Statistical Validation of a Multilingual Decipherment of the Voynich Manuscript (Beinecke MS 408)

This record contains two documents: 1. This research paper presenting a deterministic multilingual decoder for the Voynich Manuscript (Beinecke MS 408, ca. 1404–1438 CE) and its statistical validation against five classes of scrambled and synthetic controls (1,400 total replicates). 2. The complete decoded translation of all 227 folios (35,881 tokens, 82.6% classified), produced by the v14 decoder pipeline. The decoder classifies tokens using a 14-tier priority cascade across 15 source languages, organized around a Hebrew proclitic grammatical framework. No machine learning or training data is used. The 15 source languages (Hebrew, Arabic, Malay/Old Javanese, Ge'ez, Amharic, Spanish, German, Occitan, Persian, Chinese, Sanskrit/Kawi, Turkish, Tamil, Swahili, and English) correspond to documented 15th-century trade routes between Europe, the Middle East, East Africa, and Southeast Asia. Statistical validation results (one-sided permutation p-values):- Token coverage: p < 0.001 vs. all controls- Strict vocabulary coverage: p < 0.001 vs. all controls- Grammar coherence: p = 0.008 vs. word-order shuffle (Cohen's d = 2.47)- Longest confirmed run: p < 0.001 vs. content-scrambled controls; p = 0.029 vs. word-order shuffle The grammar coherence test demonstrates that the manuscript's word order carries syntactic information consistent with Hebrew proclitic grammar. The decoded content describes a water engineering and agricultural reference compiled by a multilingual trade network. The decoder pipeline and control test harness are fully deterministic and reproducible.