A photograph freezes handwriting.
A scribe did not.
The Voynich Manuscript reaches us as finished marks on parchment, but every mark was once a movement.
Quill touches skin.
Moves upward.
Turns.
Crosses.
Lifts.
Returns.
What we now call one glyph may have been made in one continuous stroke, two strokes, three strokes or a compound sequence inherited from a scribal habit we have not reconstructed.
Palaeographers have a word for the characteristic way a written form is produced:
ductus.
Ductus includes stroke order, stroke direction, rhythm, pen lifts and the habitual movements by which a writer turns an abstract sign into ink.
For Voynich, this matters far beyond identifying scribes.
If we do not know how a sign was constructed, we may not yet know whether it is one sign.
Quick Read
- Ductus is the physical sequence and manner of pen movements used to produce written forms.
- Two glyphs that look nearly identical when finished can be produced by different stroke sequences.
- Two glyphs that look different can sometimes be variants generated by the same basic movement.
- Lisa Fagin Davis’s palaeographic work on Voynich uses features such as crossbar shape, stroke continuity and pen lifts as evidence for scribal differentiation.
- A frequently discussed example is EVA-k: a bowed crossbar can result from a continuous movement from the first vertical, while a flatter horizontal crossbar is more compatible with a pen lift and new stroke.
- This is evidence about writing behaviour, not phonetic value.
- Ductus can help test whether apparent ligatures, benches and complex gallows are single units or compounds.
- It can also distinguish a character difference from a scribal allograph.
- The influential five-scribe model uses palaeographic tendencies, but the exact number of scribes remains debated; ductus should therefore be treated as evidence to test hand models rather than as proof of a sacred count.
- Digital images can hide pen-pressure, stroke-order and surface information. High-resolution, multispectral and raking-light evidence can matter.
- A future glyph model should preserve both shape and construction.
Shape and Movement Are Different Evidence
Draw a lowercase “t”.
You might draw the vertical first and add the crossbar later.
Someone else might make a looped cursive movement that produces both in one gesture.
The finished letters can look similar.
The writing processes are not.
Now reverse the problem.
One writer produces a hurried form.
Another produces a careful form.
The finished marks look different enough that a transcriber assigns two symbols.
But if both arise from the same basic stroke sequence under different speed or angle, the difference may be palaeographic rather than alphabetic.
This gives us two distinct questions:
- graphical identity: what shape is visible?
- gestural identity: how was that shape probably produced?
Voynich research often has the first.
Ductus tries to recover the second.
Why Ductus Matters in an Undeciphered Script
In a readable medieval manuscript, palaeographers have an advantage.
They know which letters they are looking at.
If two forms both spell the same Latin word in the same position, shape variation can be interpreted with linguistic context.
Voynich removes that safety net.
A shape difference could indicate:
- a different character;
- an allograph of the same character;
- a ligature;
- an abbreviation;
- a decorative initial;
- a scribal habit;
- a damaged form;
- a transcription mistake.
Ductus helps because it asks a question independent of meaning.
Did the hand make these forms in the same way?
The EVA-k Example: A Crossbar Can Preserve a Pen Lift
Lisa Fagin Davis’s formal palaeographic work made this point especially clear with the Voynich form conventionally transcribed as EVA-k.
One diagnostic question is the form of the crossbar between the two vertical elements.
In some hands, the crossbar appears bowed or smoothly descending from the top of the first vertical.
That geometry is compatible with a single continuous movement: the pen rises, turns and flows into the crossbar.
In other examples, the crossbar is flatter and more independent.
That is more compatible with the writer finishing the first vertical, lifting the quill and beginning a separate stroke.
The difference matters because it is not merely decorative appearance.
It records a motor habit.
A writer can consciously imitate a letter shape more easily than they can consistently erase every habitual micro-sequence used to produce it.
That is why ductus can be useful for distinguishing hands.
Ductus Is a Tendency, Not a Fingerprint Scanner
This warning is essential.
A human does not write every character identically.
Speed changes.
Ink flow changes.
Parchment texture changes.
The wrist changes angle near a margin.
A writer becomes tired.
A crowded line encourages compression.
