The Voynich 2026 Conference — New Methodologies Meet an Ancient Enigma
The first International Conference on the Voynich Manuscript is convening in 2026, hosted at the University of Malta and organized under the auspices of the Medieval Academy of America. It is the most significant scholarly convergence around the 600-year-old enigma since vellum dating established its 1404–1438 production window.
Key Facts
- Organizers: Dr. Colin Layfield (Senior Lecturer, University of Malta), Prof. John Abela (Associate Professor, University of Malta), with Prof. Claire Bowern (Yale University, linguistics) and Prof. Lisa Fagin Davis (Medieval Academy Executive Director) as senior advisors
- Submission process: Double-blind peer review; 750-word abstracts via EasyChair
- Institution: www.um.edu.mt/events/voynich2026/ | [email protected]
- First major interdisciplinary convening around the manuscript — unprecedented scope
Research Tracks Invited
The conference is explicitly soliciting work from four domains, each representing a distinct methodological lens:
- Historical and codicological — physical manuscript analysis, provenance, cipher-history context
- Image processing — multispectral imaging, folio recto/verso correspondence, ink analysis
- NLP and computational linguistics — word-structure analysis, language family classification, statistical fingerprinting
- AI and machine learning — LLM-based decipherment attempts, clustering, embedding approaches
This convergence reflects the unsatisfying state of individual-discipline attacks: no single methodology has been sufficient.
What New Methodologies Are Actually Showing
The Natural Language Debate (2025)
A May 2025 structural analysis (reported by BoingBoing, peer-reviewed preprint) used suffix-stripping and SBERT embeddings to reveal something unexpected: Voynichese word formation follows structured, natural-language-like patterns with identifiable prefix–root–suffix architecture. Markov transition analysis revealed section-specific vocabulary patterns — the herbal, astronomical, and "biological" sections have detectably different lexical profiles. This is consistent with either:
- A constructed/synthetic language with deliberate systematic grammar
- A real language whose morphological structure has been partially disrupted by encipherment
The Entropy Problem
The central statistical anomaly remains: Voynichese has dramatically lower second-order entropy (h2 ≈ 2) than any known natural language (h2 = 3–4). This means character sequences are unusually predictable — more like a cipher with constrained slot-filling than fluent text. However, the 2025 structural analysis shows this low entropy is localized to word boundaries, not uniformly distributed throughout words. The three-part word structure (prefix/root/suffix) explains the h2 anomaly: constrained slots at fixed positions create artificially low entropy without eliminating semantic structure.
Currier A vs. B: The Entropy Divergence Hypothesis
Prescott Currier identified two statistically distinct "languages" or scribal hands (Currier A and B) in different manuscript sections. The 2026 conference is expected to feature several papers on whether A and B reflect:
- Different languages (diglossic text)
- Different cipher keys or dice-parameter settings (supporting Naibbe-style cipher hypothesis)
- Different scribal styles encoding the same underlying language
The entropy profiles of A and B differ measurably: Currier A has slightly higher h2 (more variable) and Currier B has more stereotyped, repetitive word forms. The Naibbe cipher hypothesis (see concept voynich theories) predicts this: different dice + card settings would produce sections with different entropy profiles even from the same underlying text.
AI Approaches: Promise and Limitation
Large language model attempts to decode the manuscript hit a hard wall: without a validated segment of plaintext to anchor the model, LLM outputs are entirely unconstrained — any sufficiently flexible model can generate plausible-sounding "translations." The March 2026 Manifold prediction market set the probability of a scientifically recognized decipherment by end of 2026 at roughly 5–15%, reflecting cautious optimism from the conference itself while acknowledging the field's history of confident-but-wrong breakthroughs.
AI's actual contribution in 2025–2026 is structural mapping, not translation:
- Identifying the ~50 most frequent glyph combinations and their positional constraints
- Clustering sections by vocabulary overlap to infer chapter structure
- Cross-referencing manuscript figures (plants, astronomical charts) with contemporaneous European herbal and astronomical traditions via image embedding
The Materiality of the Voynich (Lisa Fagin Davis, 2025)
Codicologist Lisa Fagin Davis — one of the conference's senior advisors — published a 2025 analysis of the manuscript's physical construction that established:
- Five distinct scribal hands (confirming earlier analysis), each with characteristic glyph forms
- The manuscript was bound in its current order relatively late, and there is evidence the original folio ordering differed
- Certain quire signatures suggest the manuscript was used, not just produced — smudges, handling wear, and marginalia placement indicate a reader navigating the volume
This materiality evidence is important: a purely decorative hoax wouldn't show evidence of repeated use. The physical data pushes against the pure-gibberish hypothesis without confirming real language.
Where the Field Actually Stands in 2026
| Hypothesis | Status |
|---|---|
| Meaningless hoax | Weakened by structural language features and use-wear |
| Natural language (unknown/dead) | Possible; entropy anomaly unexplained |
| Constructed synthetic language | Consistent with evidence; can't be disproven |
| Naibbe-style dice cipher | Statistically plausible; lacks a confirmed plaintext segment |
| Steganographic (hidden meaning in illustrations) | Speculative; no breakthrough |
The conference's most likely outcome: a narrowing of the hypothesis space rather than a solution. The entropy anomaly localized to word boundaries, the five-scribe collaborative structure, and the use-wear evidence together rule out some theories and constrain others — but "ruled out" isn't "solved."
See Also
- concept voynich manuscript — full manuscript history and codicological facts
- concept voynich theories — all serious decipherment theories including Naibbe (2025)
- concept probability cipher connection — the deep unity of probability and cryptography that the Naibbe hypothesis exploits
- concept indus valley script — another undeciphered script where AI is helping map structure without achieving translation
- concept linear a — Minoan script; phonetics partially recovered via Linear B but language remains opaque
- concept information theory — Shannon entropy; the theoretical framework for h2 analysis