Liaisons in the Voynich text

[tavie]

For n and m you're right, inside words there are too few cases to say anything. But the rule doesn't rest on them, rk and dk have plenty of chances inside words and are rare anyway. And the real test is between words, where position plays no role: after a word ending in n, the next gallows word takes an o or a y almost always; after a word ending in l, much less. If n.k were missing just because n is always final, there would be no reason for that.

I'm not suggesting that.  It wouldn't make sense for me to say that n.k is missing because n is always final. What I was cautioning about was connecting the lack of n.k with the lack of /nk/ within words, because n is almost always final.   Maybe there is a general connection between word break pairs and inside-word pairs, but we don't know.  The rigidity of so many glyphs doesn't help us here.

It was Patrick Feaster I was thinking of, when I mentioned I remembered someone previously writing about the relation between word break pairs and inside-word pairs.  Section 4 You are not allowed to view links. Register or Login to view. is relevant to your post, in particular point 4.  The whole thing is super super long but well-written. I'd also recommend going through Emma's other blog posts on Agnostic Voynich.  She has a lot of them but they're always good reads.  You might find the posts on her theory that o, y, and a are equivalents to be relevant to your analysis.

I'm a broken record on this but I'd also suggest doing the analysis separately for different sections, given that some glyphs are more common in some positions than others.  For instance, Herbal A has a lot more words beginning with k in the main text than the others do.  And also maybe consider the distorting effect of the position on the page.  I did word break calculations a few years ago for text that had the line start, line end, and top row removed. 

Some interesting things come out when you divide the text.  Despite the overwhelming y.q dominance, Stars and Balneo have small surpluses for y.l.  For Stars, this is mainly ey.l while dy prefers dy.q.   And after cutting out the daiins from Line Start, it is interesting to see in Stars that daiin.q is close to its expected value while kaiin.q fares much worse.