The Voynich Ninja

Let's strike out on an entirely new (for me) aspect of Voynichese - the possibility that it might have a phonology. I've dallied with this on and off for years, but never really had the energy to go down this route. But recently I've had some thoughts that are encouraging me.

Let us reduce the problem to its most reductive position. I postulate the following:

• Voynichese has a phonology (ie it can be spoken).
  

Why not? In the 15th century, many written texts were designed to be spoken aloud, it was a fundamental part of the training of scribes. And if Voynichese does link back to a language, even if the script is unique, the pronunciation is likely to carry through.

• Voynichese will be a language that shares enough phonemes with our languages to be recognised in the modern era.
  

Because otherwise, this is a waste of time.

• Voynichese lacks intonations
  

The lack of diacritics or similar tone indicators appears to bear out this theory, unless line initial or line ending glyphs indicate some such thing. This helps to narrow down the target languages, it can't be a tonal language.

• Voynichese is not regular
  

The creator of Voynichese would have been working off their own assumptions; they obviously did not have access to the huge amount of research into human speech that we do. This would help explain the distinct "Currier languages" - the scribe was making it up as he went along, and incorporating new sounds into his book as he came across them.

• Sunshine was an intelligent, but medieval author
  

We must therefore take a big step backwards and try to put ourselves into the mentality of a medieval intellectual. We may have a vague idea of what we want to achieve, but we lack the knowledge base to apply any scientific system to our method and are therefore making it up as we go along. I can't stress this enough, no matter in which direction we go.

Let us therefore use modern tools to attack Voynichese. We take the IPA and take a long draught, before opening our manual upon the You are not allowed to view links. Register or Login to view..

Quote:The general principle of the IPA is to provide one letter for each distinctive sound (You are not allowed to view links. Register or Login to view.), although this practice is not followed if the sound itself is complex.You are not allowed to view links. Register or Login to view. to represent single sounds, the way You are not allowed to view links. Register or Login to view. does with ⟨sh⟩, ⟨th⟩ and ⟨ng⟩, or single letters to represent multiple sounds the way ⟨x⟩ represents /ks/ or /ɡz/ in English.

There are no letters that have context-dependent sound values, as do "hard" and "soft" You are not allowed to view links. Register or Login to view. or You are not allowed to view links. Register or Login to view. in several European languages.

The IPA does not usually have separate letters for two sounds if no known language makes a distinction between them, a property known as "selectiveness".You are not allowed to view links. Register or Login to view.You are not allowed to view links. Register or Login to view.

Among the symbols of the IPA, 107 letters represent You are not allowed to view links. Register or Login to view. and You are not allowed to view links. Register or Login to view., 31 You are not allowed to view links. Register or Login to view. are used to modify these, and 19 additional signs indicate You are not allowed to view links. Register or Login to view. qualities such as You are not allowed to view links. Register or Login to view., You are not allowed to view links. Register or Login to view., You are not allowed to view links. Register or Login to view., and You are not allowed to view links. Register or Login to view.

Question
How to identify You are not allowed to view links. Register or Login to view.?
The number of glyphs roughly corresponds to most medieval European alphabets, even if no direct correspondence to the Latin alphabet has ever been found. We can thus assume that they are discrete letters, and form phones when merged into tokens. Obviously they do not directly represent phones, any more than the 26 characters of the English alphabet represent all of the different sounds used to speak English.
However, the regularity of glyphs within tokens makes me suspect that they may have a dual function. For example, certain combinations may indicate common word start or endings, and this would help explain away the LAAFU problem.

Question
How many phones are there?
You are not allowed to view links. Register or Login to view., which are 19 vowels and 26 consonant sounds. Latin; well, it depends on which period Latin. Italian has 32 phones, Spanish just 30. An interesting table of different languages can be You are not allowed to view links. Register or Login to view..

Question
So is there enough regularity in the glyph distribution to permit its phonetic structure to shine through? Well, I think there is. I'm starting to suspect that the glyphs work on both a bigram and digram level -some represent simple sounds, others represent more complicated or regular speech patterns. This is why we see regularity in the tokens, we could be talking about a proto-Italian dialect with the traditional "sing-song" vocal endings of words, etc. Certainly more research into this area is required.

Now, how to translate Voynichese into IPA? Two methods occur to me.

The first is arbitrary match patterns. That is, translating the most common glyphs into the most common phones by language and seeing if we recognise anything. A fun project, but not one I intend to start today.

The second is cribbing. Find words that we think we recognise the words of and try to match to IPA, then see if it works with other words.
Unfortunately, I haven't yet thought of a way to do this without imposing my own cultural reference upon it. Without any target language, it's no more a shortcut than the first option as we need to try lots of matches. It could even be slower than the first option, as we are limiting our options to something that is likely to be incorrect.

So, people, ideas please.

It's a correct remark that Medieval texts were meant to be voiced out. Popular belief has it that many readers were even unable to read without saying or at least mouthing the words at the same time.

