The Voynich Ninja - Analysis of the text

Falsification-first methodology for Voynich semantic recovery — negative and positive

Tue, 28 Apr 2026 02:27:53 +0000

I've been working on a computational approach to the Voynich manuscript for about eight months. I'm posting the paper, the code, and all results data because I think the methodology may be useful to the community, and because the negative results are as important as the positive ones.

What I did: I built a 12,365-entry dictionary using logographic character-level assignments and morphological expansion. Then I tested whether the specific character-to-meaning assignments actually matter — whether shuffling which meaning goes on which character produces worse results. It doesn't. 500 permuted dictionaries produce equal or higher coherence scores. The dictionary is dead. That's the first result.

What I did next: I developed a four-gate protocol — pre-registered GO/NO-GO criteria at each step — to see what, if anything, survives falsification. Using morphological decomposition, distributional clustering (PPMI-SVD on the EVA corpus), and Levenshtein reverse-matching against medieval medical vocabulary, I identified five morphological roots whose families pass all four gates: distributional section-tracking, family cohesion above null, cluster enrichment for phonetically-anchored families, and pharmaceutical-section pull under 10,000-permutation testing with Bonferroni correction. Three roots survive even the most conservative correction.

What the five roots suggest: The manuscript contains pharmaceutical preparation vocabulary — roots phonetically matching medieval terms for water, heat, and plant roots — concentrated in the botanical and recipe sections, co-occurring with plant illustration markers (p = 0.018). The text is domain-segregated (z = −55.95 against shuffled null), not multi-layered. The encoding remains unbroken. I cannot read a single line.

What the paper also contains that may be of interest:

- The four-gate protocol run on Rugg-style Cardan grille text — all gates fail (Appendix D). The pipeline does not pass on procedurally generated text.
- Entropy decomposition: root-sequence H₂ = 5.17 bits vs full-token H₂ = 4.80 bits. Affix stripping recovers +0.37 bits. The entropy anomaly persists at the root level. I have not resolved this and I don't claim to.
- Anchor specificity controls: place-name vocabulary produces zero passing families, common Latin produces one, medical vocabulary produces five (Appendix B).
- An abandoned statistical gate (Gate 4 v1) documented as a methodological error and corrected before examining the redesigned result (§5.4).
- A multi-layer hypothesis test that came back negative — I tested my own theory and it failed.
- Illustration-text analysis showing pharmaceutical vocabulary near plant markers, a separate radial-diagram register in the astrological section, and structurally isolated nymph-figure labels with non-pharmaceutical vocabulary.

Code and data: You are not allowed to view links. Register or Login to view.

Pre-registered protocol: PROJECT_8_PROTOCOL.md in the repository
Draft paper Attached and in Github Repo.

The protocol is designed to be reproducible. Everything runs on the Takahashi IVTFF transcription. If anyone wants to run the pipeline on a different transcription or test it on their own hypothesis dictionary, the code is structured for that.

I'm particularly interested in feedback on two points: whether the grille control adequately addresses the generation-mechanism question, and whether anyone has ideas for further decomposing the root-level entropy anomaly by section. Both seem like productive next steps that I haven't pursued.

Happy to answer questions about any part of the methodology.

PAPER_DRAFT.pdf (Size: 106.81 KB / Downloads: 21)

The first glyph of every line – a hint at a cipher?

Sun, 19 Apr 2026 08:22:00 +0000

The first glyph of every line in the VMS – a statistical anomaly, or perhaps even part of the cipher?

This is a “by-product” of my statistical investigations in connection with the Bavarian hypothesis. I think this might interest others too, and I’m curious to see if you can verify it, or if it’s perhaps already known (I haven’t read about it yet). I’m posting it in the Text Analysis section because that’s where it belongs.

During the many analyses I’ve carried out, I’ve noticed that the first glyph in a line doesn’t match the translations. The effect is visible across the entire corpus and statistically distinguishes position 1 clearly from all others.
I already knew / suspected this regarding the first glyph on the page; most of you are probably aware of that. But the effect is more widespread.

A brief definition:

Anomaly rate: the proportion of tokens containing at least one internal bigram with a negative PMI – i.e. a pair of characters that occurs together less frequently in the manuscript than would be expected under independent distribution.
(PMI: Pointwise Mutual Information measures how surprising it is that two characters appear next to each other.).

Note

Only lines with at least three tokens are considered (to exclude labels and other elements; interestingly, the result is actually slightly better for lines with more than three letters (not token)).

1. For each token at line positions 1 to 5, the average length and the anomaly rate are calculated:

  Anfangsbuchstabe1.png (Size: 24.24 KB / Downloads: 298)

If we group all tokens from position 2 onwards (32,425 words), they have an average length of 4.49 and an anomaly rate of 28.2%. Even so, the first word of each line deviates significantly, with 4.91 and 41.5%.

