Voynich Minim Strings: i, ii, iii, n and Repeated Strokes

Most Voynich characters dislike being doubled.

Then two families behave as if that rule does not apply to them.

The e-like strokes.

And the i-like strokes.

They can repeat.

Twice.

Three times.

Occasionally four.

For the i-like family, the repeated strokes often sit near the ends of word-like units and are usually followed by one of a small set of terminal forms.

The most famous strings are written in EVA as:

in;
iin;
iiin;
ir;
iir;
il;
im;
iim.

They look deceptively simple in a transliteration file.

One i.

Two i’s.

A final n.

Surely that means three characters?

Older Voynich alphabets often disagreed.

Currier represented EVA in with one code and EVA iin with another.

v101 likewise has synthetic codes for many of these sequences.

EVA deliberately exposes the repeated strokes.

All three are describing the same ink.

The minim problem is not how many i’s we can see. It is whether the writer intended those strokes as repeated characters, parts of one character, or a structured quantity-like family.

This is one of the places where the Voynich script most clearly refuses to tell us what level of writing we are looking at.

Quick Read

One-sentence answer: Voynich i-like minims form a highly structured near-final subsystem in which one to several repeated strokes are followed disproportionately by a small set of terminal signs—especially the form EVA calls n—and the resulting strings behave coherently enough that older alphabets treated many of them as synthetic characters, but no accepted interpretation establishes whether minim count represents phonology, morphology, abbreviation, quantity, code value or graphic composition.

Repeated i-like and e-like shapes are exceptional because most Voynich characters rarely duplicate.
i-like strings can contain one, two, three and very rarely four repeated minims.
They tend to occur near token endings.
At the end of a token, i-like strings are essentially always followed by a small set of terminal forms.
The form EVA calls n is the dominant terminator.
n can occur alone, but only a small minority of its occurrences are unpreceded by i-like strokes.
This strongly supports the possibility that n is a final contextual form related to i, though that is not proven.
In two independent transliterations analysed by René Zandbergen, iin is by far the most frequent minim string.
In the cited ZL counts, iin represents about 58.3% of the major n-terminated i-string cases, in about 24.3%, with the rest collectively smaller.
ir and iir are also substantial families and can be visually difficult to distinguish from n-ending strings in some handwriting.
l- and m-ending minim strings are comparatively rare.
Older Currier and v101 systems often treat whole minim strings synthetically, while EVA exposes the components analytically.
Strings of e-like/c-like minims have a different length-frequency distribution from i-like strings, suggesting the two repeated-stroke systems are not functionally identical.
Minim strings strongly affect alphabet size, word length, entropy and edit-distance families.
A numeral or quantity interpretation is tempting because stroke count varies, but no stable manuscript-wide numeric system has been demonstrated.

The repeated strokes are therefore evidence of structure.

They are not yet numbers, sounds or suffixes.

Why “Minim” Is the Right Cautious Word

A minim is a short vertical stroke used as a component of handwriting.

In medieval Latin scripts, several letters can be built from repeated minims.

That means a row of strokes need not correspond one-for-one with alphabetic letters.

René Zandbergen therefore uses “symbols” or “minims” cautiously when analysing Voynich i-strings.

The visible shape is certain enough.

The graphemic status is not.

Calling each stroke “the letter i” would already answer the question we are trying to investigate.

So the safe description is:

one or more repeated i-like minim strokes followed by a terminal form.

Meaning can wait.

The Strings Live Near the Ends of Tokens

Position is one of the strongest clues.

i-like strings tend to appear in late or final token positions.

This immediately suggests several families of explanation.

suffix or inflection;
word-final orthography;
abbreviation ending;
quantity or count appended to a stem;
cipher-group termination;
generated closing template.

The position does not decide among them.

It constrains them.

A theory that assigns iin a word-initial grammatical role has a distributional problem immediately.

A theory that calls it an ending at least starts in the right physical neighbourhood.

The next job is to show what changes when the number or terminator changes.

The Final n-Like Form Is Peculiar

EVA n is especially interesting because it is mostly found after one or more i-like minims.

It can stand alone.

But standalone cases are comparatively rare.

In Zandbergen’s analysis, only around two per cent of n-like occurrences stand without a preceding i-string.

That creates a powerful possibility:

n may be the word-final form of the same basic minim represented internally as i.

If true, EVA iin may visually represent something closer to a three-minim final cluster than the sequence i+i+n of three independent letters.

