direction Changes the Search Space for Decipherment

A decipherer who allows every circular string to be read:

from any starting point;
in either direction;
with several segmentation choices;
with optional wrap-around;

has created a huge freedom budget.

A desired word can be found by rotating, reversing and resegmenting until something resembles it.

This is especially dangerous for proper-name claims on star and zodiac labels.

A strong decipherment should constrain orientation from the manuscript before searching semantic candidates.

Every free rotation and reversal should be treated as a model parameter, not forgotten after the match is found.

Direction discipline reduces false discovery.

Direction Can Help Test Cipher Hypotheses

Some ciphers reverse strings.

Some transposition systems reorder groups.

Therefore a cipher theory may legitimately claim that visible writing order differs from plaintext order.

But the transformation must be systematic.

If prose is decrypted left-to-right but circular text is reversed whenever needed, the model has gained a patch.

A real cipher should specify:

input order;
encryption order;
output writing order;
how circular/radial geometry affects the procedure.

The same rule should explain unseen loci.

Reversal is a mechanism only when it is rule-bound.

Direction Can Help Test Natural-Language Hypotheses

If Voynich represents language with relatively direct spelling, the visible direction should preserve recurring morphological asymmetries.

Prefixes should remain prefix-like.

Suffix-like endings should remain endings.

Function-word classes should occur in consistent positions.

A reversed reading that makes every known positional family less coherent is unlikely.

For circular text, compare both directions against the prose-derived morphology.

The direction that preserves the manuscript’s broader structural grammar has an advantage.

Again, direction can be inferred statistically without translating the words first.

Direction Can Help Test Generated-System Hypotheses

A procedural generator may have an explicit production direction.

Copy the previous token.

Modify its ending.

Write the next token to the right.

On a circle, continue clockwise.

If such a mechanism is real, local similarity should be stronger in the production-forward direction than backward.

Directional asymmetry becomes a test.

If copying relations are equally strong both ways, a strictly sequential generation mechanism weakens.

The geometry can therefore expose causal direction as well as reading direction.

The Start Point Is Part of a Numerical or Coordinate Theory

The Numbers article showed that circular position can encode ordinal value.

But ordinal value requires an origin.

Degree 1 must be somewhere.

Day 1 must be somewhere.

Sector 1 must be somewhere.

If a zodiac ring has thirty figures but no independently identified start, assigning label N to degree N is underdetermined up to rotation.

Direction doubles the ambiguity.

A numerical-coordinate theory therefore needs:

origin;
direction;
step rule.

Without all three, numerical values can be rotated around the circle until the desired match appears.

Direction Is One Reason Circular Text Needs Its Own Data Type

Modern IVTFF’s distinction among paragraph, label, circular and radial loci is not administrative fussiness.

It preserves different geometries of information.

A circular locus should carry:

direction;
start marker or start uncertainty;
wrap-around relationship;
circle identity;
sector relationship.

A radial locus should carry:

inward/outward direction;
centre endpoint;
outer endpoint;
associated diagram element.

Once these fields are stripped away, later statistical analysis cannot recover them from the string alone.

Representation should preserve the source’s directional affordances.

What Survives the Direction Work

Ordinary Voynich running text is strongly left-to-right and top-to-bottom.
This is a property of the visible writing system, not proof of source language or provenance.
Circular and radial text are recurring manuscript text modes and need separate geometry-aware treatment.
The current IVTFF inventory records 84 circular loci: 83 clockwise and one anti-clockwise.
The sole anti-clockwise locus is on the Rosettes sheet.
The current inventory records 142 radial loci: 75 inward and 67 outward.
Radial orientation therefore varies substantially by local context.
A circle has direction but no inherent linear start point.
Some diagrams provide start cues; others remain more ambiguous.
Rotation and reversal introduce degrees of freedom that must be controlled in comparison and decipherment.
Wrap-around adjacency is real and should survive digital transcription.
Writer movement, glyph orientation and semantic reading order are related but not identical concepts.
Page rotation and reader ergonomics can explain some orientation without semantic reversal.
Direction can be tested against manuscript-wide morphology and edge-to-edge dependencies.

What Does Not Survive as Established Knowledge

Left-to-right Voynich proves a Latin or Romance language.
Every circular locus is clockwise.
Every radial locus reads outward.
Inward writing means input.
Outward writing means output.
Clockwise means forward in the manuscript’s semantics.
Anti-clockwise means reverse or negative.
The modern first token of every circular transliteration is the historical first token.
Writing direction always equals semantic reading order.
Rotating a page or foldout implies hidden encryption.
Labels arranged clockwise form one continuous sentence.

The directional geometry survives.

The semantic arrow does not.

A Better Reading-Direction Analysis

Identify the locus type before linearising text.
Use palaeographic stroke direction where available.
Record clockwise versus anti-clockwise circular direction explicitly.
Record inward versus outward radial direction explicitly.
Identify physical start markers separately from modern transcription start points.
Preserve wrap-around adjacency.
Test all rotations transparently when no origin is secure.
Count reversal and rotation as model freedom.
Keep label order separate from prose syntax unless dependence is demonstrated.
Compare direction with q-, minim-, gallows- and edge-position grammar.
Test whether proposed semantic direction generalises across independent diagrams.
Keep page orientation and reader ergonomics in the physical model.

