Evidence of Layered Positional and Directional Constraints in the Voynich Manuscript: Implications for Cipher-Like Structure

A new computational linguistics study by researcher Christophe Parisel has applied systematic AI-driven analysis to the enigmatic Voynich Manuscript (VMS), yielding the first quantitative evidence of its unique structural constraints. The research, published on arXiv, reveals the manuscript's script operates with a two-layer directional system: character sequences within words are optimized right-to-left, while dependencies at word boundaries flow left-to-right. This specific directional dissociation was not found in any of the four natural languages used for comparison—English, French, Hebrew, and Arabic—marking it as a statistically anomalous pattern.

To test potential origins, the study evaluated two classes of structured text generators against a joint criterion of four statistical signatures. A parametric slot-based generator and a Cardan grille model (implementing Gordon Rugg's 2004 'gibberish hypothesis') were both tested across their full parameter spaces. Neither class of model could reproduce all four of the VMS's structural signatures simultaneously. While this doesn't rule out untested generative models, it provides the first concrete, data-driven benchmarks against which any future cryptanalytic or generative theory of the VMS must be measured. The findings strongly suggest the manuscript's text is governed by sophisticated, cipher-like positional constraints that are not easily explained by simple frequency-based mechanisms or hoax-generation techniques.