Even when casually looking at the pages of the VM, it’s obvious that the words aren’t generated at random: There are many repeating letter sequences, several letters or “syllables are always word-initial (“qo”, for example) or word-terminal (“dy”), some letters are always follwed by certain others etc.
Here’s my attempt to write down a “grammar” for the VM. It is valid only for the “Currier B” part of the manuscript, and will not explain all, but the most frequent of words:
The following diagram shows a grammar for “legal” (ie existing) words in the Voynich Manuscript.
As a source I’ve used the most frequent words in the recipes section (Currier hand “B”) and manually developed a grammar which would be as simple as possible while covering most of the words.
To create a word, start at the “Space” symbol and follow the arrows until you arrive at “Space” again.
The following rules apply:
Footnotes:
1) Only if not preceded by “q” in the word,
2) Only if no start group has been used,
3) Up to four of these characters
(I’m not fully statisfied with the “chke” situation, but it’s difficult to assess due to the similarity of the letters.)
Thoughts about the Voynich Manuscript 
Glyphs instead of Latin letters would illustrate the harmony observation. The writer was an artist. The artist was a scribe. For a lark they swapped jobs. If you persist with the crypto stuff anyway, a second order letter contingency table can help lead you along some rosy paths (scents, thorns and all).
I’m no linguist, so I won’t jump to conclusions over this and instead ask…. Do any other language of the period lend itself to be characterised in a similarly strict manner? Or is it possibly a byproduct of the normal complexity disappearing in an Abjad script?
Hi Björn,
To my knowledge, there are no natural languages which exhibit these features. IMHO, this leaves two options:
*) The VM is written in a constructed language, probably an “a priori” logical one, like Solresol. This would explain the word structure as an artifact of the word composition rules. It seems, though that the VM predates the first known constructed language by two centuries.
*) The VM word structure is an artefact of the enciphering process. For example, if the enciphering resulted in words in which the letters were sorted in alphabetical order (ie the plaintext word “alphabet” would be rendered “aabehlpt”), then “a” would always be word-initial, and “b” would always be either word-initial or preceded by “a”.
Unfortunately, nobody has as yet come forth with an enciphering system that would result in the complex structures we observe in the VM.
Dear Elmar,
Re: “Unfortunately, nobody has as yet come forth with an enciphering system that would result in the complex structures we observe in the VM.”
I have.
Thank you.
Don of Tallahassee