What Would Count as a Real Direction Breakthrough?

Imagine a syntax model is learned entirely from ordinary left-to-right prose.

It identifies robust token-edge transitions and structural classes.

The model is then applied to circular loci without using their current transcription direction as truth.

For 83 circles, the clockwise orientation produces the expected syntax far more strongly than reversal.

For the one Rosettes anti-clockwise locus, the anti-clockwise orientation uniquely restores the same structural grammar.

Radial loci similarly separate into inward and outward orientations that preserve common token structure.

Now direction has been validated by an independent property of the writing system.

Or a decipherment may recover a calendrical or coordinate sequence whose known ordering independently fixes both origin and direction on a ring.

That would be equally powerful.

A real direction breakthrough happens when the manuscript’s geometry and the recovered information order independently point the same way.

Primary School: Which Way Around the Circle?

Write six letters around a paper plate.

Ask a child to read them clockwise.

Then anti-clockwise.

The same letters produce different sequences.

Now ask where the sequence starts.

Without a marker, several answers are possible.

That is the circular Voynich problem in one exercise.

Lower Secondary: Separate Direction From Orientation

Write a sentence around a circle so each word remains upright relative to the tangent.

The word orientations rotate around the page.

The sequence direction can still remain consistently clockwise.

Students discover that “this word is upside-down on my screen” and “this sequence reads backward” are different statements.

Upper Secondary: Count the Search Freedom

Give students a circle of 20 tokens and a target pattern.

Allow one team to choose one fixed start and direction.

Allow another team to test all 20 rotations in both directions.

The second team finds more impressive coincidences.

Not because its theory is better.

Because it searched forty times as many alignments.

This is why rotation and reversal freedom must be counted in Voynich comparisons.

JC and Adult Readers: Circular Text as a Cyclic Sequence

At a higher level, a circular locus should be modelled on a cycle graph rather than a path graph.

A path has endpoints.

A cycle connects the final node back to the first.

Statistical measures can then be made rotation-invariant while preserving local order.

If a physical start marker exists, it can be added as an anchored node rather than assumed by the file format.

This mathematical representation is closer to the manuscript geometry and reduces artificial boundary effects.

A Parent and Teacher Guide

Begin with ordinary prose to establish the dominant visible direction.
Separate writing direction from source-language identity.
Treat circular and radial text as their own geometries.
Always record clockwise/anti-clockwise and inward/outward orientation.
Ask where a circle starts before turning it into a line.
Preserve wrap-around neighbours.
Count every allowed rotation and reversal as analysis freedom.
Do not turn inward/outward into semantic input/output without repeated evidence.

The broader lesson is powerful:

before asking what a sequence means, make sure you have not created the sequence by choosing the direction, start point and geometry yourself.

Reader Checklist: Before You Read a Voynich Sequence

Is this paragraph, label, circular or radial text?
What direction is visible from the handwriting?
Is the circular locus clockwise or anti-clockwise?
Is the radial locus inward or outward?
Is there a physical start marker?
Is the modern transcription start arbitrary?
Is wrap-around adjacency preserved?
Has the page or foldout been rotated in the digital display?
Could reader ergonomics explain orientation?
Does the proposed direction preserve q/minim/gallows positional structure?
Does it preserve final→initial token-edge dependencies?
How many rotations and reversals were searched?
Is label adjacency being mistaken for sentence order?
Does the direction rule predict unseen loci?

Frequently Asked Questions

Is Voynichese written left-to-right?

Ordinary running text overwhelmingly is. Lines proceed left-to-right and paragraphs top-to-bottom.

Does that prove a European language?

No. It identifies the visible script direction. A cipher, shorthand or adapted script can use a direction different from that of its source language.

How does circular Voynich text run?

The current IVTFF reference inventory records 84 circular loci: 83 clockwise and one anti-clockwise.

Where is the anti-clockwise example?

It is the lower short circular text in the small lower-left circle on the Rosettes sheet.

How does radial text run?

Both ways. Current reference data record 75 inward radial loci and 67 outward loci.

Where does a circular text begin?

Sometimes a graphical start point can be identified; sometimes it is less clear. A modern transcription must not silently turn an arbitrary cut point into a historically meaningful beginning.

Why does direction matter for decipherment?

Because reversing or rotating a sequence changes adjacency, prefix/suffix position, syntax and possible matches. Free direction choices greatly increase the chance of false alignments.

What is the strongest current conclusion?

Voynich uses a stable left-to-right prose convention plus geometry-specific circular and radial conventions. Reading order must therefore be represented locally rather than reduced to one universal linear sequence.

Research and Further Reading

The Final Idea

Most of Voynich begins in the reassuring place.

Left.

It moves right.

Then down.

And then the manuscript reminds us that it is more than prose.

A sentence bends into a circle.

A label points inward.

Another points out.

A foldout asks the reader to rotate the object.

One tiny Rosettes locus even turns the other way.

The danger is not that Voynich has no direction.

It has several kinds of direction.

Our job is to preserve which kind we are looking at.

Before decoding a Voynich sequence, first make sure the sequence has not been created by flattening a circle, choosing a convenient start and reading it whichever way makes the theory work.

Voynich | Everything eduKate Knows and Tested | Which Way Does Voynich Read? Direction, Rotation and the Geometry of Sequence