Two Scribes, Two Vowel Dialects.

[Brain-V]

Following on from my first post, Brain-V has converged on a specific finding I want to put to the community.

The question: does the Voynich encode a plant / non-plant distinction, and is the signal mechanical or intentional?

What the data shows:
Hand A marks plant folios with _.oii, a 4.84× enrichment over non-plant herbal folios, stable across 5-fold CV.
The pattern fires at 0.566% on Hand A plant folios and zero on Hand B plant folios.

Hand B marks plant folios with _.e.e, _.eeo, and _.o.eo, a combined rate 1.63% on plant folios vs 0.00% on non-plant herbal folios, stable across 5-fold CV.
The two hands share exactly one pattern in their respective top-10 plant-enriched sets. Hand A markers are o-vowel heavy; Hand B markers are e-vowel heavy. Same function, completely different vowel vocabulary.

On the volvelle question, given the discussions here this seems worth stating directly, I ran 1,600+ null runs across four volvelle architectures (section-level and folio-level cartridges, 26 to 500 root sizes, simple template and corpus-derived roots). None reproduced either hand's enrichment. Empirical p < 0.01 across all variants for both hands.

This isn't a decipherment claim.. It's a narrow claim: two scribes consistently mark the same folio category using different vowel conventions, and no mechanical generator tested can produce that signal.

Full preprint with methodology, CV results, volvelle null distributions, and all code: https://github.com/BuilderBenv1/brain-v
Live dashboard (all hypotheses including failures): https://brain-v-beryl.vercel.app
Limitations are in section 5 of the preprint — the non-plant herbal baseline is thin (n=7/8 folios per hand) which is the main weakness. Happy to discuss.