Currier A and B

[Anton]

I think one problem with saying whether a label or a word occurs only in Currier A or B, is that you may not be capturing the real difference.

Well that's no problem if we have a prior agreement upon which folios are A and which are B (those folios which are uncertain may be left off).

Yet the Currier B "feature" is not either of those words but rather the bigram [ed], which is over 125 times more common in B than A.

It is better to start with these two character combinations as they are more fundamental to whether a word appears in one language or another. They are also, hopefully, more diagnostic of what the difference between A and B will turn out to be.

I think that's precisely what's the fundamental problem with the A/B definition, unless those differing n-grams (for n=2 or whatever is essential for A/B) are elementary units of Voynichese. In the latter case we could suppose that one set of n-grams is used in A exclusively, another set of n-grams is used for B exclusively, and some n-grams are used in both A and B. But even that is probably not the real case, because, AFAIK, the stats are not about complete absence of certain n-grams in A or B, just about they are to a certain extent more or less frequent.

The reasoning is as follows:

a) We have labels in the VMS;
b) A label should be an expression in itself (unless a really sophisticated encryption mechanism is in place) - in other words, it represents a certain word or a phrase, in whatever way;
c) if a label (hence - a given word or phrase) is met both in A and in B, then no matter what are the n-grams stats, I don't see what's the real linguistic difference in the underlay.

If language A/B border is not the same as the folio border, the situation might have been considered in another light, but, AFAIK, that's not the case either.