Voynich Glyph Structure

This started as a comment on Nick Pelling’s Cypher Mysteries blog, where he posted some ideas on EVA-s. Unfortunately, my comment became far too long to put on someone else’s blog, and it was in need of pictures to support the text, so I have changed the focus and posted on a related topic instead.

Much attention has been given to daiin (and its relatives). This is a surprisingly frequent combination of glyphs in the VMS that has generated substantial discussion and statistical analysis. Whole papers have been written on what it might represent.

I have a fairly long list of possible interpretations, some of which I’ve posted (and some that I admit I’ve kept to myself), but in this blog I’d like to discuss something more fundamental and focus attention on the shapes that underly it.

Why I Rejected Existing Transcripts

One of the reasons I created my own transcript of the VMS text is because I interpret the shapes differently from the way they have been historically recorded. As I’ve mentioned in previous blogs, I don’t use the EVA alphabet either (it would complicate the process of searching for patterns)—I developed my own.

During this process, I chose -auv instead of -ain to represent the end of daiin. Here are some examples from folio 1r but note that even this has a caveat (there is one additional possibility that I will discuss below). Note the separation between the first two strokes and the “v” shape is more distinct than the first two strokes that resemble a “u”. You can click on the image to better see the gaps and connections:

Note how the gap between the “u” and the “v” shape is more distinct than that between the two stems of the “u”. Note also in the first two examples in the second row, that the “v” shape sometimes follows the “a” shape directly, and is separated in a way similar to “v” when it follows “u”, suggesting that the “u” is a glyph in its own right and is not necessarily an “n” shape as in historic transcriptions.

So, rather than transcribing daiin, I have been recording this pattern visually as dauv/dauw, etc. You might say, “So, what? It’s just a different shape for the same thing,” but it’s not quite as simple as that. Since the bottom right example suggests that even the “u” might sometimes be a single glyph and might at other times be two glyphs (a double-i shape), it opens up possibilities such as aiv/aiw/aiiw/aiuv/aiuw/aiiv, etc., which means there may be more variation in the daiin family than is apparent when using historic transcripts. The frequency counts are potentially all wrong.

You might still argue that the daiin shapes are positionally similar and thus less likely to vary as much as I’m suggesting (or that they might mean the same thing even if they do). That might be true if the spaces are literal, but if they are not, then the potential variations could be important to the interpretation of the text.

And Then There’s the Tail…

Yes, the tail—the upswooped shape on the end of the “v”…

These days, most swooping tails are embellishments added for aesthetic reasons. In early manuscripts, however, the upswooped tail was a convention to show that letters had been dropped from the end of a word.

We still occasionally use this form of abbreviation. For example, the words “with” or “without” are sometimes written with a line over the “w” or a slash between the “w” and “out” (w/out). This is a holdover from scribal conventions that are more than a thousand years old. Similarly, in the middle ages, in Latin, English, French, German, Italian, Czech, Spanish, and other languages, this back-sweeping tail stood for whatever ending was appropriate for that language and could represent one or several missing letters.

In the VMS, it is not known whether glyphs with tails, such as “v”, EVA-r, or EVA-s, represent individual units, multiple units, or whether the motivation for the tails is to make the text look like Latin.

Faster Your Seatbelt, EVA, It’s Going to Get Bumpier

Now hold that thought about tails, because this is where it gets gnarly (as is so typical of the Voynich manuscript)…

In Latin and other European languages, the “v” shape could be a “u” or “v” with a tail or, it could be an “i” with a tail. In other words, it might be “auv-something” or “aui-something” or, in the more ambiguous example on the lower right, it might even be “aiii-something”. If this were conventional text, the reader would know by context how to expand the swoop, or whether it were simply an embellishment.

Which brings us to a further wrinkle… if medieval conventions allow that the VMS “v” with a tail could alternately be an “i” with a tail, there is another aspect of the text that needs to be addressed, one I haven’t seen anyone mention yet…

If the glyph at the end of daiin is an “i” with a tail then EVA-r needs to be re-examined, as well. In terms of glyph design based on some internal system known to the scribes, it’s possible that the “v” is an “i” with a bottom-tail and EVA-r is an “i” with a top-tail. In fact, there are times when EVA-sh is written with a tail attached to the top of the crossbar (similar to EVA-s but without the right-hand part of the character) and sometimes to the bottom. This is more apparent in some hands than others.

Superficially, almost all the VMS glyph shapes can be traced to Latin and Greek (I’m still trying to finish my blog on this, but I’m nose-deep in examples that have to be sorted and inserted), but even if they are, it’s possible the relationships between the shapes are based on certain conventions unique to the VMS.

Some closing thoughts…

It’s true that EVA-d occurs very frequently before ai, but we have to keep in mind that more than a dozen other glyphs can directly precede ai, as well, including EVA-t and EVA-k (EVA-k more than twice as often as EVA-t).

It’s also important to consider one-to-many relationships—EVA-d is sometimes written like a c combined with Latin -is abbreviation, rather than a rounded figure-8. If it’s a ligature, it may stand for two units or something else.

I have more examples of glyphs that may not fit the assumptions made in historic transcripts, but I’ll save them for future blogs.

J.K. Petersen

© Copyright 2017 J.K. Petersen, All Rights Reserved

7 thoughts on “Voynich Glyph Structure

  1. Koen Gheuens

    I’m afraid we are indeed dealing with one-to-many relationships, or even many-to-many. That is to say, a sound can be expressed individually or as part as the bench, for example. And as you say, it seems likely that some glyphs will have a different reading depending on context or even reader intuition.

