It is well known that certain structures and patterns are typical of the beginnings of lines, and very different structures and patterns are typical of the ends of lines, in the Voynich ms text. But has anyone studied the relative entropy levels of just the beginnings of lines, and just the ends of lines, compared with the whole text? For example, what is the entropy of just the first halves of all lines of the ms text? And of just the second halves of all lines? What about just the first three vords of all lines, and just the last three vords of all lines? 

I suppose the expectation would be that all such entropy levels would be even lower than that of the whole ms text, since the beginnings of lines by themselves and the ends of lines by themselves would be expected to be even more similar to each other than the whole text is. But I wonder if that has ever been tested. There is also plenty of repetition within each line, from beginning to end, so I wonder if breaking apart line beginnings and line endings will really lower entropy so much. I also wonder if there is any significant difference in the entropy of just the beginnings of lines and just the ends of lines. 

The reason I ask is that I have come across certain groups of lines in my research where I find it easier to interpret certain parts if I only read the first few vords of each line and continue with the first few vords of the next line, ignoring the rest of each line. But this may well be an illusion on my part, which is why I am curious about the entropy statistics of just the beginnings of lines and just the ends of lines. 

It would be possible to encrypt a message by only making the first three words of each line meaningful, and padding out the rest of each line with nonsense nursery rhyme repetition of the sounds of the first three words:

meet me at fleet be mat sleet we vat
the back door he lack moor we sack poor
of jons house off cons mouse scoff nons louse
monday at noon sunday cat moon runway sat loon

This doesn't seem like a very secure level of encryption, but now combine it with a simple substitution cipher, or even better a mysterious invented script that no one else knows. Then it would become rather difficult to decipher. By the standards of the early 15th century, it would have probably been quite secure. And both the concept of simple substitution (possibly incorporating elements of a verbose cipher, as we have recently been discussing elsewhere on this forum) and the steganographic concept of hiding the words of a meaningful message within a larger nonsensical message were simple enough to have been known and possibly employed in the time period of the Voynich ms. 

Also, in filling out the nonsensical parts of each line, the author could very well have followed some of the principles of the "auto-copying" or gibberish theory that has also been discussed recently on this forum. That could have been deliberate, or it even could have happened unconsciously as the author thought of nonsense rhyming words and phrases to fill out each line. I found myself doing it as I wrote the lines above, at first accidentally and then deliberately after I noticed I was doing it. It's only natural to take "inspiration" from the other nonsense words and phrases that are already in the immediate vicinity of the line that one is filling out. And for the author and the intended recipient, it doesn't matter what those words and phrases are anyway. 

If the principle of filling out the nonsense parts was based on choosing rhyming words and phrases, as in my example above, then one would also expect the middles and ends of words to have much lower entropy (more predictability) than the beginnings of words, a statistical feature that we also observe to be present in the Voynich ms text. 

I am aware that the beginnings of lines also tend to be repetitive in the Voynich ms text, so I do not at all expect the idea I raise here to solve all the problems inherent in the difficult structures and patterns of the ms text. But I'm curious if the entropy breakdowns by parts of lines may give us some clues and leads to follow for further and more sophisticated examination of these ideas.

Came across this paper (attached below) which seems to address a concern that is becoming more and more of an issue for my ongoing testing of possible approaches.

I know that this was of interest to a group including Koen and nablator back in 2019 and Koen was going to e-mail (see this thread You are not allowed to view links. Register or Login to view.)

Koen -- did you ever hear back from him?

I know that you and nablator did some separate investigations that looked for repeated word strings and reached some conclusions, but that question is a little different than what Fincher's work supposedly shows -- as far as I could tell from this paper.

Of course, Fincher's paper suffers maybe from a lack of breadth of data -- like all of this type of work, the more languages tested the more convincing it would be.

Any other thoughts on this paper and its approach?  I continue to hone my ability to critically evaluate such work . . . 

Also, have to ask it:

Any thoughts on Fincher's conclusion that word transposition has to be involved to get the VM into a "natural language" form?

