Why and how the text could be Bavarian

[JoJo_Jost]

In reverse, it is the glyph that enters the word, the key: from a selection of y, n, r, l:

Here, the y was chosen.
Then the “d” as a branch, one of two possibilities (d or not d),
Then the “E-run,” here in the form of a single “e,”
Then the “base glyph” of this series—ch, sh, k, t, and others—here k

Working backward, this is a clear, very uniform structure, followed—in some glyphs—by additional glyphs (in some longer words) that adhere to this same structure.

And this structure is the same across all families, just with different keys, branches, and runs.

It’s so clear, simple, and uniformly structured that it still completely baffles me—and, in fact, this rigid and highly uniform structure seems incompatible with any language. At least, I can’t imagine it; this language would have to be highly structured. Here’s the diagram again:

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

   

Of course, I don't know if that means you really have to decipher the VMS by looking at it from the back, but from that perspective, the structure becomes so uniform—that's what I find interesting.

[Jorge_Stolfi]

And exactly the same applies to the sh / k / ch phenomenon. The same dependency does not imply the same step-by-step selection structure. This refers to the linguistic level; when read backward, the long-range effect—although mathematically present—is easier to “explain” because it is cumulative—just as with “oq,” only here it spans multiple glyphs, yet still without a “jump.”When reading backward, the (as mentioned, existing) structure is not noticeable; it blends in with the other distributions of the glyphs. See the diagram:

I thought that I had understood your statistics, but now I am confused again.

That diagram seems to be a first-order Markov chain: a mathematical text generation device that can be described as a graph where each node ("state") is the last letter that was generated, and there is an arrow ("transition") from each state X to another state Y labeled with the probability (or observed frequency) that the letter Y occurs after the letter X.  Is that a correct description of your diagram?

Typically people will handle the word space SP as a letter.  Then the number in a transition from any state X to SP is the probability that the word will end after reaching the state X.  Conversely the number on a transition from SP to any state Y is the probability that a new word will begin with the letter Y.

But it seems that you are ignoring the spaces, so a transition from X to Y may be inside a token or across a word space, between two consecutive tokens.  How do you record which is which?  In the second case, the number on the transition would be the probability that a word that ends with the letter X will be followed by a word that begins with the letter Y.  Is that so?

Such a model can give information about two very different processes, but only for one of them will the information be reliable.  We can discuss  that after I understand what your model is, precisely.

All the best, --stolfi

[Jorge_Stolfi]

"below the overall average” and “more frequent than the corresponding ch-form” are not mutually exclusive. That is precisely why I compared forms with identical second parts. For all 21 sh/ch pairs examined, the qo-occurrence rate is higher on the sh side. Even shor can be below the overall average and still occur before qo more frequently than chor.

I am not sure I understood this second part, but such differences between the attraction to qo-words from Sh-words vs. Ch-words can probably be explained the same way.

Namely, certain Ch-words are attracted to certain qo-words too, and certain Ch-words are repelled by certain qo-words.  When you compute the general attraction of Ch- to qo-,  you are computing the average of all those attractions and repulsions over all Ch-words and all qo-words, weighted by the frequencies of those words.  

So you seem to be saying that the average Ch- to qo- attraction comes out positive but smaller than that of Sh- to qo-.

There is nothing strange in that result.  Such "anomalies" occur in every natural language text, due to the causes I described.  Before claiming that they are unique to Voynichese, you should check whether that is indeed true.  (See my advice #2.)

All the best, --stolfi

[JoJo_Jost]

@ Stolfi 

1. It’s interesting—and frustrating—how much we seem to be talking past each other. I don’t know how else to make this any clearer with those 21 examples. I’m giving up—which doesn’t mean I’m convinced (not at all)—but there’s no point in repeating the same thing again in different words. So let’s just leave it at that We just have different views—and why not?

[JoJo_Jost]

@ Stolfi,

2. These diagrams do not show a first-order Markov chain. Each percentage value refers to the entire sequence of glyphs tracked up to that point (oqyde—with a “folger”), not just the last glyph.

So the arrow from q to y asks: How often does y follow the sequence oq? And, of course, not: How often does y generally follow q?