Therefore one bowed crossbar does not identify one scribe.
Davis herself has stressed that palaeographic conclusions rest on tendencies visible across a body of writing rather than on a single isolated form.
The correct unit is distribution.
How often does a writer choose one stroke sequence?
Under what conditions?
Which other features co-vary?
Ductus is strongest when several independent habits travel together.
The Five-Scribe Debate Makes Ductus More Important, Not Less
Davis’s 2020 analysis proposed five distinguishable scribal hands across the manuscript.
That model has been influential because it gave Voynich research a practical palaeographic partition independent of semantic interpretation.
Subsequent work has questioned whether all of the variation is best represented as five discrete people rather than fewer hands or continuous scribal change.
This debate should not be framed as:
ductus works versus ductus fails.
The real question is statistical and palaeographic:
Do clusters of stroke habits form discrete groups, gradual continua or a mixture of both?
That is precisely the kind of question ductus can help answer.
See The Scribe Problem for the disputed count.
A Pen Lift Can Change the Alphabet Count
This is where palaeography meets the Alphabet Problem.
Suppose a complex shape consists of two strokes that are always produced as one integrated movement and never occur independently.
That supports treating the complex shape as one unit.
Now suppose a visually similar complex form is made by writing one common glyph, lifting the pen, then adding another common glyph.
That supports a compound analysis.
The finished bitmap may be almost identical.
The underlying character model is different.
That changes:
- character counts;
- bigram frequencies;
- entropy;
- token length;
- candidate cipher mappings;
- morphological analyses.
A microscopic pen lift can therefore propagate into a manuscript-scale statistical result.
Benches and Gallows: One Glyph, Two Glyphs or a Construction?
The bench characters and complex gallows are obvious places where ductus matters.
A bench-like form conventionally transcribed as ch or sh may be treated by one model as two linked components and by another as one unit.
A gallows rising through a bench can be:
- a compound glyph;
- a ligature;
- a base sign plus modifier;
- one indivisible character with a complex stroke path.
Static appearance alone does not settle this.
Ductus can ask:
- which stroke is laid down first;
- whether the pen lifts between components;
- whether one line visibly crosses over another;
- whether the components retain the same movement when written independently;
- whether the compound has its own special motor pattern.
See The Bench Characters and The Gallows Characters.
A Ligature Is About Movement as Much as Shape
A ligature is not simply two letters touching.
Historically, letters can fuse because the writing movement itself joins them.
The pen leaves one form and naturally enters the next without fully resetting.
Over time the fused movement can become a conventional sign.
This creates a continuum:
two separate letters → touching letters → habitual ligature → conventional single-looking sign
Voynich may contain forms somewhere along that continuum.
If so, arguing whether the “alphabet has 20 or 25 characters” may be asking a crisp question of a historically fuzzy boundary.
Ductus Can Reveal Allographs
An allograph is a different written form of the same underlying grapheme.
English handwritten “a” provides an easy example: people can write a one-storey or two-storey form while still intending the same letter.
In Voynich, a rare form can look sufficiently unusual that a transcriber gives it a new code.
Ductus may reveal that the rare form is simply what happens when:
- the pen is not lifted;
- the crossbar begins higher;
- the writer compresses two strokes;
- the sign occurs at a line boundary;
- the hand moves faster.
This can shrink an apparent alphabet without translating anything.
The Rare Glyphs article owns the inventory problem. Ductus supplies one way to adjudicate it.
Ductus Can Also Reveal Genuine Differences
The same logic works in the opposite direction.
Two finished shapes may look like minor variants.
But if one is consistently produced:
- with one stroke rather than two;
- from bottom-to-top rather than top-to-bottom;
- with a recurring pen lift at a stable point;
- with a different joining movement into neighbouring glyphs;
then the distinction may be functionally real even if the outlines overlap.
This is why tracing only bounding boxes and pixel similarity can be misleading.
Handwriting is trajectory.
Stroke Order Is a Hidden Time Axis
A manuscript page looks two-dimensional.
Ductus adds time.
Stroke A happened before stroke B.
One line lies over another.
One curve begins where another stops.
The ink flow thickens as the nib changes direction.