This would be analogous in principle—not identity—to writing systems where a character changes form at the end of a word.

The evidence is strong enough to take seriously.

It is not strong enough to call n definitively a final allograph.

Rare standalone n remains part of the problem.

iin Dominates the Family

If repeated minims simply represented free counting, we might expect a smooth decline as the strings become longer.

The actual distribution is more structured.

For n-ending i-strings, iin is the dominant combination.

in is the second major form.

iiin exists but is far rarer.

Four-minim-plus-final forms are exceptional.

Zandbergen’s two independent transliterations produce strikingly similar counts, which strengthens confidence that the broad distribution is not one transcriber’s artefact.

The shape of the distribution matters.

It suggests preferred discrete constructions, not merely arbitrary stroke multiplication.

A good theory should explain why the two-minim-plus-final construction is so successful.

Currier, EVA and v101 See Different Units

The history of transliteration reveals the uncertainty beautifully.

Currier gave synthetic codes to many minim strings.

EVA writes the visible components analytically.

v101 often returns to synthetic treatment.

For example, the visible strings represented by EVA as:

n;
in;
iin;
iiin

receive quite different single-code or multi-code representations across the systems.

This is not a failure of transcription.

It is a record of uncertainty.

Currier asks:

Does this complete visible cluster behave like one character?

EVA asks:

Which repeated visible components can we expose consistently?

The manuscript needs both questions.

The r-Ending Family Complicates the Final-Allograph Theory

Not every minim string ends in n.

ir and iir are substantial recurring families.

They are especially interesting because their visible forms can resemble the n-ending family closely.

In some manuscript locations the distinction can be difficult.

This raises at least three possibilities.

r is a truly different terminal sign.
r and n are contextual/allographic relatives.
the manuscript contains a continuum that our transliteration forces into two categories.

If r is distinct, then minim count and terminator type may encode two dimensions.

If r and n are variants, the functional system is smaller.

This is exactly where high-resolution palaeography and inter-transcriber disagreement become useful rather than inconvenient.

l- and m-Ending Strings Are Rare—and That Matters

EVA also records minim strings ending in l-like and m-like forms.

But they occur far less often than n- and r-ending families.

This is useful negative evidence.

If every final sign freely combined with one, two or three minims, we would expect a more balanced combinatorial system.

We do not see one.

Legal-looking combinations have radically different probabilities.

That suggests:

grammatical selection;
phonotactic restriction;
abbreviation convention;
code legality;
template selection.

Whatever the mechanism, minim composition is not free.

The rare endings tell us where the rule tightens.

The e-Like Strings Behave Differently

Voynich e-like strokes can also repeat.

EVA can represent sequences such as ee, eee and eeee.

It is tempting to treat e-strings and i-strings as one general “repeat a minim” process.

Their statistics argue against that simplification.

For i-like strings, the most frequent length is not the shortest possible string under the major counting conventions.

For e-like strings, frequency decreases much more monotonically as repetition length increases.

This indicates that the two repeated-stroke systems likely have different functions or different construction rules.

Shared appearance does not imply shared operation.

The bench article touched the structural relationship of ee-like forms to mantle positions.

This article’s key point is narrower: i-strings form their own specialised distribution.

Could Minim Count Encode Number?

One stroke.

Two strokes.

Three strokes.

The numeric temptation is immediate.

Perhaps the minims are tally-like.

Perhaps in means one, iin two, iiin three.

This is a testable hypothesis.

A true quantity system should show:

ordered semantic relationships among lengths;
contexts where quantity makes sense;
parallel behaviour across different stems or records;
consistent interaction with depicted countable objects or measurements.

Our retained work did not support promoting repeated units into a universal dosage code.

That remains an important failure.

Minim count may still encode quantity in a narrower system.

The manuscript-wide numeric claim has not been earned.

Could They Be Roman-Numeral-Like?

Historical scripts often reuse simple strokes numerically.

Roman numerals make repeated I meaningful.

Accounting notation can use repeated strokes.

This establishes historical possibility.

It does not establish identity.

Voynich minim strings are embedded inside word-like tokens rather than appearing only as obvious standalone numbers.

The dominant final n-like form complicates a simple tally reading.

And the highly non-uniform frequencies do not resemble unrestricted counting.

A numeral analogy is therefore a candidate mechanism, not a solved identification.

Could They Be Morphological Endings?

