Tangential Voynich object-world experiment 08. This article does not claim that the Voynich Manuscript is a Russian nesting doll, that fifteenth-century makers knew the modern Matryoshka tradition, or that the codex contains literal nested objects. The familiar Russian nesting doll is a much later object. The Museum of Russian Art notes that its early history belongs to the late nineteenth century and that parts of its origin story remain matters of legend. That historical distance is exactly why it is useful here. We borrow only the logic of recursion, nested scale and repeated reveal. The Museum of Russian Art: A Thousand Nesting Dolls.
Previous: Automaton or Marvel · Tangential Voynich | Systemic Parallax · Voynich Research Library
What If the Same Problem Keeps Appearing Inside Itself?
Open a nesting doll.
There is another doll.
Open that one.
Another.
Each level is recognisably related to the previous one, but it is not identical. Scale changes. Surface detail changes. The act of opening changes the receiver’s model of the object. What first appeared to be one thing becomes a hierarchy.
The Voynich Manuscript repeatedly confronts us with a similar analytical temptation.
We want to decide what the unit is.
Glyph?
Multi-glyph unit?
Apparent word?
Line?
Paragraph?
Page?
Bifolio?
Quire?
Visual section?
Manuscript?
What if the correct answer is not one level?
What if Voynich has structure inside structure, and different rules become visible at different scales?
The Scale Error
Many scientific disagreements are really disagreements about scale.
A molecule is not an organ.
A neuron is not a mind.
A household is not a national economy.
A single railway station is not a metropolitan transport network.
The same system can obey different useful descriptions at different resolutions.
Voynich research risks a scale error whenever a statistic measured at one level is promoted into a statement about the whole manuscript.
Low glyph-level conditional entropy does not directly tell us the structure of page-level organisation.
Page-level clustering does not identify the function of an individual token.
Bifolio-level scribal coherence does not make every line semantically homogeneous.
The nested-object tangent therefore makes scale an explicit axis rather than a background choice.
The Voynich Scale Stack
We can define a provisional stack without claiming that every layer is functionally real:
L₀ — stroke or primitive mark
L₁ — glyph class
L₂ — recurring multi-glyph unit
L₃ — apparent blank-delimited token
L₄ — line
L₅ — paragraph or local block
L₆ — page interface
L₇ — bifolio / physical conjunction
L₈ — quire / local codicological unit
L₉ — manuscript-wide population
The important phrase is provisional stack.
Some levels may later merge. Some may split. Some visible boundaries may not correspond to functional boundaries. But placing them side by side lets us ask whether structural phenomena recur across levels.
The First Test: Cross-Scale Recurrence
A nesting doll repeats a recognisable form at different sizes.
Voynich may or may not do anything analogous.
The test is deliberately neutral:
Do the same kinds of structural distinction recur at more than one analytical scale?
For example:
- Glyphs may have privileged initial and terminal positions.
- Apparent tokens may have privileged line positions.
- Lines may have privileged paragraph positions.
- Pages may have privileged quire positions.
- Quire transitions may reorganise distributions.
If boundary sensitivity repeats across several levels, we have not discovered “nested meaning.” We have discovered cross-scale boundary recurrence.
That would become a strong invariant candidate because it is independent of any one tokenisation.
The Second Test: Self-Similarity Without Identity
Self-similarity does not require exact copies.
A coastline can have related roughness at different scales without reproducing the same bay.
A branching tree has twigs, branches and limbs organised by related but not identical geometries.
We can therefore ask whether Voynich has statistical self-similarity.
Take a structural measure M at scale L:
M(L)
Now observe how that measure changes when the scale changes.
Possible measures include:
- entropy;
- transition concentration;
- boundary asymmetry;
- degree distribution in co-occurrence networks;
- community structure;
- recurrence spacing;
- visual branching complexity;
- rarity distribution.
If a related organisational signature survives several levels, the important discovery is not that Voynich “is fractal.” That word would overclaim.
The safe residue is scale-stable organisational signature.
The Third Test: Nested Segmentation
Segmentation is one of the manuscript’s deepest problems.
We see spaces.
We see lines.
We see paragraphs.
But visual separation does not guarantee functional separation.
The nested-object world asks whether segmentation should be hierarchical.
Instead of:
text → words
consider:
text → blocks → lines → units → sub-units
Each layer may have its own boundary strength.
A blank might be a weak boundary inside a larger structural unit. A line ending might be a stronger reset. A page transition might alter only some variables. A quire boundary might be physical without being semantic.
This suggests estimating a boundary hierarchy rather than a binary boundary/no-boundary model.
The Fourth Test: The Reveal Depth of Meaning
Nesting creates delayed context.
The outer object does not tell you how many inner objects remain.
Could Voynich interpretation require a similar sequence?
Perhaps a glyph is uninterpretable until its local multi-glyph unit is known.
Perhaps a token’s role is uninterpretable until its line position is known.