  Anfangsbuchstabe2.png (Size: 20.97 KB / Downloads: 296)

On average, the first token of each line is 0.6 units longer than the token at position 2, and its internal strings are statistically significant roughly twice as often as at any other position.

2. I was then interested in the question of what happens when the first letter is removed.

The result: if one removes the first unit of each token, the anomaly rate changes to varying degrees depending on the line position:

  Anfangsbuchstabe3.png (Size: 28.87 KB / Downloads: 296)

At position 1, removing the first unit reduces the anomaly rate by 20 percentage points; at all other positions, by only 7 to 8.

The effect at position 1 is about 2.5 times as strong.

The conspicuous sequence is therefore at the beginning of the token. If one removes precisely the first unit, the remaining word behaves statistically like a normal Voynich word.

3. "The Magnificent Seven" There are mainly seven glyphs in the first position.

  Anfangsbuchstabe4.png (Size: 39.99 KB / Downloads: 296)

Total: 85.5 %

The remaining 14.5 per cent are distributed among rarer units such as ch, sh, k and others.

However, this is not the normal frequency. The overall frequency of these seven units in the corpus is significantly lower.
E.g.: p accounts for 0.84 per cent of all units in the VMS, but 7.2 per cent of all first glyphs in a line – an overrepresentation by a factor of 8.5. For s, the factor is 5.7. The distribution at position 1 therefore does not follow the general frequency of the corpus.

4. I was also interested in whether there are section-specific distributions:

Result: The seven units are unevenly distributed across the sections. Each figure indicates the proportion of lines with this initial marker that appear in the respective section:

Herbal Astro Balneo Cosmo Pharma Recipes

  Anfangsbuchstabe5.png (Size: 48.13 KB / Downloads: 297)

Each unit has its own section distribution. "o" is concentrated in Astro, "q" in Herbal and Balneo, "p" in Recipes (f103 to f116), while p is practically absent in Astro.

If the seven markers were distributed randomly across the sections, a deviation from the measured frequency would be statistically extremely unlikely. A chi-square test confirms what is already evident in the table: each marker has its own section preference.

5. Then I thought to myself, hmm, perhaps the effect is distorted by the first line.

In 72 per cent of cases, the first line of a page begins with a Gallow (t, p, k, f). That is a known fact. However, if one removes these first 227 page lines from the analysis, the key figures hardly change:

  Anfangsbuchstabe6.png (Size: 21.14 KB / Downloads: 296)

The key finding therefore stems from the normal text lines, not from the potentially ‘decorative’ page beginnings.

6. Summary

The first glyph of every line in the Voynich Manuscript exhibits statistical behaviour that differs from all other positions:
a) The first token is longer and statistically more anomalous
b) The effect is limited to the first glyph
c) 85 per cent of the lines begin with one of only seven units
d) The choice of this unit depends significantly on the section

----
Interpretation:

How this finding should be interpreted remains open. Possible directions:

- a calligraphic convention that manifests as a statistical glyph difference?

- a linguistic element (e.g. a clitic prefix) that is mandatory at the start of a line?

- what I believe: a specific hint at part of a cipher functionality. This is supported by the fact that I can already detect changes in the words following the initial letters, particularly p and t, but this is not yet statistically significant. Perhaps I’ll come back to this later.

---

In my opinion, the effect cant be dismissed as an artefact. What do you think?

Do you have any idea what this might be?
I’m interested in comments and, above all, whether this phenomenon has been described before. And, of course, in counter-evidence, hypotheses that might put it into a comprehensible context, etc...

Jojo

PS: It could at least suggest that, when attempting translations, one should try leaving out the first letter as a test.

Intentionality of the Zodiac Construction

Sat, 18 Apr 2026 19:00:40 +0000

Ladies, gentlemen, and unidentified beasts! For another thread, it became of interest for me to count the number of words in the Zodiac counter clockwise texts. The ALL CAPS signs are the sum of the 2 halves.

Folio Sign Word Count
TAURUS 159
AIRES 132
f72r3 Cancer 122
f70v2 Pisces 104
f72v3 Leo 85
f72r1 Taurus 81
f72r2 Gemini 79
f71v Taurus 78
f72v2 Virgo 77
f71r Aires 72
f72v1 Libra 72
f73r Scorpio 66
f73v Sagittarius 66
f70v1 Aires 60

Folio Sign Character* Count
TAURUS 947
AIRES 836
f72r3 Cancer 738
f70v2 Pisces 603
f72v3 Leo 539
f72r1 Taurus 493
f72v2 Virgo 491
f72r2 Gemini 461
f72v1 Libra 459
f71v Taurus 454
f71r Aires 442
f73r Scorpio 432
f73v Sagittarius 416
f70v1 Aires 394

*Includes the space as a character, counting the end of each of the three lines as a space.

I think the finding here is interesting if we're thinking about the intentionality of constructing the Zodiac pages. If we assume---and this may not be warranted on numerous grounds---that they were created with the text in mind, we would predict that the largest ones are the ones that got split. We see this! And we see the crowded ones that didn't get split. We also see one of the splits appearing at the bottom of the distribution, and the second half of each split appearing smaller.