The near-final position makes morphology attractive.

A stem could take one of several endings:

-in;
-iin;
-iiin;
-ir;
-iir.

If those are grammatical endings, the differences should correlate with grammatical environments.

One might mark number.

Another case.

Another verb inflection.

The exact semantics are not important yet.

The prediction is.

Tokens ending iin should prefer one latent syntactic environment relative to otherwise related tokens ending in.

If no contextual difference survives after token family and Currier regime are controlled, a grammatical interpretation weakens.

Morphology turns minim count into a testable variable.

Could They Be Abbreviations?

Medieval abbreviations are another strong possibility class.

On Voynich.nu’s comparison with Cappelli, some ordinary-looking minim-plus-final forms have visual parallels in medieval Latin abbreviation traditions.

That does not give the Voynich strings the same expansions.

It shows that scribes could use compact minim-based forms to represent longer material.

If iin is one abbreviation sign rather than three characters, several Voynich anomalies change at once.

The effective alphabet grows.
Character-level entropy changes.
Visible word length shrinks at the grapheme level.
The string becomes one semantic or phonological unit rather than three.

A real abbreviation model must provide stable expansions.

The final article in this batch treats that system-level hypothesis directly.

Could They Be Cipher Groups?

A verbose cipher can use repeated simple marks to build a family of code groups.

One minim count can indicate one plaintext value.

The terminal sign can choose another dimension.

The resulting code table might contain in, iin, iiin, ir, iir and related groups.

This naturally explains family resemblance without requiring phonetic similarity.

But a cipher model needs more than a possible code table.

It should explain:

why iin dominates;
why n behaves differently from r;
why l/m endings are rare;
why the family sits near token endings;
how the groups map to meaningful plaintext.

The distribution constrains the code before any key is proposed.

Could They Be a Generation Template?

A token generator can create minim families trivially.

Choose an ending slot.

Select one, two or three minims under weighted probabilities.

Choose n most often, r less often, l or m rarely.

The observed distribution can be reproduced by design.

This makes generation compatible with the surface structure.

As always, compatibility is only the first step.

Why these exact probabilities?

Why do they change by Currier regime?

Do minim variants predict document context?

Does the generation preserve information?

A generator can imitate the shape.

The remaining manuscript decides whether that imitation is explanatory.

Minim Count Changes Entropy Dramatically

EVA iin contains two i symbols followed by n.

Because this sequence is extremely common, the transitions i→i and i→n become part of the character model.

Treat iin as one synthetic grapheme and those internal transitions vanish.

This is one reason a raw EVA character count is not ideal for some statistical questions.

Zandbergen’s Cuva analysis alphabet explicitly treats some minim strings more synthetically for this reason.

There is no guarantee that Cuva is the true alphabet either.

Its value is methodological:

change the representation and see which unusual statistics survive.

If low entropy remains unusual after plausible minim resegmentation, the phenomenon is deeper than EVA’s analytical spelling.

Minim Count Changes Word-Family Geometry

Consider two token endings:

…in
…iin

Under EVA they differ by one inserted minim.

They are edit-distance neighbours.

Under a synthetic alphabet they may contain two different atomic ending characters.

The visible family remains.

The analytical relationship changes.

This matters because local-copying hypotheses can treat one extra minim as a literal mutation step.

A morphological hypothesis can treat it as a graded suffix.

A synthetic-grapheme model treats it as substitution.

One visible difference supports several mechanistic stories.

Minim Strings Can Help Test Syntax

If in, iin and iiin are related grammatical endings, their following-token distributions should differ predictably.

Perhaps one ending prefers q-bearing next tokens.

Another closes records.

Another appears before a particular label class.

This connects the minim system to the Syntax Before Semantics article.

The endings may define latent classes before we know their grammatical names.

If transition behaviour is indistinguishable after controlling for stem family and position, a grammatical-count model becomes weaker.

Again, context can decide between visual interpretations.

Minim Strings Can Help Test Currier A and B

Currier regimes differ in many token families.

The minim subsystem should therefore be analysed by regime rather than only globally.

Questions include:

Does one regime favour iin over in more strongly?
Do r-ending forms change frequency?
Do minim-ending stems have different vocabulary associations?
Can one simple transformation map the distributions between regimes?

If A/B differences can be described partly as changes in minim-ending probabilities, the subsystem becomes a window into the larger Currier transformation problem.

If the minim distribution is stable across regimes, it may represent a deeper invariant layer.

