I've noticed that these strange sentences make for a perfect constraint crib. Precisely because they usually allow for only one, or very, very few, possible interpretations. So now I’m looking for additional possible readings of these.
The background: since I’m assuming a homophonic cipher in which, above all, the vowels are each represented by multiple bridge bigrams, it’s extremely difficult to crack this cipher.
On top of that, there are likely different substitutions for consonants at positions C1 and C2—which makes it even harder.
With the help of this crip idea, I might be able to solve some ciphers, which would be, so to speak, the thread to unravel the whole tangle.
So, moving on—the next strange line:
<f8v.8,+P0> okchol k sh.<->chol.chol.chol.cthaiin.dain
k | sh.ch | o | l.ch | o | l.ch | o | l.cth | aiin
K V K V K V k V (aiin = ending)
n u/ü zz e z e zz en
nuzze zezzen
Standard German: Nüsse zu essen
English: nuts to eat
k= n (from the solution above)
sh.ch = u/ü
o= s/z/sz/zz (because these variants are used synonymously in Middle High German)
l.ch = e
l.cth ain = en (aiin still defined as an ending)
Plausible, since an original line in the "Breslauer Arzneibuch" reads:
“gip im haselnuzze zezzen. Mache im ein plaster uon haselnuzze” (Page 74)
Hochdeutsch: Gib ihm Haselnüsse zu essen. Mache ihm ein Plaster von Haselnuss.
Englisch: Give him some hazelnuts to eat. Make him a hazelnut plaster.
Of course, this is still just a proof of concept and not necessarily the actual translation.
However, if the letters found continue to align consistently with the possible solutions to the Strange Sentences, something might come of it.
And one final note: These strange sentences were, among other things, a reason for assuming that it must be a hoax. Because such repetitions do not fit with normal language. With the Vowel Bridge Model (VBM), I can clearly refute exactly that.
As you can see, the repetitive nature of the VMS can be easily explained by the VBM.
If this line is really referring to hazelnuts, the question immediately arises as to whether the plant on page 8v is a hazelnut.
The female hazelnut flower looks like this:
[
attachment=15884]
Let’s say: It's a possible match
The leaves look like this:
[
attachment=15887]
You can see the serrations in the original VMS as well—and in the small leaf of the VMS, you can also see that it tapers to a point. That fits surprisingly well.
The roots look like this:
[
attachment=15886][
attachment=15885]
Let’s say: there is a certain resemblance.
Conclusion: At the very least, this comparison isn’t entirely far-fetched. Interesting - very interesting...
And I want to emphasize that I first found the words, then posted it, and only then did it occur to me to look at the plant.
Also: there were initially several solutions (3) for the chol.chol.chol puzzle. But because of the qokedy puzzle, k=n was defined, and that narrowed the solution down to a single chain.
As a little side analysis, I was curious to see whether the opening word might confirm the hazelnut theory. To cut to the chase: the result is inconclusive, and only holds true under certain assumptions that have not been confirmed.
First of all, this strange Bank-Gallow construction maybe (!) resembles a capital H:
[
attachment=15898]
Bank-Gallow: capital H, followed by o= LSM ("deleted")| d.s | o | o.ch | s
The structure would be (H) | V1 (d.s) | C1 (o) | V2 (o.ch) | C3 (s)
with:
H = h (funny idea)
d.s = a (assumption)
o = zz / s / z (confirmed)
o.ch = e (assumption)
s = L (I don't actually think that's correct, but it may depend on the positioning. )
This would indeed result in “Hazel”; however, everything that follows is generally incompatible with “shrub / tree / nut, etc.”
Too little evidence... but that's quite a funny idea with the "H" - we'll keep it in mind.
The Theory on the Encoding of C2
<f108v.8,+P0> ysheedy.okeedy.oteedy.qokeedy.okeedy.okeedy.chedal.okar.qoteedar.oty
The section under consideration is:
oteedy.qokeedy.okeedy.okeedy.chedal
Vowel Bridge Model (VBM) vowels are:
oteedy.qokeedy = y.qo
qokeedy.okeedy = y.o
okeedy.okeedy = y.o
okeedy.chedal = y.ch
Decomposition:
y.qo | k.eed | y.o | k.eed | y.o | k.eed | y.ch
With the test values:
y.qo = e
y.o = e (for testing)
y.ch = e (for testing)
k (C1) = n
eed (C2) = n (for testing)
the result is:
e | nn | e | nn | e | nn | e
so:
ennen . nenne
Where y.ch could also be another vowel, e.g., a.
ennen. nenna.. nenna could then be something like "nenn" a "call a” etc., so there are many possibilities.
The first C2 Consonant is interesting; one would actually think it should be an n, just like the others.