    Which leads to the paradoxical situation that in order to be able to read this text, one needs to have had a lot of practice reading it.

    That’s why I keep saying that a purely text based approach won’t work. It’s just too complex, and it can’t be solved with computer logic. We’ll have to understand the imagery first and work from there… I do think the images can be understood but we’ll need a serious paradigm shift first 🙂

  2. Nick Pelling

    You should be right at home with Glen Claston’s Voyn-101 transcription, he spent a lot of time trying to see things in a broadly similar way to what you describe here.

    However, given that you haven’t read Curse yet 🙂 , what do you make of all the aiin groups on f38v?

    1. J.K. Petersen Post author

      Nick, thank you for bringing Glen Claston’s transcription to my attention. I haven’t seen it, but I will take a look as soon as I can fit it into my schedule.

      And you’re right, I haven’t read Curse (in fact I scratched my head… what curse? and then realized you must mean Curse of the Voynich). Folio 38v is definitely an interesting one. Not only are the ain repetitions almost echolalic in frequency, but there are some other interesting things going on, like the bigger gap between the ain and the prefix, the Latin “tur” character on the second-to-last line, and some of the characters I call “short-9s” drawn more distinctly here than elsewhere.

      I hope you’ll forgive me if I delay my answer until the weekend. Unfortunately, I’m currently hard-pressed for time and I would prefer to answer with visual examples.

  3. Rene Zandbergen

    Hello JKP,

    since you have transcribed a large part (if not all) of the MS yourself, you will have made the experience that transcribing consists entirely of making decisions all the time. These decisions area always subjective. They are of the nature:
    – does this small gap represent a space or not?
    – is this symbol closer to character ‘a’ or ‘b’, or is there some intermediate form that would require a new character?

    I wonder how confident you felt about the result when you were finished, because my confidence that all was correct was far from great. (The result has not been published).

    All transcription alphabets suffer from the same problem: some choices have been made about which variations constitute still the same character, and which variations deserve the use of a different character.

    I found the space problem most difficult.
    In the work of Currier and Friedman, they decided one way or the other. In the new transcription work of the old mailing list, the comma was introduced to designate an ‘uncertain space’. This reduces the issue a bit, but does not eliminate it. (It turns one possible decision into two possible decisions).

    As regards the different transcription alphabets that exist, the one question is whether the alphabet (the system) is good, and the other whether any transcriptions that have been made using it are good. Some of the most important design criteria for Eva were:
    – to be able to represent all existing transcriptions without loss of information (back and forth translation with a simple tool was possible).
    – to be able to represent all of the text in the MS (including rare characters and ligatures).

    You should have a look at GC’s “v101” as Nick already suggested. It was designed much later than Eva, but also with a specific need of GC in mind: it was very important for him which groups of strokes form one character, as he was interested in a cipher where the count of characters was critically important.

    One of your issues is easily solved using Eva. To show which of the strokes in the various shapes of daiin seem to ‘belong together’, forming one character, they should be put between parentheses. That is, da(ii)n is not the same as dai(in).
    Another problem cannot be solved, or only partly, namely the various shapes of Eva-sh.
    Here, v101 has a very large palette.

    I already once wrote about what a ‘next generation’ transcription should look like.
    It should be contained in a database, where the user can define his own transcription system (incl. alphabet). This may seem prohibitively difficult, but I don’t see it like that.
    Agreed, it is fairly complicated, and what’s more important: it may not be worth the effort.

    In the end, everyone is of course entirely free to use his own transcription system.
    It is only a pity that it is not practically possible to merge all of them.

    1. J.K. Petersen Post author

      Hello, Rene, you’ve touched on many good points.

      I tried to design my transcript with variation-interpretations in mind. For example, I created EVA-ch and EVA-sh each as two characters in such a way that it could be searched as one, searched as two, or searched along with similar-looking characters such as EVA-s (to accommodate the possibility that EVA-sh was a ligature made from existing Voynich glyphs).

      I also used slashes for alternate forms (something I borrowed from the programming world and then found out it was already being used for VMS transcripts)..

      Another thing I did with one of my VMS fonts (I designed three of them), is to put all the VMS characters in the alternate register so that I could use it in conjunction with regular Latin characters so they don’t clobber one another. One of the fonts is designed so that I can quickly switch between my transcription alphabet and Voynich characters. I wanted the whole thing to be well integrated but also flexible (I’m still not 100% satisfied, still tweaking it, but it works reasonably well).

      Another problem I wanted to overcome was that of searching lines but also being able to search portions of groups separated by spaces (or by lines/carriage returns). I have a couple of systems for this that mostly work, but I think I’ll try to brush it up so it’s more seamless.

      Pretty much everything I’m doing with respect to the VMS is a work-in-progress.

  4. Nick Pelling

    f38v is interesting for a completely different reason, one which I only noticed after staring at it for several hours while at the Beinecke a decade ago. Have a look for yourself… 🙂

    As far as transcribing goes, EVA also allowed joined characters to be denoted in capitals, e.g. Sh and Ch rather than (sh) and (ch). But that never really caught on.

    The problem of transcription, of course, is whether you have transcribed what was on the page or what you thought was on the page. And there can be a surprisingly big difference between the two…

  5. Rene Zandbergen

    Hi Nick, just for the record, this so-called capitalisation rule was already used by Jacques Guy for his Frogguy alphabet. For Eva transcriptions its only use is if one wants to represent the text using Gabriel’s font. (It really does a great job there).
    The parentheses are far more visual, and one of the options of the ‘vtt’ tool was to swap back and forth between the two notations.


Leave a Reply

Your email address will not be published. Required fields are marked *