Can LAAFU effects be modeled?

[magnesium]

On the other hand, I can imagine a system that would predictably produce line effects as a natural byproduct of its use, and that also falls well within the range of hypotheses people already entertain about how Voynichese might have worked (along the lines of Rene's You are not allowed to view links. Register or Login to view.).  Consider this set of specifications:

(1) Lines always break at word boundaries.

(2) Within lines, words are run together indiscriminately.

(3) Text is chunked for encoding into units consisting of one or more consonants followed by one or more vowels, with each "chunk" being encoded as a vord.

(4) It's possible to encode an isolated consonant or vowel (or isolated clusters of either), but this is done only as needed to satisfy rule (1).

(5) Vords that are similarly structured represent similarly structured "chunks," but not in a straightforward letter-by-letter way (imagine something like Naibbe encoding tables not being randomly interchangeable, but each encoding a different category of "chunk").

To keep this conversation going, I have come up with a quick modification of the Naibbe cipher that partially implements pfeaster's ideas. 

1. Lines always break at word boundaries.
2. Respacing proceeds deterministically, such that plaintext bigrams are consonant-vowel, and everything else is a unigram.
3. When possible, encrypted bigrams are reused on successive lines.

The cipher is otherwise identical to what I have previously described. I have encrypted Book 16 of Pliny's Natural History and the Divina Commedia this way:

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

Also note a new feature: The "respaced" versions randomly take out 3% of spaces from the idealized ciphertext, which is meant to simulate ambiguous spacing. This also has the effect of generating a small proportion of 3-grams and 4-grams, which is very handy for generating lots of word types efficiently.

I have not yet performed any pfeaster-style analysis of placement-dependent glyph variation, nor do I have time to do so over the next few days. But I wanted to get this over to you all as soon as I could.

I want to quickly mention pfeaster's Point 5, the structurally similar chunks. I agree that this would be ideal; this is why I set the Naibbe cipher's unigram word type assignments to do just this. But when I was creating the Naibbe cipher, this struck me as a challenging problem that bled into a decryption attempt. So instead, I opted to make some initial prefix and suffix assignments so that I could proceed directly to characterization of the cipher's behavior.