Multi-Level Statistical Constraints in the Voynich Manuscript: From Token Structure to Sequential Organization

This preprint presents an integrated multi-level statistical analysis of the Voynich Manuscript. It synthesizes five component studies evaluating positional constraints, comparative predictability, prefix-core-suffix conditional dependency, PCS token-generation modeling, and local inter-token arrangement. The results support the interpretation of the Voynich text as a multi-level structured symbolic system: token placement is position-sensitive, token predictability is non-random, suffix selection is strongly conditioned by the operationally defined core, a PCS model reproduces observed token forms at high matching and coverage rates, and local token sequences show constrained inter-token organization. The study does not propose decipherment, translation, source-language identification, or confirmed syntax. Instead, it establishes a reproducible structural baseline that future linguistic, cryptographic, algorithmic, or generative explanations of the manuscript should address. The associated integrated reproduction package is archived separately on Zenodo at https://doi.org/10.5281/zenodo.19876003. This paper does not claim decipherment, translation, or confirmed syntax.