The Voynich Ninja

These gallows are the field still largely unexplored.

One more consideration: sometimes (but not always!) the one-legged plain gallows (p and f) have increased spacing not only rightwards (which would be explained by the "coverage hypothesis" as described above), but also leftwards.

This suggests that at least in some of the cases these gallows were inserted afterwards - into empty spaces provided in advance. The width of the space was such as to accomodate a two-legged gallows (or even any other character maybe) conveniently; so when a one-legged gallows is inserted there, some spacing is still left to the left or to the right (or both ways), and this becomes notable.

This leads us to the concept of fillers or to the principle of several passes in which the text was written. Maybe this is true not only for the gallows but for other characters as well; one-legged gallows only naturally highlight this circumstance. In fact, many words have increased spacings within, like being "constructed" of several parts.

However, the concept of giving flourishes to letters, especially upper-case strokes, was an accepted scribal tradition in all European languages during the late Middle Ages and early Renaissance.
So the flourishes could simply be the scribe breaking up the monotony of writing as he progressed, with no other intent.
I remind you of this not to discourage the theory, but simply to point out that it's not unusual to see such flourishes in works of the era.

Yes, and certainly the embellishment is (and has been indeed, in the Voynich discussions) the natural assumption in the first place.

For a somewhat more comprehensive check of the "coverage" hypothesis, one may do the following (limiting this to plain p and f, for simplicity):

1. Locate all words containing plain p and f 
   
2. For each word, write down the order of the coverage (e.g. if the loop is covering one character, the order is 1, if it is covering two characters, the order is 2, and so on; if the gallows does not cover anything, the order is 0)
   
3. Check if the non-zero order of coverage is preserved for occurrences of a given word as identified in point 1 above. If it is, at least in some of the occurrences, then we observe a system. If it is not, then possibly these are flourishes, indeed.
   

Unfortunately, I have not done any tests in this direction and cannot provide any raw data.

Prior to doing any tests though, we have to agree about the extent of the coverage itself - should it be marked at the extent of the main loop, or at the extent of the top right ear of the gallows?! The case with t  suggests the first way. By the way, those t's with their legs in different places look to me a strong point towards the coverage system. Because obviously this behaviour cannot be attributed to embellishment.

One of the most important aspects (IMHO) of the gallows is so obvious that it is rarely mentioned. This is, that Eva-k and Eva-t, which have a similar shape, can also occur in exactly the same contexts. Take any word with one of them, replace it, and you have another valid word (with a very high probability).
Without going into a list of possibilities how this could happen, I think it is very important. 
Note that this also works for ch and sh, and r and l, just to name a few.

It does not work for k and p (or the other three similar options). Currier was adamant that p and f are never ever followed by Eva-e, though in more complete transcriptions one may find a few examples of that. In any case, k and t are of course very often followed by e. p and f might be interchageable, but I am not certain of that.

For me, this confirms the conclusion from entropy analyses and the existence of (not very strict) word patterns, namely that one should not hope to turn the Voynich MS text into a meaningful plain text by simple substitution. 
This should be phrased more precisely of course. It means that any such plain text is necessarily in a language that exhibits the same features: special entropy, word patterns, replaceable characters.

(02-01-2016, 02:42 PM)Anton Wrote: You are not allowed to view links. Register or Login to view.Yes, and certainly the embellishment is (and has been indeed, in the Voynich discussions) the natural assumption in the first place.

For a somewhat more comprehensive check of the "coverage" hypothesis, one may do the following (limiting this to plain p and f, for simplicity):

1. Locate all words containing plain p and f 
   
2. For each word, write down the order of the coverage (e.g. if the loop is covering one character, the order is 1, if it is covering two characters, the order is 2, and so on; if the gallows does not cover anything, the order is 0)
   
3. Check if the non-zero order of coverage is preserved for occurrences of a given word as identified in point 1 above. If it is, at least in some of the occurrences, then we observe a system. If it is not, then possibly these are flourishes, indeed.
   

The following image is a collage of all words starting with EVA 'p' or 'f', sorted alphabetically according to the respective EVA transcriptions:
You are not allowed to view links. Register or Login to view.


It was originally assembled for a different purpose but might help to compare features of the gallows characters. There is also one for 't' and 'k' here:
You are not allowed to view links. Register or Login to view.

I think your coverage hypothesis is interesting. It's also the type of feature that could increase the overall entropy of the text.

On the other hand, looking at the above images, i don't see really see a pattern.

