Papers by Guillaume CLEMENT
Computational Codicology Series, 2026
This paper presents a formal statistical characterization of the structural properties of the Voy... more This paper presents a formal statistical characterization of the structural properties of the Voynich Manuscript (Beinecke MS 408), focusing on the internal logic of the Pharmaceutical and Herbal sections. We identify a set of robust computational constraints—including a rigid prefix-root-suffix morphology and a systematic "herbal-to-pharma" substring composition system—that define the manuscript’s unique textual architecture.
These regularities are evaluated against the hypothesis of a personalized Tironian shorthand system. While the observed patterns are shown to be theoretically compatible with such a framework of professional abbreviated notation, we present this as a heuristic model for structural comparison rather than a definitive linguistic identification. Furthermore, we document several candidate ingredient strings that, while currently below the threshold for statistical certainty, provide necessary test cases for future cross-corpus validation. By establishing these data-driven constraints, this work aims to narrow the search space for the manuscript’s functional nature.
Toward a Phonetic Decryption of the Voynich Manuscript, 2026
Version 1 of this study presented a computational pipeline for Latin phonetic recovery of the Voy... more Version 1 of this study presented a computational pipeline for Latin phonetic recovery of the Voynich Manuscript (Beinecke MS 408), claiming an 89.3% validation rate via the Perseus Digital Library and the identification of 33 pharmaceutical terms.
This Version 2 documents a critical methodological bias that invalidates those initial claims. Analysis reveals that the sheer scale of the Perseus dictionary (exceeding 265,000 entries) allows for the statistical validation of almost any Latin-sounding syllable sequence, creating a "mirage" effect of linguistic coherence. Furthermore, the use of permutation testing demonstrates that the previously observed correlation with the Antidotarium Nicolai is a statistical artifact of Zipf’s Law rather than evidence of genuine semantic content.
While the codicological and structural observations of the first version remain valid, the phonetic decryption claims are hereby formally retracted. This corrective emphasizes the necessity of rigorous cross-validation and independent data testing in Voynich research. It redirects future efforts toward a purely structural analysis of the text (published separately, Clement 2026b 10.5281/zenodo.19543917), which interprets the manuscript as a personalized system of Tironian shorthand.
Uploads
Papers by Guillaume CLEMENT
These regularities are evaluated against the hypothesis of a personalized Tironian shorthand system. While the observed patterns are shown to be theoretically compatible with such a framework of professional abbreviated notation, we present this as a heuristic model for structural comparison rather than a definitive linguistic identification. Furthermore, we document several candidate ingredient strings that, while currently below the threshold for statistical certainty, provide necessary test cases for future cross-corpus validation. By establishing these data-driven constraints, this work aims to narrow the search space for the manuscript’s functional nature.
This Version 2 documents a critical methodological bias that invalidates those initial claims. Analysis reveals that the sheer scale of the Perseus dictionary (exceeding 265,000 entries) allows for the statistical validation of almost any Latin-sounding syllable sequence, creating a "mirage" effect of linguistic coherence. Furthermore, the use of permutation testing demonstrates that the previously observed correlation with the Antidotarium Nicolai is a statistical artifact of Zipf’s Law rather than evidence of genuine semantic content.
While the codicological and structural observations of the first version remain valid, the phonetic decryption claims are hereby formally retracted. This corrective emphasizes the necessity of rigorous cross-validation and independent data testing in Voynich research. It redirects future efforts toward a purely structural analysis of the text (published separately, Clement 2026b 10.5281/zenodo.19543917), which interprets the manuscript as a personalized system of Tironian shorthand.