[split] Does Voynichese evolve from A to B?

[Jorge_Stolfi]

You are moving the goalposts already. Many statistical jumps (at the paragraph, page and section level) have been documented in many threads...

I am not moving the goalposts, you are... 

Yes there are significant statistical differences between pages, bifolios, and sections (treating Herbal-A and Herbal B as two sections). 

But the question is about evolution of the script/language/encoding/blabber as the text was produced.  Which that paper claims is visible in the Rohonc Codex.  Its pages were almost certainly written in the order they are bound.  I understand that they detect changes in the grammar and/or the lexicon, especially at the beginnig, that are consistent with its Author perfecting the language based on his increased experience in using it.  Isn't that so ?

We still do not know the order in which the VMS sections were written, and the order of pages or bifolios within each section.  Except in a very few cases -- like, the bottom starless text of f116r was probably written after all the Starred Parags were written, and the lost parag of You are not allowed to view links. Register or Login to view. was probably its continuation.   And Pharma was probably composed (as a draft, inf not a vellum copy) before Herbal.

But we cannot even assume that page You are not allowed to view links. Register or Login to view. was the first one to be written. Or the last one.  Or that the Zodiac pages were written in the sequence that they are bound. (For one thing, the illustrations strongly suggest that Pisces was drawn after Aries, maybe also after Taurus, but before the other signs.)

One can always order the pages of a section so as to maximize the statistical similarity between successive pages, based on the assumption that words used in one page are more likely to be used also on the next page. I played around with that idea myself for a while, 20+ years ago.

However, that assumption is too weak to overcome the other causes of statistical variation.  Therefore, the resulting order is not necessarily the chronological writing orde (CWO); in fact, for some types of text -- like a herbal or materia medica -- the assumption is very likely false, and the order obtained by that method is almost guaranteed to be quite unlike the CWO.  Again, that makes as much sense as ordering a random set of books on a shelf by size, and expecting the result to be the order in which they were printed.

As for statistical differences between sections, we expect them to exist if the VMS text is a natural language in an original minimally phonetic script, unencrypted or in an encryption that is one-to-one on the lexicon.  Again, the frequency of a character or digraph is determined largely by its occurrence or absence in the few most common words; and these can change substantially between sections on different subjects, or even just different formats and styles (like full sentences vs. telegraphic, conditional vs imperative, etc.).

The statistical variations between sections and the occasional extreme variations from page to page are a problem only for theories that propose a sophisticated cipher (like Vigenère's), which should have put the character and digraph statistics through a blender; of for gibberish theories, which cannot explain why the blabber's parameters would remain constant throughout a section, but change at section boundaries and at the occasional anomalous pages (like f84v). 

For example see this thread, with an example in Q20 ... for the stars section it is even possible to point to pages dominated by vords containing 'ed' (see f111r) whereas the very next page is dominated by 'in'-vords....

Not saying that it is relevant   but the ~365 entries of the Shennong Bencao (SBJ) materia medica are traditionally grouped into 18 sections according to toxicity and origin of the product (mineral, herb, animals, etc.)  Each section is expected to have different word frequencies because of those differences in the remedies.  Not because the language or style evolvedover time.

All the best, --stolfi