A key to understand the VMS

[Torsten]

Torsten: the network of Voynichese words exists after the fact, not before it - its existence is not enough to prove anything useful by itself. No 15th century person could remotely have sat down and said "I'm going to create a dense network of words that follows a cleverly-thought-out set of transition rules".

You misunderstand what I say. In my paper I wrote for instance that the network is an "unintended side effect of the manufacturing or encoding process" [You are not allowed to view links. Register or Login to view.].

And your suggestion that the network probably came from a smaller (but now lost) network of words built up as "some drafts" by the same scribe is nothing short of saying that your chicken came from a marginally smaller chicken (and not via an egg). 

There is simply no need for a smaller network. All you need at the start is the script. The copying process results in the network and the observed strict word structure. All you have to do is to invent the first line of text for the first page. For my app I use a single line of the VMS with at least 8 words as seed to generate texts with 10 000 words with statistics similar to that of the VMS.

See for instance all the similar word within the first paragraph of page f1r:
   

I simply didn't think that the scribe has invented all the glyphs used in the script while writing the VMS. Therefore I find it only logical that the first drafts are not part of the VMS.

As far as eol goes, you don't seem to understand my point at all: you describe some transitions as being somehow preferential (e.g. air to ail) whereas others are much weaker (e.g. al to ail), but given that both have the same abstract edit distance yet greatly different statistics, to claim that it must be the network itself that drives these preferences is a pure post hoc argument.

I never said something like "it must be the network itself that drives these preferences". What I say is that the network of similar words is an unintended side effect of the text generation mechanism. 

What is actually going on under Voynichese's hood is that a highly varied set of adjacency and writing rules - some combinative (e.g. al/ol/ar/or), some generative (e.g. av, aiv, aiiv, aiiiv), some fixed (e.g. qo, dy), some outright mysterious (e.g. Neal Keys), plus others we don't even have names for yet - is in play. These rules all work together to implicitly create the network you've worked so hard to document, in ways that we only glimpse via statistics. 

Even if you call similar words combinative and generative they are still similar. Even if you call repeated gallows Neal Keys they are just repeated gallows. You are right by saying that they create the network of similar words. But in my eyes this was not done implicitly.

And so what I find most frustrating about your papers is that the study you have done on the statistics would be hugely informative and useful, were you not investing so much effort into linking them to a network-centric autocopying view of the manuscript which doesn't work as well as you think, and which - even if it were true, which it isn't - would only be sufficient to explain 10% of Voynichese's behaviours, and then only after the fact.

Statistics and observations are on itself informative and useful. We both talk about the same manuscript no matter what you or I think about it. I simply can't change the fact that the words 'daiin' and 'dain' exists seven and six times on page You are not allowed to view links. Register or Login to view. and that the words 'chol' and 'shol' exist five and four times on page f1v. In the same way your  statement "I don’t believe that 'oiin' would ever be correct Voynichese, i.e. it’s just a miscopied 'aiin'" [You are not allowed to view links. Register or Login to view.] doesn't change the existence of the words 'qokoiiin', 'qotoiin', 'otoiin', 'ytoiin', 'ykoiin', 'choiin', 'soiin' and 'soiiin' on page f37v.  

If you are interested in the statistics and observations as you say why you refuse to discuss them?

Only to deny is fruitless. It would be ok if you would say "I’m not convinced" like You are not allowed to view links. Register or Login to view., but you can't argue that the autocopy hypotheses must be wrong without giving any reason for it. Also you need some examples  of not explained features if you want to say that the autocopy hypotheses would only be sufficient to explain 10% of Voynichese's behaviours.