Countervailing that, f72r1, f71v, and You are not allowed to view links. Register or Login to view. are all very typical, and f70v1 isn't wildly smaller than most of its peers. The size of the actual zodiacs seems to cluster around 440 +/- 55 characters, with three then being above that bound. This is closer to, but not exactly what I'd expect if the text were filling the space provided. To commit to saying the text were filling the space I'd have to say they made Cancer and Pisces more crowded, either by accident or to throw us off, but I also don't see signs they were trying to jam text in.

Given the small set and the fact neither word length nor character length are exactly standard measurements, and the fact that circles vary a bit, there's quickly a limit to what statistics can tell us, and I'm not terribly surprised to get a fuzzy result.

And so my question, numbers or features of the page, what do people see here? Did the words fill the circles until they were full or were the circles crafted as containers for predetermined words? For my part, the fact the spaces are reasonably consistent within a diagram but less so between them gives me the sense I'm looking at something planned to hold the text, but it's a subtle effect and my bias anyway.

About the construction of lines in the MS

Wed, 15 Apr 2026 07:25:09 +0000

In this You are not allowed to view links. Register or Login to view., when I was thinking of how to determine if the lines and paragraph have some structure, I got the idea of creating features for the gaps that make the line splits. So see the properties of the words just before the line split and just after.

The idea was to avoid looking at the page as an image and instead work only from the transcription. I turned each paragraph into a sequence of word gaps. For every gap between two consecutive words, I asked whether it was a real line break or just an interior gap.

The model I created was deliberately simple. It did not use exact words as memorized items. It used short word-type signatures and short local segment features. On the left side of a candidate break it looked at the last word and the last two-word tail. On the right side it looked at the first word and the first two-word head. It also measured whether the candidate cut would leave a line length close to the typical line length of that section. So the score had four parts: end quality, start quality, fit to line length, and a smaller central transition term.

So, making it short, a "good line ending" means that the word just before the break, and the small tail that ends the line, look like the endings that real lines usually have in that section. A "good line opening" means that the word just after the break, and the short opening sequence after it, look like real line openings. This is not a claim about magical words, it is a claim about recurrent statistical shapes.

The main validation was done only on Herbal and Biological, because they are large enough and relatively coherent. I used paragraph-level cross-validation, so train and test never shared the same paragraph. I also repeated the whole split with several random seeds. Then I added a null control: within each training paragraph I randomly permuted which gaps were marked as boundaries, trained the same model on those fake labels, and tested on the real held-out paragraphs. If the signal was an artifact of the pipeline, the null should also score well.... It did not.

These are the main validation results.

Section	Real AUC mean	Real AUC sd	Null AUC mean	Null AUC sd
Biological (balneological)	0.891	0.057	0.459	0.119
Herbal	0.905	0.036	0.461	0.131

Those values are the most important part of the finding. The real model stays very high across repeated train/test splits. The null model collapses close to chance. That does not prove everything, but it is strong evidence that the detection is not coming from a trivial leak or from learning the same paragraphs it later predicts.

The next question was whether the score was really mixing several weak signals or whether the individual components also worked on their own.... they did.

Section	End-quality AUC	Start-quality AUC	Fit-quality AUC
Biological (balneological)	0.708	0.763	0.797
Herbal	0.752	0.763	0.780

This means that real line breaks are not only associated with a better local transition. They are also associated with better line endings, better next-line openings, and better line-length fit. In other words, the break seems to happen where the text can be segmented into two pieces that both look well formed.

The direction of the effect is also consistent in both sections.

Section	Boundary end quality	Interior end quality	Boundary start quality	Interior start quality	Boundary fit quality	Interior fit quality
Biological (balneological)	-2.019	-2.564	-1.774	-2.508	-1.531	-2.531
Herbal	-1.788	-2.349	-1.798	-2.444	-1.453	-2.663

The scores are log-like and therefore negative, so the less negative values are the better ones. Real line breaks systematically look better than interior gaps on all three dimensions.

After that I tested whether the mechanism was section-specific or partly shared. First I did a cleaner transfer between the two large sections. Herbal was used to score Biological, and Biological was used to score Herbal. Then I trained on both of them together and only then applied the model descriptively to the other sections.

Training	Test section	Boundary AUC	End AUC	Start AUC	Fit AUC
Herbal	Biological (balneological)	0.804	0.667	0.690	0.753
Biological (balneological)	Herbal	0.809	0.627	0.659	0.785

That is lower than within-section validation, as expected, but still strong. So the mechanism is not just a local quirk of one section.

When Herbal and Biological are pooled and used as a source model, the score still separates real boundaries from interior gaps in the other sections as well.

