A 600-year-old undeciphered manuscript. One prefix family — qo- — whose middle consonant tells you which dimension you're in. χ² = 348, p < 10⁻⁵⁷. Not a cipher. A notation system.
The Voynich manuscript is a 240-page illustrated codex, written in an unknown script, carbon-dated to the early 15th century. Over a hundred years of analysis by cryptographers, linguists, and statisticians has produced no consensus on what language it encodes, or whether it encodes language at all.
What everyone agrees on: the text is not random. It has word-level statistics that look like natural language. It has section structure — botanical diagrams, astronomical charts, biological baths, pharmaceutical recipes, and dense star-text pages. Vocabulary shifts systematically between sections.
The framework maps the manuscript's five major sections to specific positions in (O, R, α) space. Each section encodes information about one axis of a constraint notation system.
11,000 tokens. The void-object catalog — each botanical entry is a Pe=25 void configuration. Maximum opacity. qo+tch and qo+kch cluster here.
2,962 tokens. Cyclical zodiac notation. Pe=0.8. 93% of qo- word types have zero Astronomical tokens. A coordinate notation cannot express periodic patterns.
6,769 tokens. Flow topology diagrams — bathing figures in connecting pools. qo+lch dominates here at 84%. The coupling axis is the most specific dimensional assignment in the manuscript.
10,582 tokens. Dense prose without illustrations. Pe=0.1. qo+pch clusters here at 67% — the specification-complete notation for V* coordinates.
In a qo- word, the consonant cluster between the prefix and the root tells you which axis the following value is specifying. This is a grammar — a structural rule, not a statistical tendency.
| Medial | Dominant Section | Dimension | % in Section (ZL) | Takahashi | Enrichment | Concentration |
|---|---|---|---|---|---|---|
| qo+lch | Biological | α | 84.4% | 82.8% | 2.27× | |
| qo+lk | Biological | α | 76.0% | 74.5% | 2.05× | |
| qo+tchy | Herbal_A | O | 76.0% | 77.8% | 2.04× | |
| qo+pch | Stars_Text | V* | 67.2% | 65.6% | 2.23× | |
| qo+tch | Herbal_A | O | 54.1% | 57.8% | 1.73× | |
| qo+ckh | Pharmaceutical | constraint | 44.3% | 42.6% | 1.37× | |
| qo+[any] | Astronomical | R | 1.5% — structural exclusion. No qo- medial reaches Astronomical. | |||
Replicated in both independent transcriptions (ZL3b and Takahashi IT2a) within 3%. Cross-transcription Jaccard similarity: 0.7633 — 619 of 716 qo- word types shared between two independent transcribers who never coordinated on interpretation.
The most structurally significant finding is not what qo- does — it's what it doesn't do. The Astronomical section (f67–f73), with Pe=0.8, is the closest section to the constraint pole (Pe=0.1) on a simple number line. And yet:
The chi-squared result (χ²=348.40, df=28, p=5.24×10⁻⁵⁷) tests one question: is the association between the medial consonant cluster and the dominant manuscript section due to chance? No. Not by 57 orders of magnitude.
No control text produces equivalent morphological grammar. Dante Divine Comedy peaks at 38.4% section exclusivity. Pride and Prejudice at 36.2%. The Voynich qo+lch achieves 84.4% and is replicated exactly in a second independent transliteration.