Voynich and Padua | The Label Problem

Voynich has two visibly different textual populations.

There is running text.

And there are short strings placed beside plants, stars, figures, vessels and diagram elements.

We usually call the second population “labels”.

That word is useful.

It is also a hypothesis.

A short string beside an image does not have to name the image. It can encode a category, quantity, instruction, stage, coordinate, cross-reference or local parameter.

Quick Read

Voynich short strings occur in image-rich contexts that differ structurally from running prose.
Known medieval technical manuscripts often use short text units as part of diagrams, tables and medical images.
Zodiac Man and Vein Man controls show that labels can identify bodily zones, signs or procedural locations.
Folding almanacs combine diagrams, tables and compact explanatory text for trained users.
Laubach 1425 demonstrates a manuscript where technical images, multilingual text and diagrammatic information coexist.
Therefore “short word beside picture = noun naming picture” is too narrow.
The high-value test is whether repeated Voynich short strings recur in repeated visual roles.
If label families track image classes, positions or operations, the label system may reveal semantics before a full decipherment.

A Label Is a Relation

The word “label” makes us imagine a museum tag.

Object.

Name.

But technical manuscripts often do something richer.

A short text unit may tell the reader:

what a component is;
which sign governs it;
which stage comes next;
which category it belongs to;
which measurement applies;
where another entry can be found;
when the operation should be performed.

The label therefore cannot be interpreted without its relation to the surrounding diagram.

The Zodiac Man Shows Labels as Coordinates

Fifteenth-century Zodiac Man diagrams place signs or annotations around bodily regions.

The text is short because the picture supplies most of the structure.

The reader does not need a paragraph explaining where the knees are.

The image supplies location.

The annotation supplies category.

diagram position carries part of the meaning that prose would otherwise have to say.

This is an important model for Voynich’s zodiac labels.

A short string may be meaningful only together with ring, figure and sector position.

The Vein Man Shows Labels as Procedure

Medical almanacs often include a Vein Man showing locations for bloodletting.

Here a body image serves as an operational map.

A short annotation can specify a site or medical relation rather than name an object.

That gives us another Voynich possibility:

short text beside a figure may tell the reader what to do with the position rather than who the figure is.

Tables Compress Meaning Even Further

Wellcome MS.8515, compiled around 1425 in southern England, is a practical manual of calendar computation, astronomy and astrological medicine.

Tables carry relationships by row and column.

A cell can contain a tiny symbol or number whose meaning depends entirely on its coordinate.

This is the same principle at a higher level:

position is syntax.

Voynich’s labels may likewise be inseparable from page geometry.

Running Text and Labels May Be Different Interfaces to the Same Vocabulary

Voynich research has long noticed that labels and running text have different statistical behaviour.

That difference does not necessarily imply two languages.

Labels naturally have different distributions because they do different jobs.

Names repeat differently from verbs.
Coordinates repeat differently from prose.
Quantities repeat differently from descriptions.
Category codes repeat differently from explanations.

The correct question is functional before linguistic.

The Plant Label Problem

Short words near Voynich plants are often assumed to be plant names.

That is plausible.

But a botanical or pharmaceutical page could also label:

the medicinal part;
a preparation;
a synonym;
a quality;
a dosage class;
a cross-reference to a recipe;
a harvest time.

The Nature-Print control strengthens this caution because one plant can carry several names across languages and technical traditions.

The Vessel Label Problem

Voynich’s pharmaceutical pages include vessels accompanied by short strings.

If the vessel is a category marker, the label may name a preparation rather than a container.

If the vessel is a storage object, the label may identify contents.

If the vessel is symbolic, the label may perform an entirely different role.

Only repeated relationships can discriminate these models.

The Human-Token Label Problem

The Planet Children control shows why a label beside a human figure need not be a personal name.

The figure can instantiate a class.

The text can identify the class or parameter.

That gives us several testable models for Voynich zodiac labels:

name;
day;
degree;
decadal unit;
planetary class;
quality;
calendar segment.

A Label Should Predict Its Context

If a short Voynich string has stable meaning, it should show contextual regularity.

For example:

the same string near the same type of plant structure;
the same family near the same ring position;
the same prefix near the same vessel class;
the same ending in repeated celestial sectors.

This makes labels a high-value test population.

They can potentially anchor semantics without requiring full prose translation.

The Position-Ablation Test

One powerful experiment is simple.

Remove the image context.

Can the labels still be classified?

Then restore position.

Does classification improve?

If yes, spatial placement contains information.

This would be strong evidence that page geometry is part of Voynich syntax.

The Padua Question Becomes Secondary

Label architecture is widespread across technical manuscript cultures.

It does not by itself localise Padua.

Only after label function is recovered can geography re-enter through:

local vocabulary;
regional abbreviation habits;
specific astronomical parameters;
local plant names;
documented technical traditions.

The Label Test

Separate labels from running text without assuming meaning.
Cluster labels by image context.
Cluster by page position.
Measure recurrence across sections.
Compare morphology with running-text words.
Test whether image class predicts label class.
Test whether label class predicts a technical operation.
Only then test language and geography.

What Technical Labels Can Legitimately Support

Short text can be structurally different from prose without being a different language.
Position can carry syntax.
Human, plant and diagram labels can encode technical categories rather than names.
Voynich labels are a high-value population for functional analysis.

What They Cannot Support

that every short string is a noun;
that labels name nearby objects;
that label vocabulary is Paduan;
that different label statistics imply a separate language;
that a successful image association is a full decipherment.

Research Sources

Continue the Series

The Final Idea

The label may be the smallest useful unit in the manuscript.

Not because it is easy to translate.

Because a label has two sources of information at once: what is written and where it is placed. Voynich research should exploit both before deciding what the word means.

Voynich and Padua | The Label Problem: When Short Text Is Part of the Diagram