But there is a peculiarity in the VMS: our C2 (i.e., slot 3 of the tokens) shows an extremely strong connection with the final glyph of the tokens, the well known “dy” effect. In the VBM, this is explained as follows:
C2 and the vowel bigram are together part of one cipher. This means the left part of the vowel bigram indicates the link, and the right part specifies which variant of C2 is used. It sounds complicated, but it’s actually quite simple.
Example: C2= eed // (hypothesis)
eed is n when y.o appears next to it as e.
eed is k when y.qo appears next to it as e.
And this also applies to other C2 values like ed, eod, etc... Two markers here each provide the information on how C2 is read. That’s all it needs to be, but it turns out to be more—though I haven’t figured that part out yet. These two patterns, however, seem to hold true.
In this respect, our chain (in the e variation) yields:
MHG = "kennen, nenne" one of many possible sentences: "die es kennen, nennen es rot.."
English: "know it, call it" one of many possible sentences: " Those who know it call it red..."
What makes this theory so unique: It explains why so few forms repeat constantly. They may do so because different consonants are essentially encoded with the same glyphs.
As beautiful as all this is, and as perfectly as it fits the structure of the VMS, it naturally makes decipherment not easier.
Overview:
We have three levels in the stream with different encodings
1. The vowel bridge. Homophonic encodings of vowels.
2. C1 Simple substitution.
3. C2 moves toward polyphonic encoding; the simple substitution of C2 is distinguished by two or more types of vowel bridges.
I know this seems to looks very complicated, but as I said, it’s really very simple.
In this cipher if I have some plaintext, say "gip im haselnuzze zezzen mache im ein plaster uon haselnuzze", is there only one way to encode this as Voynichese or can there be different Voynichese sequences that decode to the same plaintext?
(29-01-2026, 06:15 PM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.Part 1 The cryptographic pitfalls of the VMS – theories
1. Noise injection In 15th-century cryptography, it was common practice to simply repeat letters and even words in order to make decryption more difficult. I assume that most of the repetitions in the VMS represent zero words, some with individual letter changes. These have not been scattered according to a pattern, but purely at random.
Example: I live in the city = I live leve lige in the city. This is encrypted and you have a problem.
Zero words are adapted to the first glyphs to make them easier to recognise.
2. Null lines: We know that the first lines of many pages have special features. I assume that these are Null lines, some of which are even based on Latin sentences. I myself have managed to squeeze several first lines into a Latin pattern, which then tilted in the second and third lines. This is a kind of psychological warfare: an attacker ( someone who wants to solve the manuscript) in the 15th century was educated. If he found a Latin clue at the beginning, he was psychologically programmed to interpret the rest of the book in Latin. This impression is further reinforced by Latin ligatures in VMS. The author laid a false trail that makes deciphering the manuscript an endless task, because the reader tries to squeeze what is actually a different language into Latin declensions.
3. Weird Lines: As I have shown here, a striking number of the last two or three lines are ‘weird lines’. I assume that when the writer had finished a text, he simply added blank lines at the end of the text that are complete nonsense (or perhaps they are incantations after all 
.)
4. I assume that the text was written in Bavarian-Austrian slang, phonetically, as was customary at the time. Furthermore, the sentences are not even grammatically correct. In the 15th century, it would have been easy for a Bavarian to realise that his village dialect could not be understood by many other people (for example, in the larger city where he had been living for decades). A perfect disguise.
5. Then the text was shortened, terms were changed and a cipher was superimposed, which massively destroyed further information in the text. This would be about as difficult to decipher as the Native American languages with their coded terms during the Second World War:
6. So I also assume that "qo" is only a marker. It defines how the following Gallow glyph or other glyphs should be read.
7. I assume that the Gallow glyphs stand for consonant clusters: In Bavarian, there are hard consonant clusters (st, sch, cht, pf, gn, br, etc.) often at the beginning of a word. If a single Gallow character stands for an entire cluster or cluster type, then the words shrink together. There are at least 8 Gallow glyphs; with a switch, there would be 16. Realistically, with a switch, perhaps 10 to 12 of these would be used more frequently. This allows you to use most of the consonant clusters.
o Possible example:
o Labial (lips): b / p = pf / br / ps
o Dental (teeth): d / t = zt / ts (z)
o Velars (palate): g / k = ks (x) / kch
o Sibilants/fricatives: s / f / h = sch / cht / sp
8. Vowel reduction: I assume that the vowels reflect the Bavarian sound of the 15th century. "a" and "o" are phonetically almost interchangeable in Bavarian (darkening). The “e” is consistently ‘swallowed’ in Bavarian or replaced by other vowels, which eliminates the typical letter frequency. This leaves primarily "o/a" "i" and "u" as the dominant vowels. This explains why statistical analyses based on standard languages come up empty here.
9. I assume that the y at the end pulls a vowel (probably "i" or "u") backwards – both as a Latin disguise (the "9" as a ligature) and to achieve a vowel disguise. So possibly a zero glyph at the end of a line and otherwise a vowel.