The quill lifts and leaves a tiny discontinuity.
This is a miniature version of the manuscript’s larger Order of Making problem.
At manuscript scale we ask drawing before text or text before paint.
At glyph scale we ask vertical before crossbar or crossbar before vertical.
Ductus is codicology at the scale of a few millimetres.
Writing Speed Changes Ductus
A hurried writer tends to reduce pen lifts.
Separate strokes merge.
Angles become curves.
Verticals lean.
Loops narrow.
A careful display initial may do the opposite.
It can acquire extra strokes and deliberate separation.
This matters for Voynich because some visual differences may correlate with document position rather than writer identity.
Paragraph initials, labels, circular text and dense prose can all encourage different movement.
Therefore a hand classifier must control for writing task.
Otherwise “different scribe” may partly mean “same scribe writing a different kind of locus”.
Parchment Can Alter the Stroke
The writing surface is not inert.
Parchment varies in thickness, smoothness and preparation.
A quill can catch on a rougher patch.
Ink can spread differently.
A hole or defect can force a changed stroke path.
Hair-side and flesh-side surfaces can feel different under the pen.
This means one odd stroke should not automatically become a new character or new hand.
Material context belongs in the palaeographic model.
See Ink, Pigment, Parchment and the Physical Evidence.
Why Digital Scans Are Necessary and Insufficient
High-resolution images changed Voynich research.
They let scholars compare thousands of forms without repeatedly handling the fragile original.
But a normal photograph can flatten information relevant to ductus.
- A shallow dry impression may disappear.
- A pen-lift seam may be hidden by ink spread.
- Overlapping strokes can become difficult to order.
- Parchment curvature can distort angle.
- Lighting can exaggerate or suppress nib-edge contrast.
Where a distinction depends on stroke order or physical overlap, multispectral imaging, microscopy, raking light and direct conservation observation may be needed.
The scan is evidence.
The object retains the right to overrule it.
Could AI Infer Ductus From Static Images?
Sometimes.
A model can learn correlations between curvature, tapering, overlap and likely stroke sequence from known handwriting.
It can cluster thousands of Voynich glyph instances by shape and estimate whether certain forms occupy continuous or discrete families.
But reconstruction from one static image is inherently uncertain.
Several stroke sequences can produce nearly the same final mark.
Therefore computational ductus inference should report probabilities and alternatives rather than silently converting one guessed stroke path into ground truth.
This is exactly the kind of problem where uncertainty should be preserved as data.
Ductus and the Grove Words
The Grove-word problem from the previous article asks why paragraph-initial tokens often acquire or favour tall gallows forms.
Ductus adds a new possibility.
Perhaps a paragraph-initial gallows is not simply an extra character.
It may be an elaborated way of entering the first token.
A new paragraph invites a slower, taller or more display-like first movement.
If that is true, some apparent token transformation could be partly a writing-gesture transformation.
The hypothesis predicts that paragraph-initial gallows should show ductus features different from identical-looking gallows in ordinary interiors.
That is measurable.
See The Grove Words.
Ductus and the Exemplar Problem
If a scribe is copying an unfamiliar exemplar, writing behaviour may change.
A familiar script can be written fluently as motor chunks.
An unfamiliar script may require slower visual copying, more pen lifts and less consistent joining.
Voynich often looks remarkably fluent.
That observation has encouraged the idea that its writers were comfortable with the script.
But fluency does not prove semantic understanding.
A trained copyist can become fluent in a notation they do not fully understand.
Ductus could help compare:
- routine high-frequency forms;
- rare glyphs;
- special labels;
- complex gallows;
- later marginal writing.
If rare forms show hesitation while common forms flow, the scribe’s motor familiarity may have been frequency-dependent.
See The Exemplar Problem.
The Pen-Lift Ledger
A serious future study could treat pen lifts as an explicit dataset.
For each selected glyph instance, record:
- folio and line;
- candidate glyph identity;
- scribe/hand model;
- Currier regime;
- document role;
- estimated stroke count;
- probable pen-lift locations;
- crossing order;
- bowed versus straight crossbars;
- entry stroke;
- exit stroke;
- connection to neighbouring signs;
- surface defects;
- confidence in each inference.