Training	Test section	Boundary AUC
Herbal + Biological	Marginal stars only	0.958
Herbal + Biological	Text-only	0.859
Herbal + Biological	Pharmaceutical	0.782
Herbal + Biological	Zodiac	0.746
Herbal + Biological	Astronomical	0.739
Herbal + Biological	Cosmological	0.727

I would still call those transfer values descriptive rather than fully validated, because those sections were not cross-validated on themselves in the same way. But they are high enough to suggest that at least part of the lineation logic is shared across the manuscript.

The last part was to look at what actually defines a good ending and a good beginning. The model is not using exact words as rules. It is using recurrent word-type patterns. So the result is better read as families of endings and openings rather than individual tokens.

For Biological, line endings are enriched in short and medium forms such as ol..ly, ol..dy, or..ry, ol..ol, ok..ky, and some larger qo... tails. Line openings are enriched in families like so..dy, so..ey, so..or, so..ol, sa..ar, sa..in, ds..dy, dc..dy, tc..dy, again with some recurring qo... heads.

For Herbal, line endings are enriched in short and medium forms such as ..am, ..dy, da..an, da..am, ok..am, ot..am, ch..ry, and some repeated two-word tails involving da..in. Line openings are enriched in families like yc..or, yc..ol, yc..ey, so..in, so..or, so..ol, dc..ey, dc..dy, ds..dy, yt..dy, tc..in, and ta..ar.

A compact summary of the strongest recurrent families is this:

Section	Good line endings tend to look like	Good line openings tend to look like
Biological	ol..ly, ol..dy, or..ry, ol..ol, ok..ky, some qo... tails	so..dy, so..ey, so..or, so..ol, sa..ar, sa..in, ds..dy, dc..dy, tc..dy
Herbal	..am, ..dy, da..an, da..am, ok..am, ot..am, ch..ry, some da..in tails	yc..or, yc..ol, yc..ey, so..in, so..or, so..ol, dc..ey, dc..dy, ds..dy, yt..dy, tc..in

So the main result is not that the manuscript has special boundary words that never appear elsewhere. It is that line breaks tend to occur where the preceding segment ends in a statistically good way and the following segment begins in a statistically good way, with the added constraint that the resulting line length also fits the section.

That is why the detection values are so high. The model is not reading the same paragraphs twice. It is learning which kinds of endings, beginnings and line-length fits are typical in training paragraphs, and then finding the same structure in held-out paragraphs. The null control shows that this does not survive random relabelling. So the effect looks real.

I think the safest formulation is this: in Herbal and Biological, line breaks are statistically structured rather than arbitrary. They can be detected very well out of sample because real boundaries have better line-ending quality, better next-line-opening quality, and better line-length fit than ordinary interior gaps. The same scoring scheme also transfers surprisingly well to several other sections, which suggests that at least part of the lineation mechanism may be shared across the Voynich manuscript.

Voynich text features summary file

Thu, 02 Apr 2026 07:13:26 +0000

Hello all,

to all of us who are digging in the analysis of the MS text, I think it might be interesting to have a brief summary of the confirmed text features. That's why I have created this file: You are not allowed to view links. Register or Login to view.

It is open to be improved with your thoughts (proven facts) (in order to keep a proper view, it is not open to edit). I thik it may be a useful tool for keep on analysing the text and not forget already proven features (and not repeat too much testing).

Feel free to comment, add fetures (via reply in the thread) and join the discussion.

(By the way, my two lines in the "generative facts" page, if you think they are not confirmed or are uncertain, please tell me)

Thank you

The structure of the Voynich text and how it may be generated

Wed, 01 Apr 2026 11:16:02 +0000

Most discussions about the Voynich already agree on a few basic points: the text is not random, similar words tend to appear near each other, and position within the line matters. What is less clear is how this structure is actually generated.

I have been working in a pipeline that lets me analyze the structure of the MS. The goal of this analysis was not to describe patterns, but to test concrete generative hypotheses against the data. The question was always the same: if this were the real mechanism, would it reproduce what we observe? Running different models through the same pipeline makes it possible to discard entire classes of explanations, not just speculate about them.

Hypothesis	Expected behavior	Observed failure
Random / weak structure	No stable local similarity or positional effects	Strong clustering and positional patterns persist
Sequential (Markov-like)	Next token predictable from previous ones	Bigram/HMM models add little or collapse
Copy–modify (parent-based)	Clear local derivations, strong nearest neighbor	Generative models produce too much similarity
Single dominant parent	One best local candidate per token	Multiple candidates with similar scores, no clear winner

The important point is not just that these models fail, but how they fail. Copy-and-modify mechanisms generate too much similarity, producing tight chains of derived forms that are not observed in the real text. Sequential models fail in the opposite direction, missing most of the structure entirely. The idea of a single dominant parent breaks down because the local neighborhood is too ambiguous: for most tokens, several nearby forms are equally plausible, with no clear winner. These are structural mismatches, not minor errors, and they rule out a large class of simple generative explanations.