(23-01-2016, 10:47 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.One of the most important aspects (IMHO) of the gallows is so obvious that it is rarely mentioned. This is, that Eva-k and Eva-t, which have a similar shape, can also occur in exactly the same contexts. Take any word with one of them, replace it, and you have another valid word (with a very high probability).
Without going into a list of possibilities how this could happen, I think it is very important. 
Note that this also works for ch and sh, and r and l, just to name a few.

It does not work for k and p (or the other three similar options). Currier was adamant that p and f are never ever followed by Eva-e, though in more complete transcriptions one may find a few examples of that. In any case, k and t are of course very often followed by e. p and f might be interchageable, but I am not certain of that.

For me, this confirms the conclusion from entropy analyses and the existence of (not very strict) word patterns, namely that one should not hope to turn the Voynich MS text into a meaningful plain text by simple substitution. 
This should be phrased more precisely of course. It means that any such plain text is necessarily in a language that exhibits the same features: special entropy, word patterns, replaceable characters.

While I cannot speak for the first of these characteristics, I would suggest that the other two are pretty normal for natural languages. Although the Voynich language my exhibit them more thoroughly than typical, I don't think they are peculiar.

The idea of gallows being used as articles is a possibility.  I would point out that this doesn't mean that spaces aren't spaces.  They are and whatever gallows are won't change that.  What it actually means is that words aren't words. 

It means that rather than encoding phonemes the characters are possibly being used as sememes.  Each glyph could be conveying semantic information and it's position in the glyph-group/word/string could change its meaning.  Rather than being articles initial gallows could be a semantic marker that say "This is a noun" or "This glyph-group is about earthly things". The gallows anywhere else in a glyph-group could have a phonemic value in some positions or a modifying semantic meaning in others.  This feature could also apply to glyphs which aren't gallows.

Or its possible it has both a semantic and a phonemic value.  (For example word in English that start with GL frequently refer to light or the eye.  Glossy, glassy, glare, glimmer, glance, etc. So GL is phonemic (two phonemes technically) in English, but there is a semantic aspect to GL as well, and this goes away if it's not at the start of the word UGLY, IGLOO, ANGLE.  It's not a rule in English, but the existence of the pattern means in could develop as a rule elsewhere.  Loan words obviously can break the pattern, but if you had a rule rather than a pattern you probably wouldn't adopt a word that broke it.

Anyway, if glyphs work this way they could encode a great deal of information, especially if repeating a glyph changed it's meaning too.

(01-02-2016, 08:42 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.

(23-01-2016, 10:47 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.This should be phrased more precisely of course. It means that any such plain text is necessarily in a language that exhibits the same features: special entropy, word patterns, replaceable characters.

While I cannot speak for the first of these characteristics, I would suggest that the other two are pretty normal for natural languages. Although the Voynich language my exhibit them more thoroughly than typical, I don't think they are peculiar.

As Jacques Guy used to say in the distant past, it will be hard to find any peculiarity that doesn't exist in some language somewhere.

On these two or three points, it may be a matter of taste whether one calls them peculiar or something more strongly.
If we take Mandarin Chinese for example, its rendition in Pinyin shows a structure that is even more rigid (I would say) than the Voynich MS. There are only some 400 allowed combinations (not counting the tones), and one can almost freely replace consonants by other ones and end up with another valid (pinyin) word.

Other East Asian languages will have various different rules.

For me the equivalent functions of k / t,  ch / sh, and l / r  remain highly suspicious. The best I can come up with is that someone was trying to render a language with unknown sounds, and had to make choices on the fly....

This doesn't work for everything though. The ending -m can appear where -l or -r can appear, but preferably at line ends. This more suggests an elaborate way of writing (say) -r,  but then there are labels that differ only by the ending -r or -m.
So far, I can only think of quite contrived explanations.

(03-02-2016, 12:19 PM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.As Jacques Guy used to say in the distant past, it will be hard to find any peculiarity that doesn't exist in some language somewhere.

On these two or three points, it may be a matter of taste whether one calls them peculiar or something more strongly.
If we take Mandarin Chinese for example, its rendition in Pinyin shows a structure that is even more rigid (I would say) than the Voynich MS. There are only some 400 allowed combinations (not counting the tones), and one can almost freely replace consonants by other ones and end up with another valid (pinyin) word.

Other East Asian languages will have various different rules.

For me the equivalent functions of k / t,  ch / sh, and l / r  remain highly suspicious. The best I can come up with is that someone was trying to render a language with unknown sounds, and had to make choices on the fly....

This doesn't work for everything though. The ending -m can appear where -l or -r can appear, but preferably at line ends. This more suggests an elaborate way of writing (say) -r,  but then there are labels that differ only by the ending -r or -m.
So far, I can only think of quite contrived explanations.

