Statistical Architecture in the Voynich Manuscript: A Phonetically Agnostic Structural Analysis

Statistical Architecture in the Voynich Manuscript:A Phonetically Agnostic Structural Analysis (Chapters 1–10) Author: Ren Kagawa / 架川 蓮 (ArchiFlow AI)Analysis Engine: Claude Sonnet 4.6 (Anthropic) + Codex (OpenAI, independent review)DOI: 10.5281/zenodo.20158064 | Analysis Period: May 11–17, 2026Format: Ten-chapter monograph (separate PDF per chapter) ───────────────────────────────────────────── OVERVIEW This ten-chapter monograph presents PASA (Phonetically Agnostic Structural Analysis), a newstatistical method for studying the Voynich Manuscript (Beinecke MS 408, c. 15th century)without any phonetic assumptions. Rather than asking "What sounds do the symbols represent?",PASA asks: "What statistical regularities in token distribution can be detected, and what dothey imply about the content categories being encoded?" The result is not a decipherment. No phonetic values are proposed, and no translation isoffered. The contribution is a proposed ontological content map: a statistically grounded,historically motivated, and falsifiable structural model of the manuscript's internalarchitecture. ───────────────────────────────────────────── THE CORE FINDING (Chapter 1) Two prefix-defined token classes show a statistically significant rank-order gradient acrossfive thematic sections: ok-class (e.g., okaiin, qokaiin) → proposed to encode Aristotelian substantia (substance) ot-class (e.g., otaiin, qotaiin) → proposed to encode passio/accidens (process/accident) Section ranking by ok/ot ratio: Pharmaceutical (3.27) > Biological (2.48) > Stars/Recipes (1.54) > Herbal (1.36) > Zodiac (0.67) Statistical result (confirmatory): Spearman ρ = 0.900 | exact one-sided p = 0.0417 | n = 5 Exploratory model-dependent: ρ = 1.000 | exact one-sided p = 0.0083 Intra-line positional asymmetry: ok-class tokens appear earlier in lines than ot-class tokens Mann-Whitney p = 4.39 × 10⁻⁴ (confirmatory) Both findings replicated across two independently constructed corpora: ZL3b (Zandbergen/Landini, 37,451 tokens) and corpus_clean.json (LSI, 37,394 tokens) ───────────────────────────────────────────── CHAPTER CONTENTS Ch. 1 — Foundations ok-/ot- binary established; PASA method introduced; five-point convergence (statistical, scholastic text comparison, phonological, astronomical, iconographic) Ch. 2 — Aristotelian Category System 8 of 10 Aristotelian categories identified in EVA prefix/suffix system: ok- (substantia ★★★★★), ot- (passio ★★★★★), yk- (relatio ★★★★), oe- (tempus ★★★★★), yt- (qualitas ★★★), ch- (actio ★★★ candidate), -am (habitus ★★★), -al/-ar (ubi ★★★) Intra-line ordering: oe- < yk- < yt- < ok- < ch- < ot- < -am (mirrors Galenic prescription sequence: time → relation → quality → substance → action → process) Ch. 3 — Galenic Humoral Physiology Nine organ/humor roots identified in the Biological section: qol (cor/heart ★★★), kedy (hepar/liver ★★★), tal (lien/spleen ★★★), lol (vena/vein ★★★), rol (pulmones/lungs ★★★), okar (arteria/artery ★★★), olkain (sanguis/blood ★★★), olkedy (bilis/bile ★★★), shol (phlegma/phlegm ★★) Key structural finding: olkedy = ol + kedy (bile root contains liver root) — transparent morphological encoding of the liver–bile production relationship in the Galenic system Ch. 4 — Planetary Astrology Seven-planet candidate roots proposed for the Stars section: otar (Mercury/Utarid ★★★), lor (Moon/Luna ★★), shes (Sun/Shams ★★ candidate), otair (Mars/Al-Qahir ★★), shal (Saturn/Zuhal ★★), shear (Venus/Zuhara ★★), chal/char (Jupiter/Mushtari ★★) Planet-eye co-occurrence significant for Mercury (p < 0.001 ★★★) and Moon (p = 0.002 ★★), consistent with medieval astromedical tradition Ch. 5 — Lunar Mansion System Folio f69v proposed as a candidate 28-mansion classification table (★★★): 28 short label lines (1–2 tokens each) in a 31-line folio = 28 Arabic lunar mansions Single oe- label (harvest/seasonal marker) consistent with seasonal dimension of mansion system Highest yk- density in the corpus found in preamble line (7/32 tokens = 21.9% yk-), consistent with a relational correspondence table Ch. 6 — Three-Layer Lexical Model Three-layer lexical origin proposed: (1) Latin formal case morphology (76% shared paradigm rate), (2) Arabic-via-Latin technical vocabulary (aiin = ʿAyn, otar = al-Tarf, okal = Al-Baldah via Toledo Translation movement), (3) North Italian/Occitan vernacular substrate (speculative) "Knowledge environment hypothesis": the manuscript's apparent polysemy reflects the unified Aristotelian ontological language used across medicine, astronomy, agriculture, and philosophy in 14th–15th century North Italy — not a uniquely designed system Ch. 7 — Prescription Formula Architecture Six-slot PFA (Prescription Formula Architecture) identified: S0 daiin = vide (★★★★★) → S1 ot- (accidens/symptom ★★★★) → S2 ok- (substantia/remedy ★★★★★) → S3 -edy (participial completion ★★★★) → S4 absent (null-result: practical text omits effect confirmation) → S1b aiin line-final = ad oculos (★★★ candidate) Three-text convergence: Pseudo-Apuleius Herbarius, Galenic corpus, and Avicenna's Canon independently support the same slot assignments Two-paradigm model: P-type (Pharmaceutical/Herbal, Currier A, ot-first) vs T-type (Recipes/Zodiac/Biological, Currier B, ok-first) confirmed by Fisher's exact p = 0.027 Ch. 8 — Full Morphological Design Five implementation forms of the Aristotelian category system: (1) Prefix (6 categories), (2) case suffix ubi (-ar/-al), (3) predicative suffix habitus (-am), (4) independent directional root (or-), (5) repetition/iconographic (quantitas) Phonotactic three-layer constraint: q at word-initial 98.9%, n at word-final 97.3%, ii at word-internal ~100% — consistent with the prefix-root-suffix template Complete 10-category implementation map: 5 categories at ★★★★ or above, 3 at ★★★ candidate, 1 (quantitas) with non-morphemic proposals, 1 (situs) unresolved Ch. 9 — Epistemological Limits Two-tier confidence system articulated: Tier 1 (statistical): reproducible from ZL3b corpus without interpretive framework Tier 2 (interpretive): requires independent historical and cross-textual validation Full falsification table for all major claims AI-assistance limitation acknowledged: statistical findings are AI-executable and independently verifiable; interpretive assignments require human historical judgment Requirements for elevating interpretive claims to Tier 1 identified Ch. 10 — Context Inference Method, Four-Layer Lexical Model, and Structural Gloss Programme [NEW — added May 17, 2026] Three contributions extending the PASA framework: (1) Context Inference Method (CIM) — formal definition CIM is a six-step procedure for identifying the functional category of an unknown token on the basis of its positional relationship to a high-confidence anchor token. It emerged inductively during Stage 2 analysis and has no direct precedent in existing Voynich scholarship. The procedure: select anchor (★★★★+) → exhaustive positional survey → functional inference → comparative text verification → three-node chain confirmation → confidence assignment (ceiling ★★★ from CIM alone; ★★★★ requires additional PASA support). Key CIM results: shedy = instillatum ★★★ (n=26 post-anchor, ×23 three-node chains); chedy = decoctum ★★★; choraiin = region-indicator token ★★★★★ (f72v2 Virgo decisive). (2) Four-Layer Lexical Model — revision of Chapter 6 The three-layer model is revised to four layers by adding Layer 0: a Greek philosophical conceptual layer (Platonic/Aristotelian) functioning as the common ontological substrate underlying Layers 1–3. Layer 0 (Greek conceptual): ok- = eidos/substantia ★★★★★ | ot- = kinēsis/passio ★★★★★ oe- = chronos/tempus ★★★★★ | chor = chōra/regio ★★★★ (Tier 2) yk- = logos/relatio ★★★★ | yt- = poion/qualitas ★★★ (Tier 2) Confidence calibration: distributional non-randomness is Tier 1; the Layer 0 model as a whole is Tier 2 (★★★) — a historically plausible interpretive framework, not a confirmed fact. The four ★★★★★-strength roots (ok-/ot-/oe-/chor) map onto Aristotle's four primary categories of natural description (substance / process / time / region), consistent with the scholastic knowledge environment of 14th–15th century North Italy. Tetrabiblos note: χώρα / χωρῶν appears as relevant vocabulary for regions and countries under celestial influence in the Ptolemaic tradition (Tetrabiblos Book II ch. 3; Robbins Loeb LCL 435, p. 115, fn. 5). Ptolemy's Geography Book I ch. 1 formally defines chorographia as the study of regional subdivisions. (3) Structural Gloss Programme — L2 completion for all 12 Zodiac folios A structural gloss assigns functional category labels to tokens; it is explicitly distinguished from translation. The four-level programme: L1 (token-level) ★★★ / L2 (line-level slot parsing) ★★★ / L3 (paragraph-level inference) ★★ partial / L4 (section-level reading) ★ not yet begun. L2 is now complete for all 12 Zodiac folios (1,322 tokens; 63.7% parse rate): ok/ot ratio range: 0.263 (Taurus) – 1.091 (Sagittarius) NAV total: 55 (okal 27 / otar 15 / otaiin 13; Virgo NAV=15 highest) Key cross-folio findings: · oe- (TMP) concentrates in autumn folios: OR=23.55, p=1.26×10⁻⁸ (Fisher exact, Bonferroni-corrected 6.30×10⁻⁸) — independent corroboration of Ch.5 identification · Sagittarius: only ok-dominant folio (ok/ot=1.091) ★★★ (single folio; requires replication) · Capricorn: only folio with HUM=0 ★★ (exploratory; requires corpus-wide verification) · Virgo (f72r2): NAV=15 (highest); S0-NAV rate 25%; okal-specialised · f72v2 (Virgo_b ★★★★★): choraiin in line 1 — decisive CIM evidence for chor=chōra Scope note: Chapter 10 does not translate any passage. All structural glosses assign functional category labels within the PASA framework; no phonetic values are prop