Evidence · Voynich Manuscript

A calibrated posterior over period-cipher settings pins down spacing and the line layer, not the key's size

Neural posterior estimation over the period-cipher family, calibrated by simulation-based calibration and better than rejection ABC, pins down stream spacing, the line layer, simplified spelling and the absence of syllable signs and positional forms in both herbal sections, but not nomenclator size, homophone counts or copied nulls; its best settings fit held-out herbal pages about as well as T39's (3-4/12 missed in Herbal A, 2-3/12 in Herbal B, against 5/12 and 3/12).

Supported0

What it means

Key questionHow was the text produced on the page?

Most likely: the line is a unit of the system

Most likelyunchanged

Part of itWhat procedure reproduces its statistics?

Most likely: each scribe ran the procedure in his own way

Most likelyunchanged

Part of itWhich combination of fifteenth-century cipher practices, if any, reproduces a single section's statistics on its held-out pages?

No answer ahead yet

Openunchanged

Answer this evidence bears onLatin as a re-spaced homophonic stream (period practices)

Bears on1 for, none against

Still possibleunchanged

Next steps 5

Give the T39 signs a word-internal grammar (slot-position sign tables learned from the training pages) and refit: the remaining misses (glyph predictability, one-off words, edge dependency) all say the manuscript's words are more rigidly built than freely joined signs.

~6 h

On: Which combination of fifteenth-century cipher practices, if any, reproduces a single section's statistics on its held-out pages?

—not judged

Homophone rung (R2 of T44): the search now sits 64 nats from the true key yet 30 % of sign uses are wrong and coverage is flat at 0.47 (bar 0.487). Before more search, diagnose which element pairs the wrong rows confuse (confusion matrix of the best state against the planted key) and whether a richer Latin model (4-gram, or word-level dictionary scoring inside EM) separates them; then rerun R2 with the same effort and the trigram-objective hopping started earlier.

~6 h

On: Which combination of fifteenth-century cipher practices, if any, reproduces a single section's statistics on its held-out pages?

—not judged

Segmentation-aware solver for the stream family: a segmental hidden Markov model (signs of one to three glyphs as emissions, the sticky homophone column as a latent state) fitted jointly with the key, validated on rung B of M-T40 before any use on the manuscript.

~16 h

On: Which combination of fifteenth-century cipher practices, if any, reproduces a single section's statistics on its held-out pages?

—not judged

Sharper null for decodes: run the fitted generator on shuffled Latin words, which keeps the cipher's structure (edge dependency, line layer) and removes plaintext order; a reading must beat that, not only the shuffled manuscript.

~3 h

On: Which combination of fifteenth-century cipher practices, if any, reproduces a single section's statistics on its held-out pages?

—not judged

Ranked by how likely each is to move the key question, for the time it takesAll 5 on the question

Sources 3

Fast ε-free Inference of Simulation Models with Bayesian Conditional Density Estimation

Papamakarios 2016

SRC

Truncated proposals for scalable and hassle-free simulation-based inference

Deistler et al. 2022

SRC

Validating Bayesian Inference Algorithms with Simulation-Based Calibration

T et al. 2018

SRC

Cite it

Mysteries Department. (2026). A calibrated posterior over period-cipher settings pins down spacing and the line layer, not the key's size [Evidence EVD0DCHSNX5R3VX]. https://mysteriesdept.com/r/EVD0DCHSNX5R3VX

@misc{mysteriesdept_evd0dchsnx5r3vx,
  title     = {A calibrated posterior over period-cipher settings pins down spacing and the line layer, not the key's size},
  author    = {{Mysteries Department}},
  year      = {2026},
  note      = {Evidence, catalog key EVD0DCHSNX5R3VX},
  url       = {https://mysteriesdept.com/r/EVD0DCHSNX5R3VX}
}