At the same time, some effects are very robust. Local similarity is real and strong: words share substrings and cluster in form space. Position within the line has a clear impact on length, prefixes, and suffixes. But these signals do not translate into a simple mechanism where one word determines the next. Token-level models struggle precisely because the system is not organized primarily as a chain of local decisions.

The structure becomes clearer when moving to the level of the full line. If lines are represented as whole objects, using their internal properties (number of tokens, length distributions, entropy, positional patterns), they fall into a small number of latent types. These types are not imposed manually, but learned directly from the text. They correspond broadly to different functional roles, but also reveal variation within them. These data-driven line types also show persistence across consecutive lines, suggesting that the manuscript is organized as sequences of line-level states, not just as a stream of loosely connected tokens.

The text is not governed by simple sequential rules. Token-to-token models fail to capture the structure, even when extended beyond basic Markov assumptions.
It is not generated by copy-and–modify or parent-based derivation. These mechanisms overproduce similarity and impose chains that are not present in the data.
There is no single dominant local source for most tokens. The local neighborhood is too ambiguous, with multiple equally plausible candidates.
The strongest and most stable structure appears at the level of the full line. Lines form a small number of latent types with distinct formal profiles and non-random sequencing.

A useful way to think about it is the following:

The line type defines a space of possible forms.
The local context restricts this space further by favoring forms that are compatible with nearby words.
But within that constrained space, the final choice is weakly determined. Many candidates are acceptable, and no single one is strongly preferred.

This explains why local similarity is strong but does not translate into clear parent-child relationships, and why token-level models struggle while line-level structure is much more stable.

With this analysis I try not to show that the Voynich is structured, which was already suspected, but to narrow down the class of mechanisms that can plausibly generate it. Simple sequential models and naive copy-and-modify processes do not fit. Models that operate at the level of line-level states, combined with a local compatibility field and weak selection within that field, are much more consistent with the data.

22% of 38,000 vords

Sat, 28 Mar 2026 05:28:10 +0000

I was going to do a whole study using python, but it was easy to find this statistic at You are not allowed to view links. Register or Login to view.. The voynich is abnormal with a huge anomaly of 22% of initial vowel being only "o" of the first letter of a vord out 38,000 vords. Some say the voynich is 37,000 words so the % could be a little bit higher.

What does this suggest to me along with other signs from the text. You are not allowed to view links. Register or Login to view. word before and after test show the Voynich at 100 above 20 which is random and languages at 280-300. It is low entropy. No one has been able to decode it from substitution that is reproducible. I believe that the first letter of the vord being a vowel is 22-23% of the text words. There are those that say it is meaningless, but the voynich does have structure; therefore the conlang angle. As for a language or a cipher that will present itself reproducible the same anywhere from substitution, I would find this hard to believe. You will always eventually output unintelligible syntax.

This suggests to me that the Voynich is an attempt at rules based conlang and a messy one maybe. If you have a comment against the conlang angle and believe it behave more like a known language. What evidence against this do you have to support your idea?

Here is a reference for English along with letter placements in words

You are not allowed to view links. Register or Login to view. by Position Within Word

"t" was the highest with 15.9%. and that is not a vowel. "a" was 11.68%

If you agree with me then you might also feel like I do that the meaning was lost with the people involved. So we may only get subjective feelings from Voynich on its meaning from imagery. The text may elude us.

What is the current status of the transcription?

Wed, 25 Mar 2026 21:02:34 +0000

I’m having trouble with the Bavarian cipher—I’m finding interesting and coherent words that, in terms of how frequently they appear, also match other Middle High German recipe books; actually, a lot of it fits together. But there seem to be certain letter combinations that sometimes work and sometimes don’t.

And that’s when I noticed this example: 9v, line 5.

9v7.png (Size: 176.97 KB / Downloads: 179)

I’m concerned with the “cc” and “ee.” If I’m seeing this correctly, I could read the first “cc” as a “tc,” while I could read the second “cc” as a “cz,” and the “ee” as a “ce”... or maybe none of that is correct, but if it is, that would be discouraging—because we have the “a/o” problem, problems with the “Gallows”... R problems... and that would completely turn all the statistical and other analyses upside down.

And so I’m asking myself:

What is the most current transcription, or the most accurate one?

Does it even make sense to ask that?

Is Eva a dead end that thousands of people have walked into, when Stolfi is already doubting it (if I’ve understood correctly)?

I’m a bit at a loss right now...

-------------
In this context, I should also mention this writing exercise: You are not allowed to view links. Register or Login to view.. 0.6.5.jsp?folder_id=0&dvs=1774358703990~945&pid=19073483&locale=pl&usePid1=true&usePid2=true

Screenshot 2026-03-25 220649.png (Size: 50.26 KB / Downloads: 156)

Milano is the key

Sun, 22 Mar 2026 04:37:16 +0000

In the next 24 hours we will see !
If my Theorie is right !
I found the key of the Voynich Kodex !
Stay tuned

This is impossible for a language

Thu, 19 Mar 2026 04:19:00 +0000

No known language produces a 140 to 167 prefixes for its most frequent word. The word would not make sense. If you disagree show the list of words with the language and the prefix in front of that word and it probably does not have to be the most frequent word.

