The Voynich Ninja

One more interesting manuscript recently digitized, with many charts.

You are not allowed to view links. Register or Login to view.

Note the swapped T-O map in 2r.

Note also the chart in 20r where almost every label starts with "p" (I presume, the abbreviation for "per"), which somewhat reinforces my idea of Voynich label prefixes as operators.

Hi Anton,
in 2014, You are not allowed to view links. Register or Login to view. discussed the idea that star labels may include a prefix corresponding to the Arabic determinative article 'al'. He identified the possible prefix with EVA:ot.
While it is quite rare to see the same prefix occurring in most labels in a Latin manuscript, this is indeed quite common in Arabic manuscripts (the | character on the right-side of each label).

[attachment=4876]

These are the histograms for the most frequent  initial characters of labels in You are not allowed to view links. Register or Login to view. and Voynich Pharma and Zodiac. I have transcribed with '+' the barred-p standing for 'per' in the Latin ms.

[attachment=4879]

The diagrams are overall comparable, with the most frequent character clearly standing out as an outlier. For an Arabic manuscript, I think a single character would appear at the begin of almost all labels (~100%).
Plots for the most frequent initial bigrams are quite different. The Latin ms has a flat distribution, as is expected since the second character actually is the first character of a second word following "per". In the VMS (in particular in the Zodiac) there is a strong predominance of 'ot' / 'ok' with respect to other bigrams.

[attachment=4878]

I ask myself: what is the prefix?

• if it is just 'o', it is unclear why it does not attach uniformly to all characters (this preference for gallows appears in the whole ms, not only in labels)
  
• if it is 'ot' as Stephen thought, it is not clear why 'ok' behaves so similarly
  
• if 'ot' and 'ok' are two different prefixes, their nearly identical distribution is hard to explain.
  

We also face the more general problem of word length. Voynichese words have an average length that is comparable with that of European natural languages and labels appear to be normal words, many of which also occur in the main text. As You are not allowed to view links. Register or Login to view., there is strong evidence suggesting that q- is a prefix attached to o-words. But if we assume that o- and qo- (or ok- ot- qok- qot-) are not "truly" part of the words, the average word length becomes very short (in particular in Currier B, where qo- is so frequent). Of course, the problem of word length is made worse by low entropy, which, as Rene pointed out,  suggests that words should be "compressed" in some way.

Also, the higher frequency of o- words in the labels is symmetrical to the lower frequency of q- (You are not allowed to view links. Register or Login to view.). It's more "q- is added to label-words in the main text" than "o- is added to main-text-words in labels".

In my opinion, the behaviour of the o- prefix in labels and in the main text could be suggestive of an artificial language in which there is a deeper relationship between grammar and morphology than in natural languages. In a natural language, "operators" can be applied to certain words and not others on the basis of their grammatical features, not their initial character (though it is true that, in some languages, the initial character can alter the operator).
Random and esaily rejectable examples, just to give an idea of what I mean:

• masculine nouns start with ot-, feminine with ok-, neutral with other prefixes;
  
• masculine nouns are constrained to start with a gallows, o- marks the nominative case, qo- marks the accusative;
  
• masculine nouns are constrained to start with a gallows, o- marks the singular, qo- marks the plural.
  

I am sure I did not explain myself clearly and this is largely due to the fact that I find the subject complex and hard to grasp. I cannot think of anything that could really explain all the observations. In my opinion, the Latin diagram you pointed out is great both for its similarity and difference with Voynichese labels. Thanks again for sharing it!

Hi Marco,

I've selected those two diagrams because they accidentally relate to some distict discussions in the forum: the swapping of fields in T-O maps and the null-labels. The "null-labels" is Mark Knowles's idea which we discussed recently in the thread about labels. One of his arguments was that if so many labels start with "o", then those may be just null-labels - which are just a subset of the more general set of null-vords introduced to confuse the future would-be decipherers (like ourselves).

The diagram of this Lullian MS is a good example that many labels starting with the same character is a perfectly valid scenario in itself, and it does not necessarily invoke the need for the null-label concept.

About the prefixes. Yes, I remember that Stephen's idea about "al-", in fact I think I referred to it in the labels thread. In star diagrams like f68 that would be appropriate because many star names are Arabic.

My idea has been that o-, qo-, y-, and I already don't remember which other prefix, stand for operators like basic function words which can be equally applied to nouns and perhaps other parts of speech. This in itself presumes the synthetic-language scenario, which may or may not be based on a nomenclator in respect of nouns. If it is, then one can imagine possible reasons for predominance of ot- and ok- over other starting bigrams. Actually, if we strip the o-, then that predominance means that there is the predominance of vords (let's say nouns) starting with "t" and "k". Hence I would suspect that vords starting with "t" and "k" - not all of them, perhaps, but only those occurring as parts of labels,- are two nomenclator sets of notions homogenous in a certain respect. Like, to describe objects of the set A, the author used "t" as the starting character, and to descibe objects od the set "B", the author used "k" as the starting character. So the starting character (the prefix being stripped) is a sub-nomenclator marker.

I have not done anything to confirm or disprove this idea, it's just a raw idea out of the box!

How many these subvords (those starting with k, and those starting with t) are there in the labels? Is it tens, scores or hundreds?

(11-10-2020, 07:48 PM)Anton Wrote: You are not allowed to view links. Register or Login to view.How many these subvords (those starting with k, and those starting with t) are there in the labels? Is it tens, scores or hundreds?

Hi, Anton:

Hand counting from Stolfi's label list here: You are not allowed to view links. Register or Login to view.

there are about 130 different unique labels starting with "ok" and about 160 different unique labels starting with "ot."

Of course, with hand-counting and various transliteration disagreements I am likely off a bit with both of these numbers -- but it gives you a sense of scale.

Thanks,

Michelle

Thanks! That's pretty many. Does not look like a short list of objects.

Using Takahashi Transliteration for labels with more than 1 letter, i get :
Total label count: 947
Distinct labels i.e set: 733
dodgy labels 23

Labels that begin with 'o' : count