Two sounds which seem functionally equivalent within words and can be freely switched is relatively normal in languages. The sounds may appear in the same word environment and have contrastive distribution--they can replace one another to make a new word. Often those sounds share some fundamental relationship, or at least have some commonality: vowels can replace vowels but not consonants, stop replace stops but not sonorants, and so on. Phonologically this is all fine.

The main difficulty is that the words themselves may or may not exist. Like you say with Mandarian Chinese, the number of possible syllables is rather small and many (or most?) exist as words. Some language have an even smaller number of possible syllables, and making any permissable swap is highly likely to give a real word. Yet even in English, which has a complex word structure and a lot of "gappiness" in what words actually occur (tens of thousands of syllables are possible, supposedly), we can still identify real "sounding" words from unreal ones: glik might be something, but kgil isn't.

It would seem we're simply dealing with a language where: 1) [k, t], [ch, sh], and [r, l] are phonologically related sounds, and 2) the number of possible syllables is relatively small. I'm happy with both of these points, especially as I believe there are elements of featuralism in the Voynich script (You are not allowed to view links. Register or Login to view.) and these pairs of characters have graphic similarities too.

As for the line end letter [m], it is really a part of a bigger issue that includes line beginning transformations, what Currier would have called the "line as a functional unit". They've vexed me for some months now though I've sought to get a handle on them. I can only just about describe the problem, and any solution to the whole is far away (though likely very important). I think there is a systemic problem which causes them both, possibly something to do with sandhi, pausa, and the messy transition from oral to written culture. I'll write about it before long, but I'm not sure if it is anything more than a shaky guess.

It's worth looking at some numbers for character substitutions in different languages.

The following gives the percentage of words that are still valid after a character substitution, in some texts of roughly equivalent lengths.

For example, starting with Pliny:

Pliny - Natural History (Latin):
m replaced by s: 213/3074 7.0%
o replaced by i: 226/3634 6.0%

In Pliny's vocabulary, if we take a word containing an "m" and replace it with an "s", we get a valid word 7% of the time. There are other valid substitutions but they yield less than 5% - i'm only listing the most significant ones.

Bible (Latin):
m replaced by s: 136/1887 7.0%
o replaced by i: 107/1828 6.0%

Very much in line with Pliny above.

Dante - Divine Comedy (Italian):
a replaced by e: 348/4187 8.0%
a replaced by o: 377/4187 9.0%
e replaced by a: 320/4046 8.0%
i replaced by e: 318/3789 8.0%
o replaced by a: 383/3521 11.0%
z replaced by t: 21/220 10.0%

Vowels with vowels and consonants with consonants, as Emma noted.

Melville - Moby Dick (English):
j replaced by b: 12/149 8.0%
j replaced by p: 15/149 10.0%
j replaced by r: 11/149 7.0%
j replaced by s: 14/149 9.0%
w replaced by t: 71/955 7.0%
z replaced by d: 9/127 7.0%

Bible (Hebrew):

<kaf> replaced by <he>: 68/274 25.0%
<kaf> replaced by <vav>: 80/274 29.0%
<kaf> replaced by <yod>: 47/274 17.0%
[font=Courier New]<kaf> replaced by <mem>: 50/274 18.0%[/font]

Hebrew has the highest values by far...

Voynich Manuscript - A Folios:
f replaced by d: 35/132 27.0%
f replaced by k: 54/132 41.0%
f replaced by p: 45/132 34.0%
f replaced by t: 55/132 42.0%
k replaced by t: 310/992 31.0%
m replaced by l: 89/200 45.0%
m replaced by r: 99/200 50.0%
p replaced by k: 119/303 39.0%
p replaced by t: 118/303 39.0%
r replaced by l: 239/777 31.0%
t replaced by k: 310/833 37.0%

Voynich Manuscript - B Folios:
f replaced by k: 90/223 40.0%
f replaced by p: 67/223 30.0%
f replaced by t: 83/223 37.0%
k replaced by t: 428/1566 27.0%
m replaced by l: 113/241 47.0%
m replaced by r: 118/241 49.0%
p replaced by k: 177/574 31.0%
p replaced by t: 184/574 32.0%
r replaced by l: 380/1208 31.0%
s replaced by c: 376/1313 29.0%
t replaced by k: 426/1017 42.0%

...but the Voynich is even higher.

Given that the glyphs for "m" and "r" are so similar, it's possible that some of these pairs represent the same character.

In any case, if they are in fact different characters, we can at least rule out some vowel/consonant assignments.

You can see the distribution for valid (green) and invalid (red) words for these assignments here:
You are not allowed to view links. Register or Login to view.

For example:
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.
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.
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.