167 Unique values found

(863) [b]daiin[/b]
(60) o[b]daiin[/b]
(44) cho[b]daiin[/b]
(42) qo[b]daiin[/b]
(32) che[b]daiin[/b]
(23) sho[b]daiin[/b]
(21) y[b]daiin[/b]
(16) ch[b]daiin[/b]
(15) she[b]daiin[/b]
(11) cheo[b]daiin[/b]
(9) ol[b]daiin[/b]
(9) to[b]daiin[/b]
(8) oteo[b]daiin[/b]
(5) sheo[b]daiin[/b]
(5) pcho[b]daiin[/b]
(5) po[b]daiin[/b]
(5) oto[b]daiin[/b]
(4) chee[b]daiin[/b]
(4) pche[b]daiin[/b]
(4) opche[b]daiin[/b]
(3) ko[b]daiin[/b]
(3) opy[b]daiin[/b]
(3) ro[b]daiin[/b]
(3) so[b]daiin[/b]
(3) sh[b]daiin[/b]
(3) lo[b]daiin[/b]
(3) pch[b]daiin[/b]
(3) yto[b]daiin[/b]
(3) dal[b]daiin[/b]
(3) te[b]daiin[/b]
(3) tcho[b]daiin[/b]
(3) al[b]daiin[/b]
(3) p[b]daiin[/b]
(3) l[b]daiin[/b]
(3) qote[b]daiin[/b]
(3) ote[b]daiin[/b]
(3) op[b]daiin[/b]
(3) [b]daiin[/b]dy
(3) qoke[b]daiin[/b]
(3) qoe[b]daiin[/b]
(3) oke[b]daiin[/b]
(2) py[b]daiin[/b]
(2) fcho[b]daiin[/b]
(2) dcho[b]daiin[/b]
(2) dol[b]daiin[/b]
(2) oko[b]daiin[/b]
(2) kcheo[b]daiin[/b]
(2) tch[b]daiin[/b]
(2) olch[b]daiin[/b]
(2) olke[b]daiin[/b]
(2) or[b]daiin[/b]
(2) keo[b]daiin[/b]
(2) okeo[b]daiin[/b]
(2) teeo[b]daiin[/b]
(2) okeeo[b]daiin[/b]
(2) teo[b]daiin[/b]
(2) psho[b]daiin[/b]
(2) qoee[b]daiin[/b]
(2) yche[b]daiin[/b]
(2) oee[b]daiin[/b]
(1) cha[b]daiin[/b]
(1) [b]daiin[/b]y
(1) oka[b]daiin[/b]
(1) yo[b]daiin[/b]
(1) qocho[b]daiin[/b]
(1) eto[b]daiin[/b]
(1) shfy[b]daiin[/b]
(1) fol[b]daiin[/b]
(1) chol[b]daiin[/b]
(1) t[b]daiin[/b]
(1) [b]daiin[/b]ol
(1) ctho[b]daiin[/b]
(1) chech[b]daiin[/b]
(1) kal[b]daiin[/b]
(1) ypch[b]daiin[/b]
(1) tolch[b]daiin[/b]
(1) do[b]daiin[/b]
(1) taro[b]daiin[/b]
(1) shep[b]daiin[/b]
(1) psh[b]daiin[/b]
(1) ksho[b]daiin[/b]
(1) chokcho[b]daiin[/b]
(1) dcheo[b]daiin[/b]
(1) ycho[b]daiin[/b]
(1) cpheo[b]daiin[/b]
(1) seo[b]daiin[/b]
(1) qofy[b]daiin[/b]
(1) qoto[b]daiin[/b]
(1) ytcho[b]daiin[/b]
(1) yctho[b]daiin[/b]
(1) [b]daiin[/b]o
(1) oe[b]daiin[/b]
(1) yteo[b]daiin[/b]
(1) opal[b]daiin[/b]
(1) sholfos[b]daiin[/b]
(1) yke[b]daiin[/b]
(1) tae[b]daiin[/b]
(1) oteeo[b]daiin[/b]
(1) oeeeo[b]daiin[/b]
(1) chl[b]daiin[/b]
(1) odee[b]daiin[/b]
(1) daror[b]daiin[/b]
(1) eeo[b]daiin[/b]
(1) o?[b]daiin[/b]
(1) qopy[b]daiin[/b]
(1) pol[b]daiin[/b]
(1) ycheeyty[b]daiin[/b]
(1) [b]daiin[/b]ls
(1) sal[b]daiin[/b]
(1) keeo[b]daiin[/b]
(1) ararcho[b]daiin[/b]
(1) chto[b]daiin[/b]
(1) opo[b]daiin[/b]
(1) ocheo[b]daiin[/b]
(1) cho[b]daiin[/b]dy
(1) qofch[b]daiin[/b]
(1) kche[b]daiin[/b]
(1) ???[b]daiin[/b]
(1) shol[b]daiin[/b]
(1) doleo[b]daiin[/b]
(1) fo[b]daiin[/b]
(1) se[b]daiin[/b]
(1) lkeo[b]daiin[/b]
(1) qokcheo[b]daiin[/b]
(1) qoteo[b]daiin[/b]
(1) eeeo[b]daiin[/b]
(1) cheot[b]daiin[/b]
(1) qop[b]daiin[/b]
(1) sotch[b]daiin[/b]
(1) yshe[b]daiin[/b]
(1) tsho[b]daiin[/b]
(1) qe[b]daiin[/b]
(1) cheolch[b]daiin[/b]
(1) ypo[b]daiin[/b]
(1) pshe[b]daiin[/b]
(1) lee[b]daiin[/b]
(1) lote[b]daiin[/b]
(1) cheee[b]daiin[/b]
(1) okee[b]daiin[/b]
(1) ykee[b]daiin[/b]
(1) qotee[b]daiin[/b]
(1) qoch[b]daiin[/b]
(1) qopche[b]daiin[/b]
(1) sheeo[b]daiin[/b]
(1) kch[b]daiin[/b]
(1) lsh[b]daiin[/b]
(1) pcheo[b]daiin[/b]
(1) cheodoii[b]daiin[/b]
(1) otchee[b]daiin[/b]
(1) okche[b]daiin[/b]
(1) qokee[b]daiin[/b]
(1) chdo[b]daiin[/b]
(1) qokeeo[b]daiin[/b]
(1) ofche[b]daiin[/b]
(1) chdyp[b]daiin[/b]
(1) tche[b]daiin[/b]
(1) qoko[b]daiin[/b]
(1) ycheeo[b]daiin[/b]
(1) tcheo[b]daiin[/b]
(1) ykeo[b]daiin[/b]
(1) cheoltche[b]daiin[/b]
(1) qeoo[b]daiin[/b]
(1) olche[b]daiin[/b]
(1) oshso[b]daiin[/b]
(1) otcheo[b]daiin[/b]
(1) dche[b]daiin[/b]
(1) ychee[b]daiin[/b]