The search takes place within the lines of the glyph stream, with spaces effectively skipped. In principle, this also captures the sought-after sequence within longer tokens—which, of course, is very rare after oq. But a standalone word is not assumed.

The displayed spaces are counted separately and represented by the gray bars. “Space 99%” means: In approximately 99% of cases, a space follows the sequence in question—regardless of which glyph follows next. The figure therefore does not refer exclusively to the continuation with y but is backward-looking.

The graph simply describes the observed continuations in the text—I can also display them with any other glyphs that follow a given glyphcluster (this is a very flexible program that works with any actual glyph sequence, both forward and backward).

   

[JoJo_Jost]

But I can add two more tables:

Distribution of shXXX words (exact tokens only):

   

And the same, but with possible prefixes:

   


It's interesting that, across so many different variations sh always most strongly draws out qo (almost—only once, in “o,” is k higher).

@ Stolfi, you describe how the numbers are derived from word-pair frequencies. But so far, this assumption doesn’t explain why the differences across so many different forms systematically point in the same direction.

That’s where we differ. And you say it’s possible that these are word pairs, while I say I think it’s unlikely.

That’s really all there is to it, and that’s where we disagree. Thats all 

But what’s much more important is whether this is possible in Chinese—and that brings us right back to your theory...


P.S. If you also take “t” into account, then in the endings shown above, ‘t’ is always lower than “sh”. Even with “p” (though there are very few such endings).

[Jorge_Stolfi]

Distribution of shXXX words (exact tokens only):

Sorry for insisting, but since you are claiming that your observations refute the Chinese Origin theory, I must make sure that I understand them, and check whether that is indeed the case.

I understood that you have two claims about the VMS. The first claim is that a Sh at the start of a word "attracts" (or "is attracted to"?)  a qo at the start of the next token.  The second claim is about the attraction of words that start with Ch compared with those that start with Sh; let's discuss this second claim in a later post.

I would rephrase your first claim as there is a positive correlation between event S0 -- a token starts with Sh -- and event Q1 -- the next token starts with qo. 

As noted in my reply to @chenxiang, about 15% of the words in the VMS start with qo.  Therefore, that claim is true for a given word ShXXX if more than 15% of its occurrences are followed by a qo-token.  If that fraction is less than 15%, the correlation is negative: the word ShXXX actually repels qo-words.

So, from the first table in your post, we can say that Shaiin, Shedy, Sheedy, Shey, Sheody and Sheol show positive correlation (attract the qo- prefix), while Shar, Shol, Shor and Shy have a slightly negative correlation (repel the qo- prefix).  Do you agree?

So it does not seem correct to say that "the Sh- prefix attracts the qo-prefix".  The attraction or aversion to qo is a property of some specific Sh-words, not of the Sh prefix.

And you also claim that, when we consider all Sh-words together, we see a net attraction.  That is, on the whole, a Sh-word is followed by a qo-word more than 15% of the time.  But that attraction is the weighted average of attractions and repulsions by the various Sh-words (including all those that are not listed in your tables).  The average is positive only because the Sh-words that attract qo are more numerous and/or have a stronger effect than the Sh-words that repel qo.  Do you agree?

Do you think that the observations so far are incompatible with the Chinese Origin theory or the SPS=SBJ claim?

I have just looked at the frequencies of the pairs "ShXXX YYYY" where XXX is edy, eedy, eey, ol and or, and YYY is anything. I see that the attraction or repulsion depends not only on the specific Sh-word, but on the specific qo- and non-qo-words that can follow that Sh-word.  For instance, while Sheol and Sheey have similar levels of qo-attraction (~37% by your table), the qo-words that can follow one are rather different than those that can follow the other.

I think I see some other interesting patterns in those counts, but they are too complicated and uncertain to discuss here.  Suffices to say that those counts seem to confirm my hunch that qo is a symbol that means "and", like our &, and is in fact a "barrier" for word-word correlations.

All the best, --stolfi

[JoJo_Jost]

No, I’m not claiming that it refutes the Chinese theory! I know too little about the Chinese language to make such a claim—I thought I’d made that clear enough in several places.

