The Voynich Ninja

Hello everyone,
I’m new here. I’m not a linguist or cryptographer, my background is not academic.
This is not a decipherment attempt. What follows is just a structural observation based on pattern-recognition and basic morphological reasoning.
I’ve been looking at Voynichese from a pattern-recognition perspective: not as encoded phonemes, but as possible morphological units.

What caught my attention is that several recurring sequences behave more like morphemes (in the typological sense) than like components of a substitution cipher.
Many analyses have noted that Voynichese has:

• lower entropy than a typical monoalphabetic substitution,
  
• very stable internal word patterns,
  
• recurring sequences such as qo-, che-, -dy, -iin,
  
• and clear positional constraints on certain glyph groups.
  

Instead of treating these as “oddities” of an underlying phonemic text, I wondered what happens if we flip the model:

Core idea (structural, not semantic)
What if Voynichese doesn't encode letters or phonemes at all, but behaves more like a morphological system, where each "word" is a bundle of morpheme-like units (Prefix + stem + suffix), somewhat analogous to polysynthetic or strongly agglutinative languages?

Voynichese “words” may function as semantic/morphological bundles, similar to polysynthetic or highly agglutinative systems, rather than representing a letter-based encoding.
This idea might help explain some well-known features:
• very stable internal structure in many tokens
• frequent recurring sequences (qo-, che-, -dy, -iin)
• strong positional constraints
• low entropy inconsistent with simple substitution
• vocabulary shifts across sections
To illustrate the structural idea (not the semantics), here are a few examples in EVA:

1. The qoke- family: qokedy, qokeedy, qokain, qokaiin, qokal
These share:

• Initial element: qo-
  
• Stem-like core: k(e/a)
  
• Variable endings: -edy / -dy / -ain / -aiin / -al
  

This pattern resembles a fixed stem with multiple aspect/state suffixes, common in polysynthetic morphology, where endings encode nuances like iteration, completion, plurality, state, etc.
Even without knowing the semantics, the structure is consistent.
2. The -hedy cluster: shedy, chedy, ychedy, lchedy, okedy (overlapping pattern)
These share:

• a stable -hed- / -ched- / -ked- type stem
  
• variable onsets (s-, ch-, y-, l-, o-)
  
• a highly stable final element: -dy
  

In morphology, this is classic behavior of a productive suffix attaching to multiple stems.

• Position: final
  
• Stability: invariant
  
• Distribution: high frequency
  
• Combination: attaches widely
  

These are foundational criteria for morpheme identification in unknown languages.

3. The ol–olol–olkeeody family
These show:

• ol
  
• olol (reduplication-like extension)
  
• olkeeody (stem expansion + final suffix)
  

Reduplication and recursive stem-building are common in polysynthetic and agglutinative languages but are unusual in substitution ciphers unless artificially engineered.
The recurrence and structure again suggest morphological productivity.

Why polysynthetic-like?
Not because Voynichese is one of those languages, but because:

• tokens behave like semantic bundles rather than phoneme sequences
  
• many stems appear to be non-phonotactic but internally consistent
  
• affix-like sequences have clear positional rigidity
  
• the writing flow looks natural for whole morphological units
  

This shifts the analytic model away from:
encoded alphabet → encoded syllables → encoded phonemes
and toward:
prefix (class/process marker) + stem (core process/state) + suffix (aspect/iteration/state)
A pseudo-polysynthetic system could be invented, constructed, or hybrid, the origin doesn’t affect the structural behavior.

What this is NOT:

• Not an argument about semantics
  
• Not claiming the text is natural language
  
• Not claiming decipherment
  
• Not pushing a specific meaning system
  

Just a structural possibility that might be testable through:

1. morpheme segmentation algorithms
   
2. co-occurrence analysis
   
3. positional modeling
   
4. affix-grammar approaches
   
5. typological comparison (Eskimo-Aleut, Algonquian, Chukotko-Kamchatkan, etc.)
   

If such analysis has already been attempted, I’d appreciate pointers.
If not, maybe this model offers another angle for people working on statistical or computational methods.

Thanks for reading and for any thoughts.
 Astra Lumen

Hi Astra.  Welcome to the forum but we have a clear rule that we cannot accept theories built with the help of an LLM like ChatGPT or Claude.

(03-12-2025, 11:58 PM)tavie Wrote: You are not allowed to view links. Register or Login to view.Hi Astra.  Welcome to the forum but we have a clear rule that we cannot accept theories built with the help of an LLM like ChatGPT or Claude.

Hi tavie, thanks for the note and no worries, the idea and analysis are entirely my own. I didn't use Chat GPT or any LLM to generate a theory, the post reflects my personal observation and pattern-recognition approach while reading about morphological system. If the framing was too polished, that's on me! Since English is not my first language sometimes I draft in my native language and run parts of the text through Google Translate or an online dictionary to make the phrasing clearer. That might be why structure looked polished.
If anything needs rephrasing to fit forum standards, I'm happy to adjust.

Sorry but this all looks extremely like the output of an LLM, not Google translate or a dictionary.  We have members who use a translator, and it doesn't output at all like yours does.  

The forum can be used to share thoughts on things like apparent suffixes - many people have commented on Voynichese word structure - but they cannot be the product of an LLM.