Voynich Manuscript-transcription

Wed, 18 Mar 2026 02:21:46 +0000

For researchers looking to copy-paste You are not allowed to view links. Register or Login to view. voynich.

You are not allowed to view links. Register or Login to view.

Order of the pages

Tue, 17 Mar 2026 18:40:17 +0000

Would you be surprised if I told you that the order of the pages of the Voynich Manuscript does not matter. Each page forms its own entity, independent of the other pages.

About the generation of similar words

Sun, 15 Mar 2026 19:17:09 +0000

This weekend I have been exploring the internal structure of word variants in the Voynich manuscript with a fairly simple approach. The goal was to try to answer the question of the generation of words from the same family. When very similar words appear next to each other, do they generate in successive chains of transformation, or do they form radial families around a base form?

I think this distinction is important because some models of generation proposed for the Voynich (such as Torsten Timm's model) assume that a word is often generated by modifying the immediately preceding word, that is, a successive or sequential generation.

Below I briefly explain the method used and the main results.

Basic idea: When reading the Voynich text it is common to find groups of very similar words within a short space. For example: qokedy, qokeedy, qokeey, qokeeydy.

This type of sequence suggests that they could be variants of the same base form. To study this systematically, the text is simply treated as a linear sequence of words. For each word:

- It is considered a potential core.
- The next 40 or so words are examined.
- If any word has a small edit distance (Levenshtein ≤ 2), it is added to the same group. I call these local groups of variants "bursts".

Reconstructing relationships: Once a burst has been identified, the next step is to look at how the variants relate to each other. For each new variant, we look at whether it is more similar to the core or to a previous variant within the same burst. From this, we can reconstruct a small dependency structure. Three main patterns emerge.

- Radial (star) structure:

variant
|
variant — core — variant
|
variant

Most variants are derived directly from the core.

- Linear (chain) structure

core → variant → variant → variant

Each form is derived from the previous variant.

- Mixed structure: a combination of the two.

Control texts: To find out what is normal in real texts, the same procedure was applied to three control texts: De Docta Ignorantia (Latin), Alchemical herbal (Latin), Culpepper's herbal (English). In these texts the bursts tend to be: small, mostly radial, with very few linear chains. Typical values:

- star structures: ~0.68–0.80
- linear chains: ~0.003–0.007
- average burst size: ~2–3 variants

Results in the Voynich: Applying exactly the same procedure to the Voynich manuscript, we obtain approximately these global values:

- non-isolated bursts: ~4512
- average burst size: ~3.45
- average depth: ~1.56

Structural distribution:

- star: ~0.63
- mixed: ~0.36
- chain: ~0.01

This means that bursts are larger than in the control texts, linear chains are still very rare and there are more mixed structures. Mixed structures usually have this form:

core → variant
core → variant
variant → variant

That is, many variants derive directly from the core, but some also derive from other variants.

