Two of the least commented-on aspects of the Voynich Manuscript’s “Voynichese” alphabet are (a) its symmetry and (b) its partitioning into quite well-known (but distinct) usage groups. For example:
* the four gallows characters, where EVA t and EVA k are almost always interchangeable, while the single-leg shapes for EVA p and EVA f closely mirror the double-leg shapes for EVA t and EVA k. (And let’s leave the strikethrough gallows aside for the moment.)
* the EVA aiin family of letter groups, which all operate in a very specific way: there are no contexts where ain appears that you wouldn’t also see aiin or even aiiin.
* the ar / or / al / ol group, whose members seem to appear within words in much the same way as each other. The air and aiir letter groups might also be related to this set, though this isn’t
not 100% clear. Similarly, -am often seems to me (with a hat tip to Emma May Smith, who discussed -m recently) to be something closer to a combination of ar and hyphen, i.e. that -am at the end of a line often resembles the end of the first half of a word broken in half by the line-ending (and where the second half of the word is at the start of the next line, but disguised with an extra letter inserted before it).
* the -dy and -y word endings, which both seem to be cut from almost exactly the same cloth.
* the e / ee / eee / ch / sh / eo group, which seem to me to function slightly differently between A and B pages.
* the qo group, which almost universally seems to operate as a prefix. In those places where we get l- words, we also get qol a lot: and where l- words don’t appear, we get almost no instances of qol.
Cross all the above instances out, and what remains is a very sharply reduced set of usage groups, such as d- words (in particular daiin, which seems to operate in a mysterious world all of its own), o- words (particularly in front of gallows), and y- words.
What about EVA s?
But if you do do this kind of crossing out, you also won’t find a comfortable place for EVA ‘s’ to go. In fact, to my eyes EVA ‘s’ appears to be the single most anomalous character in the Voynichese alphabet: there’s a strong case to be made that it is the most ‘exposed’ single glyph of all of them, and – by that same token – the one we should spend most time on trying to understand. What I’m saying is that EVA s might well be the weakest link in the Voynichese chain.
If you remember to put aside all the completely different ‘sh’ characters (sharing ‘s’ for both of these glyphs was, in my opinion, a foolish mistake in the design of the EVA transcription scheme, *sigh*), you find that ‘s’ occurs about 1.71% of the time in A pages, and about 1.00% of the time in B pages. If you remove any ‘as’ or ‘os’ pairs (as being probably miscopied or mistranscribed ‘ar’ / ‘or’ pairs) from these stats, these figures go down to 1.34% and 0.83% respectively.
And yet some A pages have numerous s characters (e.g f14r, f15r, f24r), while others have one or fewer s characters (e.g. f14v, f18r, f19r): that this single statistic can differ so much between the two sides of the same folio is something that hasn’t really been noted before, as far as I can recall. [Unless any Lorites out there care to show me the precedent I’ve missed: in one of Friedman’s groups, no doubt.]
All of which incidentally reminds me of something that Glen Claston told me he noticed when he was making his transcription (but which I now can’t find in my email archive, *sigh*): that Voynichese had different clusterings of letter usages that would seem to go into and out of fashion (almost as if one kind of ‘mode’ was active now, and then a different mode active later), sometimes by paragraph, sometimes by page. If this is correct, then perhaps ‘s’ is an active part of some ‘modes’ but not others – just an idea.
What about saiin vs daiin?
I find it interesting that sdaiin occurs only once (on f66r), while sdain, sdaiiin, dsain, dsaiin, and dsaiiin don’t occur at all: yet saiin occurs 144 times.
If s- is some kind of prefix token here, then it seems that so too is d-, and in a way that makes the two avoid stepping on each other’s toes.
My suspicion (for what it’s worth) is that while both work as prefix tokens, they in fact code for two quite different classes of mechanisms: and, moreover, that both prefixes are more meta-linguistic than linguistic in any useful sense.
And what about the first column?
EVA s also has a strong tendency to appear as the first letter of a (non-paragraph-starting) line, particularly in Balneo B pages – but this may possibly be because Balneo B tends to have longer paragraphs than elsewhere.
Combine this (a) with the well-known observation that the first word on each line tends to be slightly longer on average than all the other words on a line, and (b) with Philip Neal’s suggestion that the first letters down some Voynich Manuscript pages might well be a vertical ‘key’ or something similar, and you get an interesting possibility to consider: that line-initial ‘s’ may specifically operate as a null that the writing system needs to prepend to certain (typically short) words.
I was thinking about this today, triggered by a Voynich Ninja forum discussion: I wondered if it might be possible to construct a statistical experiment to test my suggestion that line-initial s- might function as a null character that gets prepended to certain short words (such as aiin).
According to the tentative model I have in mind, the (aiin : daiin) ratio for non-line-initial words should be roughly the same as the (saiin : daiin) ratio for line-initial words. And perhaps it would be good to then repeat broadly the same test for non-line-initial (ar : dar) vs line-initial (sar : dar), etc.
However, I don’t have the right counting tools to do this easily: can anyone please run this test? Thanks!