There's the among linguists well-known anecdote from Augustine about his mentor Ambrosius:

Quote:"When Ambrose read, his eyes ran over the columns of writing and his heart searched out the meaning, but his voice and his tongue were at rest. Often when I was present—for he did not close his door to anyone and it was customary to come in unannounced—I have seen him reading silently, never in fact otherwise. I would sit for a long time in silence, not daring to disturb someone so deep in thought, and then go on my way. I asked myself why he read in this way. Was it that he did not wish to be interrupted in those rare moments he found to refresh his mind and rest from the tumult of others' affairs? Or perhaps he was worried that he would have to explain obscurities in the text to some eager listener, or discuss other difficult problems? For he would thereby lose time and be prevented from reading as much as he had planned. But the preservation of his voice, which easily became hoarse, may well have been the true cause of his silent reading."

Agustine was perplexed by Ambrose's ability to read in silence. In fact he tried to emulate the skill, but only got to the point where he was silently mouthing along with the text as he read. That is to say, for the (early) medieval reader, text and sound were one. I'm not entirely sure if this had evolved much by the 15th century. 

This is, in my opinion, an argument against overly complex ciphers, so worthwhile to point out. 
I don't know how it will help us further though. Aren't we still in the situation where we try to guess sound values for glyphs, as we've been for a hundred years?

Quote:Aren't we still in the situation where we try to guess sound values for glyphs, as we've been for a hundred years?

Exactly, as in my first method.

What we need is a dictionary of IPA transcriptions for common words across different languages. We could then run a pattern match across Voynichese to see if we start producing anything similar.

Maybe this isn't going to be very useful, but just consider this possibility..
What if the author consistently left out the first letter (or consonant) while writing to make it difficult to read for the people who didn't know the system? The idea obviously comes from Derek Vogt's findings (which I still cant get enough of, even though I believe he lost his way somewhere in the middle and ended up in no man's land), and maybe keeping this in mind might help those trying to assign sound values, and such a thing might give rise to a lot of single glyph words, which the VMS has in plenty.

I myself have tried it while trying to transcribe the MS in my own way, but it never really helped.. Maybe others trying to do the same can benefit by keeping this in mind..

Interesting point Common_Man.

I think this is a very interesting topic. Obviously I have my own views on the subject, and they are incorporated into my own theory which has its own separate thread. But here I can make some general observations, language-neutral and theory-neutral, based on my analysis of the Voynich character inventory.

Some Voynich characters are very rare, and if they represent distinct phones, they must be very rare phones. I have in mind [z], [v], and [x]. They are rarely even mentioned in discussions of Voynich characters. [g] is rare as well, although not quite so much as the other three.

A key set of Voynich characters occur in strictly restricted positions in the text: [p], [f], [m], [q], (i), [n], usually considered together as [(i)in]. It is hard to argue that any of them could be the only representation of any particular phone. 
[Note: I cannot type (i) with square brackets because it italicizes the entire text of the rest of the post.]

Then we have the characters that have the appearance of ligatures: [ckh], [cth], [cph], [cfh], and also the various forms of [sh]. I argue that these ligatures are much more likely to represent combinations or variant forms of phones, as is typical of ligatures, rather than the primary or only representation of any particular distinct phone.

These considerations drastically reduce the size of the set of Voynich characters that could each potentially represent distinct phones. I count only 11 of them! :

[a], [ch], [d], [e], [k], [l], [o], [r], [s], [t], [y]

Now we see one possible reason why the statistical properties of Voynich text consistently appear to be similar to those of Hawaiian: the latter is one of the few languages in the world, along with other Polynesian languages such as Maori, that have such a small inventory of phones and letters.

I hope readers can see from this analysis that the idea of characters with multiple or "ambiguous" phone values is not just some arbitrary and idiosyncratic idea that I got in my head for my own pet theory, but rather it follows directly from an objective and language-neutral analysis of the entire Voynich character inventory itself.

Of course there are many ways that 11 characters could represent all the phones of some language, but I think a breakdown for consonants such as 1) liquids, 2) nasals, 3) sibilants, 4) labials, 5) dentals, 6) gutturals is a natural and logical idea, and then similarly for sets of vowels such as 1) front vowels, 2) back vowels, 3) rounded vowels, etc., although there will be some overlap of these vowel categories.

These considerations would apply to any language, not just one particular language in one particular theory.

Geoffrey

How do you know when you get the sounds right if you dont know the language? If you guess the vowels correctly you could end up with something that sounds like speech but still doesnt mean anything, just as voynichese resembles writing in its own ordering but not so much when you use it to encode English, at least not with the letters as assigned.

(06-04-2019, 07:21 PM)davidjackson Wrote: You are not allowed to view links. Register or Login to view.Question
So is there enough regularity in the glyph distribution to permit its phonetic structure to shine through? Well, I think there is. I'm starting to suspect that the glyphs work on both a bigram and digram level -some represent simple sounds, others represent more complicated or regular speech patterns. This is why we see regularity in the tokens, we could be talking about a proto-Italian dialect with the traditional "sing-song" vocal endings of words, etc. Certainly more research into this area is required.
[...]
So, people, ideas please.