Either outcome matters.

The Manuscript May Be Counting Strokes Without Counting Things

This is a subtle but important possibility.

Repeated minim count may distinguish forms without representing literal numerical quantity.

Many writing systems use stroke count contrastively.

One stroke forms one sign.

Two strokes another.

The relationship is graphical, not arithmetic.

Under this model, in and iin may be related because they share a construction rule while representing unrelated phonological or code values.

This is an important alternative to “one, two, three”.

Visible gradation can be structural without being numerical.

The Correct Model Should Explain Why Some Long Strings Do Not Exist

Minim systems are constrained not only by frequency but by missing combinations.

Three-or-more-minim strings are rare, and not every possible terminal combination appears.

This is crucial.

A free tally system should allow broad extension.

A grammar may cap forms at a small paradigm.

An abbreviation system may have conventional signs only for certain common sequences.

A cipher table may define only specific code groups.

A generator may weight some transitions almost to zero.

The missing long strings therefore help distinguish unrestricted count from closed paradigm.

The absences are part of the grammar.

What Survives the Minim Work

i-like minim strings are a distinctive recurring Voynich subsystem.
They are strongly biased toward token-final or near-final position.
One, two and three repeated minims occur, with four-minim cases very rare.
n is strongly associated with preceding i-like strings.
The distribution supports—but does not prove—the idea that n may be a contextual final form related to i.
iin is the dominant n-ending minim construction in major modern counts.
in is the second major form.
ir and iir are substantial alternative terminal families.
l- and m-ending minim strings are comparatively rare.
e-like repeated strings have a different repetition-length distribution from i-like strings.
Currier, EVA and v101 embody genuinely different hypotheses about whether these strings are atomic or composite.
Minim segmentation materially affects alphabet size, word length, entropy and word-family analysis.

What Does Not Survive as Established Knowledge

EVA i is proven to be a phonetic letter i.
EVA n is proven to be a separate phonetic letter n.
n is proven to be the final allograph of i.
in, iin and iiin mean one, two and three.
Minim count is a proven dosage system.
Minim strings are proven grammatical suffixes.
Currier’s synthetic characters are proven to be the correct unit level.
EVA’s analytical minim decomposition is proven to be the correct unit level.
r-, l- and m-ending strings have decoded values.

The repeated-stroke grammar survives.

The decoded meaning does not.

A Better Minim Analysis

Use “minim” or “i-like symbol” before assuming a letter.
Compare analytical and synthetic transliterations.
Separate minim count from terminal-sign identity.
Map token position.
Preserve ambiguous n/r readings.
Compare i-like and e-like repetition rather than pooling them.
Control for Currier regime and hand.
Test whether count predicts syntax, quantity or document role.
Measure entropy under alternative minim segmentations.
Require any numeric, grammatical or cipher interpretation to predict unseen minim strings.

What Would Count as a Real Minim Breakthrough?

Imagine a decipherment discovers that otherwise stable token families differ only in minim count.

One-minim, two-minim and three-minim endings correspond to one ordered grammatical feature across hundreds of examples.

The same rule predicts syntax and unseen forms.

That would turn stroke count into morphology.

Or palaeography may show that in, iin and iiin are conventional whole glyphs comparable to a compact historical abbreviation family, with each expanding consistently under a recovered plaintext.

That would make them synthetic abbreviations.

Or a cipher model may demonstrate a stable table in which minim count plus terminal type selects code values and reproduces the observed frequency hierarchy.

That would make them encoded groups.

The breakthrough is not noticing one, two and three strokes.

It is discovering why changing the number of strokes changes the behaviour of the token.

Primary School: One Stroke, Two Strokes, Three Strokes

Draw three invented symbols:

|)
||)
|||)

Ask a child what changed.

The number of strokes.

Then ask whether that must mean one, two and three.

No.

The forms could be three different letters built from the same design.

This is the basic Voynich minim problem.

Lower Secondary: One String, Three Transcriptions

Show one compound symbol to three groups.

Group A writes it as M.

Group B writes it as IIN.

Group C writes it as a two-feature code: count=2, ending=N.

Ask which is correct.

Without knowing the writing system, all three are models.

Students see why transcription is interpretation.

Upper Secondary: Test the Number Hypothesis

Suppose one stroke means one unit, two strokes two units and three strokes three units.

Ask what should happen if the hypothesis is true.