10. I assume that short words were omitted on one side, if possible (articles), or were added to the other words on the other side. This is typical of Bavarian, for example:
Standard German: Hol die Wurzel
Bavarian: Hoi d'Wurzn
Phonetic/VMS style: hoid wurzn
At the same time, words in Bavarian are very often truncated at the end (endings are omitted).
So the words have mostly 4-7 Letters.
----
That's it for now, enjoy. I still need more time for part 2 – it's going to be very exciting, or not. I know part 1 won't stand up to your criticism, but as always, I don't care at all .
In general, after looking at, your approach is actually unusual and stands out from all the "German" hypotheses. I don't even have any objections... However, as I have identified, Voynichese is simpler and represents a numerical substitution (I have written a post about it). Perhaps we should try working with the Bavarian language and compare the statistics.
I also think that we should focus more on
how t
he manuscript is written rather than what language it is written in.
(06-06-2026, 06:43 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.In this cipher if I have some plaintext, say "gip im haselnuzze zezzen mache im ein plaster uon haselnuzze", is there only one way to encode this as Voynichese or can there be different Voynichese sequences that decode to the same plaintext?
That’s not so easy to answer. If you were to take just this one sentence out of context, as far as I understand it so far, there would only be one way to decode it. That said, I don’t yet know how flexible the homophone representation of the vowels is. Since there are several different vowel bridges for a single vowel, the vowel could vary. But that’s typical for a homophone cipher.
On top of that: The whole thing flows in a stream, without word boundaries or punctuation. And then it would probably also depend on what word came before it.
Example
The German sentence: “Es ist gut” (It is good) begins with a vowel. If the preceding word were “Haus,” (house) the vowel in the stream might have to refer to the last consonant. But that could also alter the entire rhythm of the sentence. I can’t say for sure at the moment how far that goes.
After examining these repetitions for cribs about further patterns of behavior, I want to build a generator that uses this model to translate German sentences into VMS, in order to gain a deeper understanding of where the “problems” with this cipher lie and to determine whether VMS also has these problems, or how VMS might solve them under certain circumstances.
At the moment,
I am unable to solve C2. C1 can be easily determined using frequency analysis; the vowel bridges related to C1 could likely also be solved—I already have some ideas—but if what I’ve written is correct, C2 should not be decipherable without further clues or similar Cribs.
Regardless of whether it is actually the VMS cipher:
The VBM shows that, simply by looking at the text from a slightly different perspective—without adding or removing anything (except for the word boundaries)—it is quite clear that the text can based on perfectly ordinary language, and that all, or at least most, of the distinctive features of the VMS fit this description—including the word repetitions.
(06-06-2026, 08:18 PM)ololololo Wrote: You are not allowed to view links. Register or Login to view.In general, after looking at, your approach is actually unusual and stands out from all the "German" hypotheses. I don't even have any objections... However, as I have identified, Voynichese is simpler and represents a numerical substitution (I have written a post about it). Perhaps we should try working with the Bavarian language and compare the statistics.
I also think that we should focus more on how the manuscript is written rather than what language it is written in.
Unfortunately, what you quoted there is already outdated... I’ve completely changed the cipher model. Unfortunately, I can’t delete the old one. But according to the forum rules, if I continue to assume it’s MHD / Bavarian, I have to stay in my thread, which is why I’ve continued posting here even though I’ve changed my approach.
I’ve read your Thread about numerical substitution and I’m not convinced yet.
Sorry...
(The new theory starts with this post: You are not allowed to view links.
Register or
Login to view.)
(06-06-2026, 08:45 PM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view. (06-06-2026, 08:18 PM)ololololo Wrote: You are not allowed to view links. Register or Login to view.In general, after looking at, your approach is actually unusual and stands out from all the "German" hypotheses. I don't even have any objections... However, as I have identified, Voynichese is simpler and represents a numerical substitution (I have written a post about it). Perhaps we should try working with the Bavarian language and compare the statistics.
I also think that we should focus more on how the manuscript is written rather than what language it is written in.
Unfortunately, what you quoted there is already outdated... I’ve completely changed the cipher model. Unfortunately, I can’t delete the old one. But according to the forum rules, if I continue to assume it’s MHD / Bavarian, I have to stay in my thread, which is why I’ve continued posting here even though I’ve changed my approach.
I’ve read your Thread about numerical substitution and I’m not convinced yet. Sorry...
(The new theory starts with this post: You are not allowed to view links. Register or Login to view.)
Yes, I'm aware of it, I've studied your new theory, and I want to say that I once thought about it in a similar way (however, I felt that the text behaves somewhat simpler). Your approach is more accurate than many, but at the moment
it's just an approach (as is mine
)...
However, you should find more correlations with the manuscript text and improve your model to get more accurate results. Even if there are similarities in the statistics between the MHD and VMS you have encrypted, you should not stop there.