Hello everyone,
I’ve been examining the statistical behavior of recurring token patterns in the Voynich manuscript using an EVA transcription. The focus here is on tokens of the form
cheXdy and more generally suffixes such as
-ody /
-dy.
This is not an attempt at decipherment or interpretation, but rather a statistical and positional analysis of token behavior.
1. Non-equivalence of similar forms
Forms such as:
- chedy
- cheedy / cheekeody
- cheody
are often treated as variants, but their contextual distributions differ significantly.
A neighborhood overlap test between
chedy and
cheody yields ≈ 0.09, suggesting strong contextual separation rather than free variation.
2. Internal variation (cheXdy)
The internal segment X (e, ee, o) appears to correlate with different contextual neighborhoods:
- chedy / cheedy → dense network (qok-, qol-, qot-)
- cheody → more dispersed and distinct context
This suggests that
cheXdy forms are not interchangeable and may have distinct structural roles.
3. Distribution of -ody forms
Tokens ending in
-ody:
- appear in local clusters
- sometimes repeat multiple times within a single line
This clustering suggests a non-random distribution, possibly reflecting a localized structural mode.
4. Regime variation across sections
Comparison between two folios:
- f57r (herbal section):
→ more fluid structure
→ longer forms (e.g. cheekeody)
→ internal variation within tokens
- f70v (circular diagrams):
→ more modular structure
→ shorter forms (e.g. okody, otody, dy)
→ higher repetition
This suggests that different sections may follow different generative regimes.
5. Positional behavior of -dy / -ody
Preliminary observations on f70v:
- In longer lines:
→ -dy / -ody appears mostly in middle positions
→ possible internal or compositional role
- In shorter lines:
→ -dy / -ody appears more often at the end
→ possible closing or terminal function
6. Working hypothesis
The manuscript may reflect a constrained generative system, where:
- some tokens act as structural operators
- others carry variable content
- behavior depends on position and section context
This could be compatible with:
- a context-sensitive grammar
- or a finite-state system with constraints
7. Open questions- Is the positional behavior of -dy / -ody consistent across other folios?
- Do -ody clusters correspond to specific functional zones?
- Can these patterns be modeled with a constrained Markov process?
I would be very interested in feedback, especially from people who have worked on positional statistics or EVA-based analyses.