Robust Positional Constraints in the Voynich Manuscript: A Computational and Statistical Analysis

The Voynich Manuscript has long resisted interpretation, with ongoing debate as to whether its text encodes language, cipher, or non-linguistic structure. This study presents a computational and statistical analysis of positional constraints in the Zandbergen–Landini EVA transcription (ZL3b). Three pre-specified metrics are evaluated: line-initial token bias, paragraph-initial prefix concentration, and line-final token bias. The analysis employs multiple permutation-based null models, including global, line-internal, and position-restricted shuffling. Across 1, 000 permutations per model, all three metrics remain highly significant (empirical p = 0. 000999 in every case), indicating strong positional constraints that cannot be explained by random token placement. These findings establish robust structural properties in the Voynich Manuscript and provide quantitative constraints on any proposed generative model.