Attached Files

  WPPA.doc (Size: 207 KB / Downloads: 114)

This post is a follow-up to the "VCI" alphabetic interpretation of the script of the Voynich ms text, based on a verbose cipher analysis of the script. I posted about my VCI system You are not allowed to view links. Register or Login to view., and later modified it into a form I call "Slavic VCI" as described You are not allowed to view links. Register or Login to view., all of which I attempted to summarize You are not allowed to view links. Register or Login to view.. 

The following table presents the 25 most frequent "vords" in the Voynich ms text, first in their standard EVA transcription form, and then in my "Slavic VCI" alphabetic interpretation. Please note, for example, in line with the verbose cipher analysis, that EVA [aiin] and [ain] = Slavic VCI <o>, EVA [ol] = Slavic VCI <z>, EVA [or] = Slavic VCI <w>, EVA [qok] = Slavic VCI <n>, etc., and EVA [y] is treated as a null character, marked as Slavic VCI <#>. 

The table then presents both Upper Sorbian and Czech words, suffixes, and syllables that appear similar to the Slavic VCI forms. For many of the Czech words it is also possible to include their rankings on a modern Czech word frequency list, which one can find You are not allowed to view links. Register or Login to view.. Finally of course an English gloss is presented for each Upper Sorbian / Czech word or morpheme. 

Naturally the correspondences are not perfect in all cases, but I find the number and proportion of very close resemblances striking. It is most difficult to be precise about the distinctions between the numerous sibilant letters and phonemes in these languages: s, z, š, ž, c, č, ć, etc. Now if it were written in literary Czech, which is well known for its phonemically precise alphabet established in the early 15th century, and if we knew for certain that no diacritical markings had faded over the centuries in the copy of the Voynich ms that we have today, then we would indeed expect to see great precision in all of these details in the script of the ms text. But if on the other hand it were written in a dialect of Sorbian, which had little or no written history at that time, and if some diacritical markings in the ms may possibly have faded over the centuries, then a certain level of imprecision in the indication of these sibilant phonemes in the readable ms text that we have would be more understandable. 

Geoffrey

   

I'm sure I'm not the first person to consider this idea, but one way to deal with the difficulty of analyzing the ubiquitous vord-final glyph EVA [y] is simply to treat it as a null character. That is to say, perhaps it has no phonetic content whatsoever. Maybe it's a word-divider, or maybe it's something else. Perhaps the author simply liked the way that glyph looked as a very common medieval Latin ms suffix (and sometimes prefix) abbreviation, and decided to use it to end (and sometimes begin) words or syllables or whatever the vords are, wherever he pleased. 

This analysis would greatly increase the variety of possible vord endings in the Voynich ms text. In most languages words and syllables can end with more than just a small handful of letters and phonemes. 

I'm curious what kind of statistics Koen's and Marco's verbose cipher analysis would produce for h2 conditional entropy and h1 entropy, if all the [y]'s are simply disregarded throughout the ms text? 

For those who are curious, I took a modified version of my You are not allowed to view links. Register or Login to view., which following Rene I should call an alphabetic interpretation rather than a transcription, and applied the reading of EVA [y] as a null character. I have marked it as <#> so that it is still clear exactly where these nulls are in the text. Also, in this particular version of my modified VCI interpretation, I have focused on the possibility of a West Slavic reading of the text (Czech, Polish, Sorbian, et al.), so the modifications of VCI have been made with such languages in mind. They include [d]=<c>, [od]=<h>, isolated [o]=<ch> (voiceless velar fricative /x/), [or]=<w> (in Polish and Sorbian a certain type of Slavic "l" sound is pronounced /w/), [al]=<as>, [am]=<ás>, and [ain]=<o>. These are all minor changes in the details of the phonemic interpretation. 

