A key to understand the VMS

[Sam G]

In natural languages a word normally (cf. poems) is used because of its meaning and not because it is similar to a previously written one. The result that the words are arranged such that they co-occur with similar ones is therefore not compatible with a linguistic system.

Well I still don't see why not, even in poetry, and you ignored the (basically more important) suggestions that it could be due to reduplication or agreement.

Moreover similarly spelled words occur frequently one above the other.

This is not uncommon in meaningful texts either.  And obviously we can expect more of it by pure chance when the words are more similar in general.

To change words which have already been written requires no additional tools.  It is reasonable to assume that with some training it is very efficient to copy a text using the text itself as a source. For doing so it is not necessary to switch between an external source and the text being written. For someone in the early 15th century it would therefore have been possible to use such a method to generate the text of the Voynich Manuscript while writing.

Well, earlier in the thread I tried making some points about how the relative frequencies of the words don't look like they were produced by your proposed auto-copying method (e.g. my first post in this thread), and you said you didn't want to talk about it because you hadn't mentioned it in this thread.  Instead you wanted to talk about the properties of the word network only.  But now that I've shown that the word network is not really that different from monosyllabic Mandarin, it seems like we're back to the auto-copying idea?

Alright then, here's my original question again:

For the most frequently used word daiin it is possible to build a tree of similar words starting with o- or qo-. The frequencies for this words are very characteristic:

Level 1:                                          daiin (863 times)
Level 2:          okaiin (212 times)         |          otaiin (154)        |  odaiin (60)
Level 3: qokaiin (262 times), kaiin (65) | qotaiin ( 79), taiin (42) | qodaiin (42)
There are more words starting with ok- or ot- then with od- and the word starting with qok- is even more frequent then the word starting with ok.

So qokaiin is more common than okaiin, but otaiin is more common than qotaiin.  Why should this be true, from the standpoint of your auto-copying hypothesis?  If the scribe is just randomly adding a q in some cases, shouldn't the ratios be roughly equal?