So I’m just interested in how these peculiarities fit with that. Since I know you don’t speak Chinese either, I let this question sit for a long time until "chenxiang" showed up—who does speak Chinese—and I basically asked him whether this “peculiarity” fits with Chinese. That’s how this whole thing came about. He said no, but I think we’d really need a Chinese linguist here, to clear ist. That’s why I wanted to let the discussion rest, but I admit that with my last post here, I undermined that myself—sorry about that… ?

I really don’t see the point in splitting this “sh” phenomenon into two separate ones. To me, it’s ONE phenomenon: namely, that words containing “sh” at the beginning or in the middle are more likely to be followed by “qo” than by “ch,” “k,” or “t.” I’m not at all concerned with whether they attract or repel “qo” as individual words—in my opinion, that distinction leads nowhere.

It’s simply a matter of a glyph at the beginning or in the middle of a token having an effect on the following token, and that this effect varies in strength among different glyphs (sh, ch, k, t) while the token structure remains otherwise the same. That alone would be interesting. But the truly crazy thing about it is that “sh” is significantly higher in all these tokens, and the distribution of the other glyphs also follows a pattern (the two exceptions may be related to the low n-values).

Whether individual forms lie above or below the overall average is simply irrelevant to this analysis.

This is an important and interesting phenomenon. And I think it’s hard to reconcile with a language. That’s why I’m interested in whether this could be the case in Chinese. It’s not really compatible with European languages. I’m not aware of such a rigid pattern in any language.

Of course, I understand that you don’t want to see it that way, because it’s difficult to explain in a single language and therefore possibly doesn’t fit with your Chinese theory. And I also understand that you want to downplay it by referring to “individual cases,” but that doesn’t work. You can’t make the phenomenon go away that way.

----

And if “qo” = “and,” how do you explain that “qo” is followed by a “k” 59% of the time and by a “t” 21% of the time—which together already add up to about 80 percent? That’s going to be quite difficult, too….isn't is?

   

[JoJo_Jost]

The discussion with Stolfi then continued in the Chinese thread, where it was also concluded for the time being. To get back to VBM:

I was able to establish the structures up to the “runs” (= eee / iii, etc.) very uniformly in the Reva reading (Reverse Eva) see here: You are not allowed to view links. Register or Login to view.

Now still missing the part that follows these runs. And there seemed to be quite a bit of confusion here. But I’m looking for a tabular format. And here’s a first rough draft.

The reverse reading shows what follows the word beginnings listed on the left side of the chart (for those who only know Eva, just read them as word endings and then, of course, backwards )

And here you can clearly see the difference between the e runs and the iian (aiin) runs, as well as ria (air) and lo ro.

The “hits” enclosed in the squares indicate the possible table. But, to be clear, this is all still a bit vague and rough.

But in my opinion, it is now possible to make out a fairly well-defined table structure in the VMS.

   

Note: Calculated on Herbal B. The selection criterion was purely based on volume: all states with at least about 190 tokens. There are a total of 142 states (key + d + run); the 18 shown contain 18,464 of the 20,344 tokens, or 91%. The remaining 124 states collectively account for 9% and individually have fewer than 190 instances (roe 155, ye variants such as yeee, loo, nii, saa, etc.), for which percentage values in a 22-column row would amount to nothing more than noise.

[JoJo_Jost]

I wasn't sure where to post this german Herb-book—but it bears similarities to the VMS in that the plants are barely recognizable and sometimes only very brief descriptions were written about each individual plant. Unfortunately, it is dated 1462, but the description of the book is interesting:


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

Summary

    Johannes Hartlieb’s Herb Book (Book of Herbs) from 1462 is essentially an excerpt from Konrad von Megenberg’s Book of Nature (Book of Nature), which was written a century earlier and was the first natural history work in the German language. Hartlieb’s focus is on plants—mostly herbs—and their medicinal uses. What makes the Herb Book special is the side-by-side presentation of text and images. The high cost of such a richly illustrated book makes it unlikely that it was actually used by physicians or apothecaries of the time. In any case, the botanical inaccuracies in the 160 illustrations would have made it difficult to identify specific plants in nature. The book also contains ten additional illustrations of animals that are not relevant to the subject of plants and medicine. These factors suggest that the book was created for representational purposes. Hartlieb (c. 1410–68) was the long-serving personal physician to Duke Albrecht III of Bavaria.