I have selected the first "recipe" in the "Recipes" section, the first four lines of folio page f103r:

<[P]jecas  rjc#  #péj#Pi#  be#  #srje#  mé#  mas  rjec#  #rjcas  co  dz  cas  c#>
<co  rjet  iPjic#  casd#  Bjec#  Perjz  jeP  al  bi#  ras  sté#  ral  o  d  jec#>
<#rjco  rjét  jeb#  idas  jec#  itj#  w  wz  do  jas  b  tal  b  j#s>
<#jec#  nec#  dec#  né#  de#  icalz  sb#  jecal  as#>

I find it curious that the first word <[P]jecas> rather resembles a widespread Slavic form meaning 'you (sg.) bake': Czech "pečeš", Upper Sorbian "pječeš", Lower Sorbian "pjacoš", Polish "pieczesz", etc. 

This is just one word of course, but the more general idea of EVA [al] = Slavic VCI <as> = 2nd person singular verb ending may possibly apply to a number of vords that appear to be rather appropriately spread out throughout these lines: <mas>, <#rjcas>, <cas>, <ras>, <idas>, <jas>. In several cases it is easy to recognize similar-looking Slavic forms, such as "maš" meaning 'you have' and "říkáš" meaning 'you say'. Naturally the phonological details may vary depending on the language and the dialect. 

This Slavic reading and interpretation is of course highly speculative, and this is only a very preliminary rendition of just four lines of text. But the main thing I find interesting here is the more natural appearance of the text when EVA [y] is simply treated as a null character. 

Geoffrey

A case for Gibberish and no one should suffer anymore by the hands of the Voynich Author

1. VMS contains a low use of 1-2 tokens
   
2. NO evidence for punctuation in the VMS whatsoever.
   
3. Tokens repeat geometrically to often compared to languages
   
4. Even in the modern era there is no decryption for the VMS
   
5. If this was an invented language it still should obey a normal language rhythm for token repeats
   
6. otol is a label for a star, plant and a empty pipe
   
7. Lastly what likely would be consonants can be found at times as 1 character tokens
   

The Voynich vord ending [-dy] is well known for its ubiquitous presence in the ms text. But recently I have found some rather striking statistics about another interesting vord ending:
[-ckhy]. 

Of course [-ckhy] cannot be anywhere near as frequent as [-dy], because the ligature [ckh] itself only occurs with modest frequency in the ms text. But what is striking to me is just how large a portion of all occurrences of [ckh] in fact appear as part of this particular vord ending [-ckhy].

I have followed here Koen's and Marco's verbose cipher analysis ideas, by which [ok] is treated as a single unit distinct from [k]. Thus I treat [ockh] as a distinct unit different from [ckh]. My analysis here focuses on [ckh] and [-ckhy] without [o] before them. 

It is probably well known that vord-final [-dy], occurring 6690 times, constitutes over half of all 12505 occurrences of [d] in the Voynich ms text.

I find it notable that vord-final [-ckhy] (without preceding [o]), occurring 360 times, also constitutes over half of all 706 occurrences of [ckh] (without preceding [o]) in the ms text as well.

No other glyphs share this level of frequency of occurrence as vord-final [glyph+y]. Vord-final [-ockhy] makes up 42% of all [ockh]. Curiously, the preference is reversed for [cth], where final [-octhy] makes up 43% of all [octh], but final [-cthy] without preceding [o] only makes up 37.5% of all [cth] without preceding [o]. There would be no reason to expect such discrepancies between [ckh], [ockh], [cth], and [octh] if the ms text were merely generated by some kind of automatic self-copying algorithmic method. 

It is interesting to note that the special top-line ligature glyphs have a somewhat less frequent occurrence before final [y]: vord-final [-cphy] only makes up 23% of all [cph], and final [-cfhy] only 31% of all [cfh]. Like [cth], but unlike [ckh], these percentages rise somewhat with a preceding [o]: final [-ocphy] makes up 39% of all [ocph], and final [-ocfhy] 40% of all [ocfh]. It should be noted that the sample sizes are getting extremely small for these last glyph sequences: Here 40% means just 6 out of 15. 

Thus we have the curious fact that [ckh] is the only one of the ligature glyphs which is more likely to occur before vord-final [y] without a preceding [o] than it is with a preceding [o]. 

