Discussion of "A possible generating algorithm of the Voynich manuscript"

[nickpelling]

To be precise, EVA is neither a pure stroke transcription nor a pure glyph transcription. *Sigh* It's actually somewhere in the middle, where some component shapes are glyphs and some are strokes.

But even though verbose cipher glyph pairs make sense to my mind, I really struggle to see how such strongly-structured super-common pairs can be explained linguistically. Even though it is so visually tempting to read EVA o and a as vowels (it's why the EVA letters were chosen, after all), the strong pairing is just too mechanical, too unlinguistic to work.

What, then, of Torsten's ligatures? Some, like EVA qo, won't raise many objections. EVA ch and sh were designed in, along with c-gallows-h, so no problem there either. But ar / al / or / ol / am aren't so straightforward: because if you follow these through to their logical conclusion (as I tried to do in Curse), you have to also read o- +g as similar pairs: and from there to y- + gallows. At which point pretty much everything becomes strongly paired.

More generally, the or / ar / ol / al ligature group seems visually inconsistent with the an / ain / aiin / aiiin group. In particular, given that r starts with exactly the same linear downstroke as i, there really should be no stroke harmony adjacency behaviour difference between their usage. But the statistics shout otherwise very loudly. How is it that the hypothetical autocopyist was able to maintain that statistical consistency over so many pages of text, and yet evolve the patterns of usage (from Currier A to B) so subtly?

But even though I can see why it is visually appealing to imagine that EVA -dy can be freely substituted for EVA -dy, it is oddly asymmetric that EVA -dy is almost never substituted for EVA -d . Similarly, if EVA d- starts words with a curve, why do we see EVA dy- so rarely? Even something as apparently obvious as -dy hides a wealth of behaviors that are all linked together.