Perhaps a line’s role is uninterpretable until the page interface is known.
Perhaps a page’s role is uninterpretable until its manuscript region is known.
This is a conditional-information problem.
For a unit X, ask how prediction improves as larger context C₁, C₂, … Cₙ is supplied:
H(X | C₁) > H(X | C₁,C₂) > …
If adding page, hand, line position or visual regime sharply reduces uncertainty, the system has contextual reveal depth.
Meaning remains unknown. But dependency depth becomes measurable.
The Fifth Test: Does the Outer Layer Predict the Inner?
In a deliberately nested design, outer structure may constrain what can exist inside.
Page morphology may predict token populations.
Scribal hand may predict local orthographic preferences.
Quire membership may predict visual grammar.
Or these associations may be weak and asynchronous.
The frozen matrix already suggests that some manuscript classifications align strongly while distributional clusters do not simply collapse into the same divisions.
The nested experiment asks us to quantify parent-to-child predictability:
P(inner class | outer class)
If outer classes strongly constrain inner behaviour, a hierarchical model may be appropriate.
If not, the manuscript may be better represented as overlapping networks rather than a strict tree.
The Sixth Test: Recursion Versus Repetition
Repetition is not recursion.
This distinction matters.
A wallpaper pattern repeats.
A recursive structure contains related instances of a rule inside larger instances of the rule.
Voynich contains a great deal of recurrence. To claim recursion, we would need stronger evidence:
- similar structural relations at different scales;
- predictable nesting boundaries;
- parent-child constraints;
- repeated transformation rules rather than repeated surface forms;
- a coherent stopping condition.
Without those, “nested” remains only a metaphor.
The Seventh Test: The Stopping Rule
Every physical nesting doll eventually ends.
There is a smallest doll.
Scientific analysis also needs stopping rules.
How far should we decompose Voynich?
If every glyph can be decomposed into strokes, and every stroke into pen motions, and every motion into microscopic trajectories, decomposition can continue indefinitely while moving farther from functional units.
The correct unit is not necessarily the smallest observable unit.
We need a functional stop criterion.
One candidate:
Stop decomposing when further subdivision no longer increases held-out predictive power enough to justify the added complexity.
This converts the Russian-doll metaphor into model selection.
The Eighth Test: Nested Surprise
The nesting doll is not merely hierarchical.
It is sequentially revelatory.
Each opening confirms the pattern—another related object—while also creating uncertainty about how long the pattern will continue.
This is a delicate cognitive balance:
enough repetition to create expectation + enough variation to keep attention alive
That balance may become central to our final fascination article.
Voynich repeatedly gives the observer something familiar enough to classify, then strange enough to resist exhaustion.
A plant-like drawing.
But not quite a plant we can securely identify.
A star-like circular system.
But not quite a familiar astronomical diagram.
Word-like strings.
But not quite ordinary words under our present segmentation assumptions.
Whether this balance was historically intentional is unknown.
But the balance itself can be measured as prediction-with-residual-surprise.
Where the Russian-Doll Model Fails
The literal model fails completely.
Russian Matryoshka culture is centuries later.
The Voynich Manuscript is not a set of physical dolls.
Its apparent units may not form a clean hierarchy at all.
Several of its strongest classifications overlap asynchronously, which may defeat a strict tree model.
So if we force every level into neat containment, the tangent will lie to us.
The correct outcome may be that only some dimensions are nested while others are cross-cutting.
That is a useful failure.
Hostile Controls
Run the scale tests on known systems:
- ordinary natural-language texts;
- formulaic liturgies;
- known ciphers;
- computer source code;
- musical scores;
- random and pseudo-random generated sequences;
- known hierarchical manuscripts;
- known compilations with overlapping rather than nested organisation.
If every structured sequence produces similar apparent recursion, the metric is generic. If Voynich exhibits an unusual cross-scale profile, that becomes a real target for explanation.
Metaphor Removal
Delete Russian doll.
Delete nesting doll.
Delete Matryoshka.
What survives?
- explicit multi-scale analysis;
- cross-scale boundary recurrence;
- scale-stable organisational signatures;
- hierarchical boundary strength;
- contextual reveal depth;
- parent-to-child predictability;
- recursion-versus-repetition tests;
- functional stopping rules;
- prediction-with-residual-surprise.
World Return
The final imaginary doll is closed.
The manuscript returns to the table.
But we are no longer obliged to ask which single level contains the secret.
The answer may live in relations across levels.
A unit may become useful only when its parent context is known.
A page may make sense only inside a physical and visual neighbourhood.
A local pattern may be interesting precisely because a related transformation appears again at a larger scale.
And the act of repeatedly revealing related-but-not-identical structure may itself help explain why certain objects hold human attention.
Next: Tangential Voynich | Treat the Manuscript as a Conversation Piece or Heirloom — what if the object’s later social life, story and repeated rediscovery become part of what it is, even when that was not its original purpose?