What about the labels?
According to the theory of the numerical cipher, one word from the original text corresponds to several words in Voynichese, the number of which is equal to the number of letters in the original word. Encrypting the word and putting everything together won’t work, because it will no longer be possible to decipher it.
But then why are labels present? How can they be explained?
I won’t assume that a heavily shortened word, a word without vowels, etc., is being encrypted. I can explain this by recalling the nomenclator’s logic.
For various proper names (most often words like “king,” “London,” etc.), the nomenclator used special numbers that were not part of the cipher alphabet, and they were encrypted not letter by letter, but as a whole.
There is a hint in Astrology regarding the use of this technique:
[
attachment=17089]
Of course, these could simply be the first letters of certain words, since they wouldn’t fit in full (although in that case it’s unclear whether it would even be possible to decipher the required word), but nevertheless, this example is of interest.
In general, the labels sometimes don’t differ much from other Voynichese words:
otcheodar olcheesey otolchcThy ochory dcheoldy. This may indicate that these are the same Roman numerals, just very large ones that are not part of the cipher alphabet.
I just thought—could there be letter addition in the proposed numerical cipher?
Let’s say U/V = 21. Instead of writing XXI as a single number, we can combine several letters, e.g. the letters o -
ch and a =
y. Then, when deciphering, we will get not u/v, but the combination oa. At the same time, the number will be the same (when adding).
This could explain the variety of words, which my cipher doesn’t take into account very much.
This could be verified using the You are not allowed to view links.
Register or
Login to view. example...
Let me clarify a bit about why I believe that an anagrammatic permutation, with sufficient effort, is very difficult to decipher.
Somehow, I encrypted the word “apple” as 445673. You can guess that 4 = p, but it’s unclear with the other letters. First, the number of digits is 4, and there are only 3 remaining letters, so you’ll also need to figure out where the two‑digit number is hidden. Secondly, you don’t know in what order the letters are arranged — ALE, ELA, AEL, LEA, EAL?
Thus, even if I translate 445673 as “apple,” I won’t be able to decipher this word.
While studying Roman numeral systems and ciphers, I found two features. that could greatly help with learning the VMS cipher (but I still don’t know how
). These are the general statistical features characteristic of Roman numerals as a whole.
1). The asymmetry of character distribution: Due to the constant ratio of 14:5 in Roman numerals, the density of symbols of the first rank (I, X) in the ciphertext is significantly higher than that of the fifth‑rank symbols (V, L). Well, essentially, it’s the same as the 14:5 law: in each decimal place from 1 to 9, the symbols of the unit type sum up to 14 occurrences, while those of the quinary type sum up to 5 occurrences (for example: I - 14 times, V - 5 times). This results in a constant frequency asymmetry of 14:5
Simply put: in a ciphertext, regardless of which letter is the most frequent and which number corresponds to it, the most frequent element will be I.
2). Since the cipher alphabet is based only on basic Roman numerals and does not encode any digrams such as IV/VI, the Shannon entropy
will be very low compared to the plaintext (I learned this from Gemini!):
[
attachment=17358]
Given that the encoding principle in VMS differs from simple digit encoding, there is still a match. Its entropy is quite close to 2.3!
So far, this is the most amazing thing for me.
(23-08-2026, 03:54 PM)ololololo Wrote: You are not allowed to view links. Register or Login to view.While studying Roman numeral systems and ciphers, I found two features. ... 1). The asymmetry of character distribution: Due to the constant ratio of 14:5 in Roman numerals, the density of symbols of the first rank (I, X) in the ciphertext is significantly higher than that of the fifth‑rank symbols (V, L). ... 2). Since the cipher alphabet is based only on basic Roman numerals and does not encode any digrams such as IV/VI,
The possibility that the Voynichese words could be numbers in some "Roman-like" number system You are not allowed to view links.
Register or
Login to view.. But the only similarities to Roman numerals were (1) the distribution of word lengths in the lexicon was equally compact and symmetrical, and (2) the words seemed to consist of a number of slots in a fixed order, and each slot could be empty of filled with a small set of alternatives specific for that slot.