It is probably not surprising that vord-final [-ey] is also rather frequent, making up 27.6% of all [e]. All other glyphs occur much less frequently before vord-final [y]: final [-oshy] is 19% of all [osh], and final [-ochy] is 15% of all [och], but these percentages decline significantly to 6% and 9% without the [o]. All other glyphs, with or without preceding [o], occur less than 10% of the time before vord-final [y]. 

===============

Therefore I was curious to look into this very frequent vord-final [-ckhy] (without preceding [o]) more deeply. In fact, it turns out that the vast majority of occurrences of vord-final [-ckhy] without preceding [o] comprise just five Voynich ms vords: [chckhy] (140), [shckhy] (60), [checkhy] (47), [ckhy] (39), and [sheckhy] (35). Together they make up 321 of the 360 vord-final [-ckhy] without [o]. No other such Voynich vord type occurs more than 3 times in the ms text. 

We thus arrive at the striking conclusion that these five Voynich vords [chckhy], [shckhy], [checkhy], [ckhy], and [sheckhy] constitute a sizable 45.5% of all [ckh] without preceding [o] in the ms text. 

Even the famous [chedy], [shedy], [qokeedy], et al., cannot match this feat: The five most frequent vord-final [-dy] vords [chedy], [shedy], [qokeedy], [qokedy], and [dy] constitute only 17.4% of all occurrences of [d] without preceding [o] in the ms text. 

For comparison I also checked these five specific [-ckhy] vords with [cth] substituted for [ckh]: I found 79 [chcthy], 31 [shcthy], 28 [checthy], 111 [cthy], and 20 [shecthy]. Together they constitute 34% of all [cth] without preceding [o]. Most striking is the much greater frequency of [cthy] itself as a vord, compared to [ckhy] itself. All the other vords occur more frequently with [ckh] than with [cth]. Again, these are unexpected discrepancies that cannot be easily explained by an automatic self-copying algorithmic method. 

Finally, I checked these five [-ckhy] words with the preceding [o] inserted: I found 21 [chockhy], 5 [shockhy], 10 [cheockhy], 13 [ockhy], and 2 [sheockhy]. Together they constitute 25.4% of all [ockh]. 

In conclusion, I think it may be worth looking more deeply into the five Voynich vords [chckhy], [shckhy], [checkhy], [ckhy], and [sheckhy] that make up a striking 45.5% of all occurrences of the ligature [ckh] without a preceding [o] in the ms text. Also, the one vord [cthy] that occurs more frequently than [ckhy] may be worthy of particular attention and investigation as well. 

Geoffrey

If you made the VMs, what would it be? How is the information presented? Why were these images and this text created - either in whole or in part? - where there have already been such a variety of suggested interpretations. And what can be learned from this manuscript as it is investigated, in the current circumstance?

If you made the VMs, you would already be familiar with the old traditions and with the relevant science of that historical era. 

If you made the VMs, you would pretty much need to confine information, references, methods and techniques to the first part of the 1400s - unless some excuse is provided. Just to avoid anachronism. While some complexity and sophistication must be allowed.

If you made the VMs, is it false and meaningless? Is it intended to be scientific or literate? Is it a facade for a disguised reality- encoded, encrypted - or whatever?

My view is that the VMs is the best surviving example of an attempt to embody Roger Bacon's first rule of obfuscated communication, which is *not* to let the fact of obfuscation be plainly obvious as in many early code systems. And for the investigator having the necessary familiarity with tradition (heraldry) and with science (botany) as would be contemporary with the parchment dates, then one of the first cracks in the VMs facade takes the shape of a nebuly line.

I'm just opening a thread for discussion, but I believe people should stop trying to think of translating the manuscript using known languages and instead look for patterns in the text and try to relate what may be said in it to the pictures. Then maybe we could understand this unique language. For example if a certain word structure comes up often they may be connectives etc.

Just an idea, I'm not sure if people have already tried this I just registered to the website.