Liaisons in the Voynich text

[quimqu]

I rechecked every number of this thread from scratch, with a new count straight from the ZL file and none of my earlier scripts. I have edited #19 and #21. The main table of my first post is right, and so are its checks (the same word, the three hands, Currier A and B, rings and labels, two positions back, the line break, the two thirds). But some sentences there said more than the numbers do, and I can't edit it anymore, so here are the corrections.

1.

it is written ytaiin, 23 times out of 23 [...] it is written taiin, bare, 21 times out of 30

Those are ytaiin against bare taiin only. Counting the other forms too: after n, r, d, s, m there are 63 otaiin, 23 ytaiin, 12 qotaiin and no bare taiin; after o, y, l, 66 otaiin, 59 qotaiin, 21 taiin and 9 ytaiin. The same goes for the kaiin, aiin, okaiin and ar figures: each one is the form with the glyph against the bare form.

2. The table works the same way: each row is the words with the glyph against the same words bare, so the y row and the o row don't add up. After n, r, d, s, m, the gallows words are 4% bare, 10% with y, 45% with o, 16% with qo and 24% with other starts (ch, sh...).

3.

When the pair is fine, he mostly doesn't.

True for y, q and d (23 to 34%), not for o. Before a gallows the o is the usual form even without a clash (70%). What the clash does there is make the bare word almost disappear and bring in the y.

4.

q is a "close" form

This one is wrong. The numbers hold (after a clash, q is at 82% when the space is 1 to 1.5 x-heights and 32% above 3), but the space in front of a q-word is ordinary. What sits far is the bare o-word after a y, the clash left unrepaired: median space 2.0 x-heights, against 1.6 for words in general. Part of that is the o itself (bare o-words after other endings are at 1.8), so what the clash adds is about a quarter of an x-height. It is not a slot kept for a q: a q is 2.1 x-heights wide, and gaps that big are as rare in front of these o-words as anywhere else (about 5%); the whole distribution just shifts a little. So in the last paragraph it is the clash left unrepaired, with a bit more room, as if at a small pause, that fits the liaison picture, not "the q that only appears when the words are written close together". The rest of that paragraph holds: y and o before a gallows are there at any distance after a clash and become more frequent with a wide space when there is no clash, and d doesn't care about the distance.

5. In Herbal A the bare a-word is rare after anything, so the d doesn't show there (97% with a clash, 95% without). That goes for my #16 too, where I said the direction holds in every section.

6. Also in #16 I gave y with a different filter than the table, which mixes things. With the table's way of counting: y before a gallows after r 63%, after d 80%, after s 81% (only 25 and 36 cases for d and s), against 23% after o, y, l. Without the line edges, y goes up a bit, to 77% / 25%; q, d and o don't move.

Everything else stands.