David, you mention the idea of the glyphs working on both a bigram and digram level. Even though this is not the idea of the theory that I currently hold, as you know, I think this is quite a logical and reasonable and plausible idea as well. I have even posted about the bigram theory idea on this forum.

Since you raise the idea of bigrams and you ask for ideas, I would like to share the following material with you. This is from the files of my own material on prior hypotheses of mine that I now consider obsolete and incorrect. I had a tentative working hypothesis about Voynich bigrams and Hebrew letters / phones. The facts and the data did not match the hypothesis, so in the end I rejected the hypothesis. In particular, the hypothesis required the changing of too many spellings of basic Hebrew words in a way that simply would not happen in the medieval period, since anyone then who knew Hebrew at all, would know its traditional standard written form.

But it may be of interest to you and others to see the kind of tentative correspondences that the hypothesis proposed between Voynich bigrams and Hebrew letters / phones. I attach to this post two files: One with the encryption from Hebrew letters / phones to Voynich bigrams, and the other with the decryption from Voynich bigrams to Hebrew letters / phones.

Even though I now believe this hypothesis was wrong, I do think that if any bigram theory could possibly work, it would probably end up with a correspondence table that looks rather like the two files that I attach here. I just want you and other researchers to be prepared for the messiness that is likely to ensue if and when you resolve to proceed down this path. I wish you better luck with the idea than I had with it!

Geoffrey

Very interesting thread, but let me throw in a caveat or two.

In the development of the logic of the opening post, it is of course perfectly acceptable to start with some assumptions, but let's look at them.

The first assumption is that the text is meaningful. All of it in fact. The logic of this assumption is quite reasonable: the discussion would become meaningless it it weren't. But it is an assumption nonetheless. By the way, with the phrase "all of it" I mean that it is also assumed that the text isn't a highly verbose encoding, such as (or example) every Voynich word would stand for one letter only. Or even every line.

The second assumption is that there is a reasonably close mapping of one Voynich character to one sound. Now I know that I did not phrase this very accurately, but I wouldn't know how to do it. In many (if not most) languages, small groups of characters are used to represent a single sound. The sound of English 'sh' for example, in IPA an integral sign, is written as a single character in many languages (scripts), as two characters in English and three in German (sch). So I understand that this would also be considered in the scope of the question of this thread.

The combination of these two assumptions really seems very reasonable. I daresay that the vast majority of people approaching the problem of the Voynich MS are making this combination of assumptions. However, I consider it entirely possible that this combination of assumptions is incorrect, and this would then explain the lack of success in understanding the Voynich MS text.

All this is very similar to the classical example of the guy looking for his lost keys near lanterns only.

But never mind. The exercise, even if it has a serious chance of being meaningless, is still interesting enough.

The first question related to phonology really has to be the separation into vowels and consonants. This has already been tried. The available algorithms (Sukhotin's, the use of Hidden Markov models) do lead to a split of the characters into vowels and consonants, but they don't tell us how successful / reliable these answers are. From my own on-going dabbing into HMM analysis, I can only conclude (prematurely), that this split isn't actually anywhere as successful as for languages like Italian or Latin. It doesn't really work.

Even while the Eva transcription alphabet makes the Voynich text almost pronounceable, it isn't really. I mean, someone from Russia or Poland might have no problem to pronounce a Voynich word like chcthy, but this doesn't work for everybody.

Another risky thing is to try to decide things on the basis of the shape of the character. This clearly does not work for the Latin alphabet. Just looking at majuscules, O and Q appear very similar, but obviously have no relation whatsoever. Same with E, F and L. It may therefor not work for Voynichese. However, here, we have the very funny effect that similar-looking shapes may appear in similar contexts. This is a key observation that requires an explanation, and I am not so sure that such an explanation can be a linguistic / phonetic one.

A few years ago I wrote a post about You are not allowed to view links. Register or Login to view. It isn't based on the script or how the glyphs work together, but rather on what sounds languages tend to have. It's a starting point from which to begin thinking, if you want to pursue idea. The number of glyphs in the script represents the minimum number of sounds, I would guess, but there might be quite a few more.

One way to identify which glyph might be which sound is to study how the glyphs fit together. The sounds within languages tend to fit together in particular ways. These are characteristic of individual languages but follow certain principles. I've written about You are not allowed to view links. Register or Login to view. which is a helpful start to understanding some basic of phonotactics.

On the issue of speech and writing, and how the spoken words might influence the form of writing in early literacy, I've written about You are not allowed to view links. Register or Login to view. This also feeds into (with some change in opinion on my part) the idea of You are not allowed to view links. Register or Login to view.

I'm happy to discuss any part of the (potential) phonology of the Voynich text.