Why and how the text could be Bavarian

[JoJo_Jost]

Thank you for this demonstration, however for a statistical test it will take a reasonable amount of encoded text (ideally the size of the actual manuscript) to be able to compare the results to Voynichese. 

Note that even your short example is not very Voynich-like, with sequences like edrol, chkoaiin, opedy, that don't appear in the manuscript, even though they seem to correspond to normal frequent words and combinations of characters, as far as I can see. So far I'm under impression that this method might collapse very quickly when applied to a longer text. I'm not sure I understand how you derive the claim that "The Vowel Bridge Model fits remarkably well into the statistical peculiarities of the VMS".

You're not serious, are you?

1. If you ask for a preliminary forward test at such an early stage, and then immediately argue that only a manuscript-length statistical test would count, why ask for the preliminary test in the first place?

2. A forward test at this stage obviously requires provisional assumptions. If one of them is wrong, imperfect segments are exactly what one should expect. Treating that as a refutation of the model is a very weak and unfair argument.

The ff(C1) -> p assumption illustrates exactly this point. I explicitly marked it as extremely risky because it does not fit the frequency analysis well. So if non-Voynich-like structures appear around p in this example, that is not surprising. It shows the weakness of that particular provisional value, not of the model as a whole.

Frankly, I was more surprised by how much of the generated line looked structurally close to Voynichese than by the fact that a few segments did not.

3. Even the plaintext spelling is not fixed in MHD / FNHD. If the scribe wrote “or” (ear) as “ohr,” then the word “or” would not exist in the underlying text and therefore cannot be expected to have a direct VMS counterpart. The same applies to “minzen,” which could have been written as “minze.” So criticizing the generated output as if every normalized plaintext word must have a direct visible VMS equivalent is methodologically unsafe. 

4. One point you have clearly overlooked is this: at the level of argument you are using, even producing 200 pages would not prove the model. The VMS is clearly based on a slot-like system, as Stolfi has shown. If one simply maps frequent plaintext elements onto frequent Voynich glyphs, bigrams, and trigrams within that slot system, a Voynich-like surface will inevitably emerge. Treating that as proof would be circular.

----
For me, the forward test did exactly what it was supposed to do: it showed that the proposed structure is operational. In other words, you can actually use the VBM rules to transform a German medical sentence into a Voynich-like surface. 
That does not prove the model. I never claimed that it did. Anything beyond that would be an exaggeration.

But using a few imperfect segments from a first provisional forward test as a refutation of the claim that the VBM fits the statistical peculiarities of the VMS is simply bad methodology! It does not engage with the actual statistical argument; it only shows that parts of the current forward table are still incomplete.

Sorry, oshfdk, but please stay fair.