Voynichese and Copiale and The main advantage of Voynichese
Up until this point, I have had some difficulty describing the hypothetical numerical cipher involved in the VMS. I doubted that such a cipher had any historical precedent or would even work.
But after studying the Copiale cipher a bit in my spare time, I thought it actually resembled my assumptions. Through comparison, I will try to describe the numerical cipher to you and let you see how certain parallels with Copiale might work:
1).
...each ciphertext character stands for a particular plaintext character, but several ciphertext character may encode the same plaintext character.
I assume something similar happens with numbers. Examples from my mini‑cipher: XV = o/qo and VI = an/dy. Again, as I was saying, the Voynichese script may allow for different ways of writing the same numbers.
2).
In addition, some ciphertext characters stand for several characters or even a word.
I’m not sure about a few specific symbols, but if the numeric cipher is adapted for Latin, it could well encrypt some short prepositions or parts of words (for example, endings like -us, -rum).
Moreover, I assume that if a symbol encodes a contraction, it must encode something else as well.
3).
Seven ciphertext characters encode the single letter e.
On the contrary, one character can encode several. I already learned about this, but I still put it in a separate point.
4).
All the unaccented Roman characters encode a space.
No, there’s nothing like that in the numeric cipher. I suppose there isn’t such a wide range of symbols with a single meaning.
And now, what we have is a well‑developed theory of how the proposed numeric cipher works:
Plaintext - Numeric Alphabet - Cipher alphabet
1). The cipher alphabet is not intended for encrypting letters, but for encrypting Roman numerals. It covers a basic set of symbols (I, V, X; larger numbers like C are also possible) and some of their combinations (similar to my mini‑cipher: VI =
an/dy; V =
a, I =
n, but IV =
ok/ot). Missing combinations are recorded by using the available symbols.
I think the principle behind forming special combinations is a bit arbitrary — for example, VI =
dy, then logically IV =
yd, but such a bigram apparently doesn’t exist in VMS. In my cipher, IV =
ok/ot. Essentially, here we have the misleading letter o, which equals 15 individually, and the gallows. But we can do it more simply — write IV as
yk/yt, simply replacing
d with another symbol.
A very small set of symbols covers a wide range of meanings — letters, words, nulls, special characters, etc. Without any clues regarding the role of a specific symbol, a very complex cipher is obtained, but they can be easily introduced. For example, in my mini‑cipher,
yk/yt is nonsense. You understand that
y (equal to I individually) does not carry any meaning due to the dummy symbol
k/t, which, in essence, is a service symbol in this case.
2). The numerical alphabet consists of a set of Roman numerals (which is very surprising and unexpected
). Most likely, the author first selected the meanings for the letters and only then proceeded to expand the alphabet with new morphemes, words, etc.
This comparison shows that the hypothetical numerical cipher and Copiale are quite similar in principle. In turn, the Copiale cipher (in my opinion) is fundamentally similar to classical medieval homophonic ciphers. The principle of nulls, various signs and words is not new in itself:
[
attachment=17506]
Thus, we can say that a numeric cipher would be quite modern for the 15th century, as it would be fundamentally similar to homophonic ciphers.
...Why didn’t we solve it then?
In my opinion, the reason for the resilience of VMS lies in its two fundamental features:
1).
The cipher alphabet encrypts only numbers. Because of this, when discussing Voynichese, we are discussing numbers, but not letters. We don’t know what the numbers themselves encode or how they are distributed in the text (I mean some special rules that may not be visible to us). And we have no clear hint about the text. For example, could
daiin be a letter, a preposition, an abbreviation, or a full word?
2).
Lack of differentiation in symbols. This is a glaring feature. In the Copiale cipher, three categories of characters can be условно distinguished: Latin characters — nulls, Greek characters — vowels, abstract symbols — consonants (this is not entirely accurate, but it serves for clarity). And the turning point in deciphering was the removal of characters from one of these categories (Latin nulls) from the text. In the case of Voynichese, removing the symbols will not lead to anything, as we risk losing half of the text. If there are nulls, they may be symbols from the cipher alphabet that are not arranged according to the rules (such as
ch in my examples of encrypted texts). But even this won’t free us from the burden of deciphering the meanings of the symbols…
What can statistics tell us?
Let’s try to compare the indicators of Zipf’s law and entropy for VMS and Copiale.
The Copiale text, unprocessed for nulls, does not conform to Zipf’s law (this also happens due to homophones, but to a lesser extent). The stats just flatten out. But VMS conforms.
Copiale h1 = 5.5-6.0, Copiale h2 = 5.0. VMS h1 = 4.0, VMS h2 = 2.5 (or around two, I don't remember exactly).
This merely confirms the supposed features of the numeric cipher...
