Conditional Structure and Generative Dependencies in Voynich Tokens: A Prefix–Core–Suffix Analysis

This record presents the study: "Conditional Structure and Generative Dependencies in Voynich Tokens: A Prefix–Core–Suffix Analysis" This work investigates the internal generative structure of Voynich tokens using a prefix–core–suffix decomposition framework. Each token is segmented using a strictly positional rule. The term "core" is intentionally used instead of "stem" to avoid any linguistic or semantic assumptions; it does not imply a morphological or semantic unit, but refers only to the middle segment defined by position. The study evaluates conditional dependencies between token components through entropy measures, prediction accuracy, and controlled structure-breaking experiments. The results demonstrate a strong asymmetric dependency structure, where suffix selection is highly predictable from the core, while the prefix provides weaker constraints. While not constituting a definitive generative model, these results provide strong evidence against independent or random formation mechanisms. These findings extend prior work on structural constraints by providing direct evidence of cross-component dependencies, suggesting that the observed patterns are unlikely to arise from independent processes. Related preprints: Chang (2026a), Robust Positional Constraints in the Voynich Manuscript Chang (2026b), Language-like Predictability and Structural Divergence in Voynich Tokens A reproducibility package (software) is available at:https://doi.org/10.5281/zenodo.19765403 Note:The Zandbergen–Landini EVA transcription (ZL3b) is not redistributed. Users should obtain it from the original source:https://www.voynich.nu/