Longer stroke counts should appear where larger quantities are possible.
Ordering should be semantically consistent.
The same counting rule should work across different contexts.

Then ask what evidence would falsify it.

The student has converted visual resemblance into a predictive theory.

JC and Adult Readers: Factor the Minim Family

At a higher level, minim strings can be represented by two latent variables.

Variable 1: number of minims.

Variable 2: terminal type.

Then ask whether token behaviour factorises along those dimensions.

If minim count has a consistent effect across n, r, l and m endings, count may carry independent information.

If terminal type dominates and count does little, the apparent numerical dimension may be secondary.

If each complete string behaves atomically, a synthetic-grapheme model may fit better.

This is a clean way to turn visual construction into model comparison.

A Parent and Teacher Guide

Call the repeated marks minims before calling them letters.
Notice that they cluster near token endings.
Separate stroke count from final-sign type.
Compare Currier, EVA and synthetic representations.
Do not assume one/two/three means quantity.
Use missing combinations as evidence.
Require the winning interpretation to predict context and unseen forms.

The broader lesson is excellent:

repeated visible parts can form a structured family without revealing what dimension the family encodes.

Reader Checklist: Before You Explain iin

Are you treating EVA i as a known letter?
Is n a separate character or possible final allograph?
Which transcription system is being used?
How many minims are visible?
Which terminal form follows?
Is the string token-final or internal?
Does the same count behave similarly with r, l or m endings?
Does Currier regime change the distribution?
Could the family represent grammar rather than quantity?
Could it be abbreviation or cipher grouping?
How does a synthetic representation change entropy?
What absent strings does the theory predict should not exist?

Frequently Asked Questions

What is a Voynich minim?

It is a small i-like repeated stroke treated cautiously as a visible component rather than assumed in advance to be an independent letter.

What are in, iin and iiin?

They are EVA analytical representations of common visible minim strings with one, two or three i-like strokes followed by the n-like terminal form.

Is iin one character or three?

Unknown. Currier and v101 traditions often treat comparable visible forms synthetically, while EVA exposes their components. The functional unit has not been established.

Why is iin so important?

It is the dominant n-ending minim construction in major modern counts and therefore contributes heavily to character statistics, word families and entropy.

Could n be the final version of i?

It is a serious possibility because n occurs overwhelmingly after i-like strings and only rarely alone. It remains unproven.

Do minim strings represent numbers?

Not established. The changing stroke count makes number-like interpretations tempting, but the distribution is highly constrained and no universal quantity system has been demonstrated.

Are e-like repeated strokes the same system?

Probably not in a simple sense. Their repetition-length distribution differs materially from that of the i-like strings, suggesting different functions or construction rules.

Why do minims matter for decipherment?

Because treating minim strings as components or atomic units changes the alphabet, word length, entropy, morphology and possible cipher mechanisms.

What is the strongest current conclusion?

The i-like strings form a highly structured final-position subsystem with preferred lengths and terminators. Their functional interpretation remains open.

Research and Further Reading

The Final Idea

The minim strings are tiny.

They carry an enormous methodological burden.

One stroke becomes two.

Two becomes three.

A final stroke changes shape.

Another final form appears.

The combinations are not equally likely.

The longest ones are rare.

The family lives near the end of tokens.

This is too structured to ignore.

But structure is not yet number.

Not yet suffix.

Not yet abbreviation.

Not yet cipher.

The manuscript is clearly counting strokes. The unsolved question is what dimension of its information system those strokes are counting.

Continue Through the Voynich Research Map

This article is one specialist node in eduKateSG’s larger Voynich research library. Return to the canonical master to see the physical, visual, textual and historical evidence in one continuous argument.

Structure is evidence; structure is not translation. The master keeps resemblance, statistical structure, historical possibility and decipherment claims at separate evidentiary levels.

Next Best Routes

The Alphabet Problem — decide which unit hierarchy must be recovered before minim count can be interpreted.
Rare Glyphs — compare a frequent repeated-stroke subsystem with the script’s low-frequency edge cases.
Spaces and Word Boundaries — move from uncertainty inside visible tokens to uncertainty between them.

Deep bridge: The Model Is Part of the Message shows why the same minim string can be modelled as repeated primitives, one compound, or a count-plus-terminator system—and why each representation pays a different complexity cost.

Voynich | Everything eduKate Knows and Tested | The Minim Strings: i, ii, iii, n and the Problem of Repeated Strokes