Data and reuse

How researchers represent the writing

A typed letter is a label for a shape, not a decoded sound.

From ink to a searchable label

EVA is a modern system developed by Gabriel Landini and René Zandbergen with contributions from Jacques Guy. It records shapes without establishing pronunciation. EVA o does not mean anyone has established the sound “o”. Zandbergen’s explanation and illustrated alphabet tables show the shapes beside their codes.

The same shape can be grouped differently

These are alternative modern encodings, not translations. Neither settles how many historical characters the writer intended. See the alphabet comparison, Table 1.

Why this matters for the results

Word length and symbol frequency depend on the chosen encoding and treatment of uncertain readings. A count that includes both alternative readings is not automatically a count of distinct marks on parchment. Read the named transcription and counting rules beside each result.

A familiar shape is a comparison, not an origin

Some forms resemble Latin letters or medieval abbreviations. Resemblance alone establishes neither ancestry nor meaning. The writing-system overview and abbreviation comparisons provide historical context, not a decoding key.

The reviewed symbol-origin inventory establishes no dated antecedent for a Voynich sign. This describes the reviewed sources, not proof that no antecedent exists. This guide does not reproduce those counts as a census of physical glyphs.

The full research record

Explore all 51 graded findings, tested hypotheses, scripts, results and rerun instructions at github.com/thisisntjon/voynich (code under MIT, text and results under CC BY 4.0).

Data credits

Every count of words or glyphs in these records is taken from published transcriptions of the manuscript, not from a new reading of the scans. They are in René Zandbergen's IVTFF 2.0 format, from voynich.nu:

• ZL3b: the Zandbergen-Landini transliteration, the primary text used in these runs.
• RF1b: a reference transliteration built from the GC and ZL transliterations.
• IT2a: Takeshi Takahashi's transliteration, as included in Jorge Stolfi's 1999 interlinear file.

The Currier A and B labels and Currier's hand ids, and Lisa Fagin Davis's scribe (hand) attributions, are used as these files record them. In ZL a comma marks an uncertain space; the project parser reads it as a word break unless a run says otherwise. Plant-resemblance judgments and other prior work are credited on each record. I link these sources and do not redistribute them.

Inside the download

Files

Evidence files

42 project files that the records cite, published whole, with local machine paths replaced, or as the cited lines only. public-evidence.json says which, and lists the references that stay private and why.

Check a download

Compare a file's checksum with the table above. On macOS or Linux run shasum -a 256 records.json. On Windows PowerShell run Get-FileHash records.json.

shasum -a 256 records.json
# Windows PowerShell:
Get-FileHash records.json -Algorithm SHA256

Where the records come from

All graph findings inventoried from existing graph, not all experiments independently reviewed. Two findings included only as correction context.

The README explains export preparation, public reproduction and private-source limits.