However, Voynichese words were quite unlike Roman numerals in the order of slots and the number of different choices for each slot. Hence that "14:5" ratio would not apply.
In that page I proposed an alternative number system that could perhaps be similar to that used in the VMS. But then I realized that You are not allowed to view links.
Register or
Login to view. and many other East Asian languages, so I dropped the "numeral" idea...
All the best, --stolfi.
(23-08-2026, 04:54 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view. (23-08-2026, 03:54 PM)ololololo Wrote: You are not allowed to view links. Register or Login to view.While studying Roman numeral systems and ciphers, I found two features. ... 1). The asymmetry of character distribution: Due to the constant ratio of 14:5 in Roman numerals, the density of symbols of the first rank (I, X) in the ciphertext is significantly higher than that of the fifth‑rank symbols (V, L). ... 2). Since the cipher alphabet is based only on basic Roman numerals and does not encode any digrams such as IV/VI,
The possibility that the Voynichese words could be numbers in some "Roman-like" number system You are not allowed to view links. Register or Login to view.. But the only similarities to Roman numerals were (1) the distribution of word lengths in the lexicon was equally compact and symmetrical, and (2) the words seemed to consist of a number of slots in a fixed order, and each slot could be empty of filled with a small set of alternatives specific for that slot.
However, Voynichese words were quite unlike Roman numerals in the order of slots and the number of different choices for each slot. Hence that "14:5" ratio would not apply.
In that page I proposed an alternative number system that could perhaps be similar to that used in the VMS. But then I realized that You are not allowed to view links. Register or Login to view. and many other East Asian languages, so I dropped the "numeral" idea...
All the best, --stolfi.
I provided the general properties of Roman numerals, not a specific cipher. Naturally, rule 14:5 may or may not apply. This detail seemed quite interesting to me overall, as it essentially complicates frequency analysis.
(23-08-2026, 04:54 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.But then I realized that You are not allowed to view links. Register or Login to view. and many other East Asian languages, so I dropped the "numeral" idea...
Oh, I’ll add one more thing — couldn’t this be just a coincidence due to the fact that the VMS cipher is based on numerical logic? If you managed to create the proper number system, then it’s possible that someone else managed it too. But for some reason, I’m very skeptical about the idea of a “number system”. I think the author has no need to invent one.
(24-08-2026, 09:47 AM)ololololo Wrote: You are not allowed to view links. Register or Login to view.Oh, I’ll add one more thing — couldn’t this be just a coincidence due to the fact that the VMS cipher is based on numerical logic? If you managed to create the proper number system, then it’s possible that someone else managed it too.
Indeed, the number system I proposed in that first webpage was NOT meant to be THE coding used in the VMS. Besides the "slot with empty alternatives" structure (which resulted in the compact lexeme length distribution), it tried to vaguely imitate only the occurrence pattern of VMS gallows letters: one per word, with similar glyphs on both sides. But I did not even try to see whether that number system would match the VMS statistics and structure in any other way.
If the VMS indeed used a codebook cipher, I expected that its encoding of the numbers would be quite different from that scheme.
Quote:But for some reason, I’m very skeptical about the idea of a “number system”. I think the author has no need to invent one.
I abandoned that idea at that time because a codebook cipher would be extremely hard to write and read. It would be good for diplomatic or military letters or other short documents where secrecy was vital and justified the work; but it seemed unnecessary for a medical book, even if it was a hoax or contained "heretic" or precious secrets.
IIRC correctly, someone did find a book (a Masonic "missal"?) that was entirely in codebook cipher. But I don't recall how long it was, and I wonder how many other examples there have been.
All the best, --stolfi
(24-08-2026, 12:11 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.If the VMS indeed used a codebook cipher, I expected that its encoding of the numbers would be quite different from that scheme.
I don’t rule out a modified substitution, as it’s easier to reproduce special effects like LAAFU.
(24-08-2026, 12:11 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.IIRC correctly, someone did find a book (a Masonic "missal"?) that was entirely in codebook cipher. But I don't recall how long it was, and I wonder how many other examples there have been.
If we’re talking about the same book, then only the verbs are encrypted there.
But they’re Masons, and for them this is generally familiar…
A witch doctor has nothing to do with it.