Then test whether the distribution clusters by:
- proposed hand;
- glyph class;
- page function;
- line position;
- quire;
- time/production sequence.
This would let the manuscript tell us whether shape categories correspond to stable motor categories.
What Ductus Does Not Prove
- It does not translate a glyph.
- It does not identify a sound.
- It does not prove one visible form is one linguistic letter.
- It does not by itself prove five scribes.
- It does not by itself prove one scribe.
- It does not distinguish author from copyist.
- It does not prove the writers understood the content.
- It does not turn every pen lift into a character boundary.
What It Can Give Us
- A physically grounded test of glyph segmentation.
- Evidence for or against allographic equivalence.
- Independent features for scribal-hand comparison.
- Clues about writing speed and fluency.
- Evidence about ligatures and compounds.
- A bridge between static transcription and the actual movement that produced the script.
- A way to test whether boundary-specific forms are genuinely separate characters or altered writing gestures.
Primary School: Draw the Same Letter Two Ways
Ask a child to draw a capital A without lifting the pencil.
Then draw another A using three separate strokes.
The final shapes can be similar.
The movements are different.
That is ductus.
Secondary School: Hide the Stroke Order
Have three students write the same symbol ten times.
Photograph only the finished symbols.
Ask another student to group them by writer.
Then reveal video of how each symbol was made.
The movement evidence can make distinctions clearer than the outline alone.
JC and Adult Readers: Treat Stroke Sequence as a Latent Variable
For a static manuscript image, true stroke order is often hidden.
Model several possible orders.
Score them using:
- ink overlap;
- curve continuity;
- nib-angle consistency;
- known writer habits;
- comparison with repeated instances.
Then propagate uncertainty downstream.
If a glyph segmentation depends on one uncertain ductus reconstruction, the statistical conclusion should remain correspondingly uncertain.
Reader Checklist: Before You Split a Voynich Glyph
- Does the form recur?
- Do its apparent components occur independently?
- Is a pen lift visible or only assumed?
- Which stroke crosses which?
- Could one continuous movement produce the whole shape?
- Does the same hand construct it consistently?
- Does the form change with speed or page position?
- Does parchment damage affect the trace?
- Does another transcription segment it differently?
- Does the statistical result survive both segmentations?
- Is the distinction palaeographic or claimed to be semantic?
- What observation would prove the current segmentation wrong?
Frequently Asked Questions
What is ductus?
Ductus is the characteristic manner in which writing is executed: stroke order, direction, rhythm, pen lifts and related movement habits.
Why does it matter for Voynich?
Because the manuscript is unread. Ductus offers evidence about glyph construction and scribal behaviour that does not require knowing what the signs mean.
Can ductus identify the scribes?
It can contribute powerful features to hand identification, especially when several tendencies co-vary over large samples. The exact number of Voynich scribes remains debated, so no single ductus feature should be treated as a personal fingerprint.
Can it tell whether ch or a complex gallows is one character?
It can help. Stroke continuity, pen lifts and repeated construction patterns can support either a single-unit or compound interpretation, but function also depends on broader distributional evidence.
Does a pen lift always mark a new glyph?
No. Many ordinary letters require multiple strokes. Pen lifts are evidence about construction, not automatic character separators.
Research Foundations
- Lisa Fagin Davis — How Many Glyphs and How Many Scribes? Digital Paleography and the Voynich Manuscript, Manuscript Studies 5.1 (2020).
- Lisa Fagin Davis — Voynich Paleography, keynote paper in the 2022 International Conference on the Voynich Manuscript proceedings.
- VOY2022 Proceedings, for the broader palaeographic and computational research context.
- René Zandbergen — The Voynich Writing System.
The Final Idea
A glyph is a fossil of a movement.
The ink preserves the result.
Palaeography tries to recover the gesture.
For Voynich, that gesture may tell us whether two signs share a writer, whether two shapes share a character identity, whether a compound is really compound and whether the alphabet we count on a screen corresponds to the units the scribe actually wrote.
Before asking what a Voynich glyph means, sometimes the better question is: what did the hand actually do?