To test whether this structure depended only on the Voynich vocabulary or also on the actual order of the text, I repeated exactly the same analysis on a permuted version of the manuscript, keeping the same words within each page but shuffling their order. The result is that the bursts do not disappear, but they do become slightly smaller, shallower, and more radial. In the actual text there is a higher proportion of mixed structures and slightly more internal links between variants. This suggests that the Voynich vocabulary already favors families of similar forms, but that the actual order of the text adds additional local organization.

The analysis also shows that the phenomenon is not homogeneous throughout the manuscript. The astronomical and zodiacal sections tend to have more radial bursts, while the biological-balneological section, and partly also Herbal and Text-only, show more mixed and deeper structures. This suggests that the local behavior of variant families changes according to the section.

Interpretation: The results suggest that local families of variants in the Voynich are not primarily organized as long chains of sequential transformation, that is, the dominant pattern does not appear to be simply a word repeatedly modified step by step. Instead, they are more like local fields of related forms, in which several variants form around a base form, and sometimes some of these variants also give rise to further modifications.

Schematically, the pattern would look more like this:

core
├ variant
├ variant
│ └ variant
└ variant

than a simple chain of the type:

core → variant → variant → variant

What mechanism might generate this? One possible interpretation is that the scribe was working with local families of forms rather than a simple linear succession of modified copies. In this scenario, a word appears and, within the same immediate context, variants are generated by:

- small orthographic modifications
- reuse of similar forms
- occasional modifications of other recent variants

Such a mechanism would naturally produce:

- local groupings of similar words
- larger bursts than in normal control texts
- partially interwoven, not strictly linear structures

Furthermore, the permuted text control suggests that this pattern does not depend solely on the manuscript's word repertoire. The Voynich vocabulary already favors families of similar forms, but the actual order of the text seems to add a somewhat deeper and more mixed local organization.

Page One Voynichese "Signatures"

Fri, 13 Mar 2026 14:06:20 +0000

Something about You are not allowed to view links. Register or Login to view. has recently started bothering me. It looks like four paragraphs. Each seems to end with what to me looks like a signature. They probably aren't signatures but I don't know what else to call them.

I'm not talking about the disputed mystery erased "Tepenec" at the bottom. I'm talking about the Voynichese text.

1:
ydaraishy<$>
ydaraishy
ydaraishy
ydaraiCishy ( ydaraiCishy )

2:
dain.os.teody<$>
dain os teody
dain os teody
dain or teody ( dain or teody )

3:
otol.daiiin<$>
otol daiiin
otol daiiin

4:
dchaiin<$>
dchaiin
dchaiin

What are these words? Could they be the names of four authors -- each giving his own introduction to the work?

Only the first one, ydaraishy ( ydaraishy ), seem to be unique within the manuscript. daiin.net has it as ydaraiCishy ( ydaraiCishy ).

daiin.net has the second one as dain or teody ( dain or teody ).

daiin.net has the third one the same as voynichese.com, otol daiiin ( otol daiiin ). I did note that the first word, otol ( otol ), is the same word as the label for Elemental Fire / Yellow bile on You are not allowed to view links. Register or Login to view. .

How fast could a scribe write a Voynich like text?

Fri, 13 Mar 2026 11:57:26 +0000

Hello,

a certain question that me and one of my friends just had was: what would be the expected time for a scribe to write a Voynich-like text? Specifically, I'm wondering because while we have some inkling (hah!) about rates of work of professional scribes in the medieval times (it seems to be 4-8 folios a week depending on the source), this would be a faithful copy of a meaningful text!

So, perhaps producing text would take a different amount of time depending on how the Voynich was created. Here are some interesting things that could impact it, broken down by theory:

A faithful copy of an earlier draft. Presumably, we could take the standard rate of text production as a good rule of thumb, though the Voynich alphabet seems easier to copy than gothic script, for instance.
A sloppy copy of an earlier draft. This would probably take less time, but presumably one would expect more errors being corrected - though perhaps the scribe(s) didn't care. If that is the case, what could be the rate of production that would satisfy whatever patron asked the scribe for the copy?
The first draft of a meaningful text: where are the corrections? Composing meaningful text takes time, too, so it could take longer than copying.
Intuitive gibberish, as in the Gaskell & Bowern paper 'Gibberish after all.' Presumably, since one neither has to copy anything nor worry about exactness, but simply put quill to paper and maybe reference other nearby words while writing, this would be the fastest method. With that stipulation, what's the fastest one could do?

The reason I'm wondering is we're putting a lot of stock into this being an extreme amount of effort; but maybe, if the manuscript is meaningless either due to being a sloppy copy or an outright gibberish hoax, the production period would be relatively brief. This has some implications for whether it would be worth writing it as a hoax or for sale.

Cheers