A new line
of inquiry.

Keeping uncertainty visible—and correcting the image record

The f1v chapter now has a conservative comparison of all ten loci in IT 2a and ZL 3b. After spacing markers are removed, drawing interruptions retained, and the first supplied alternative selected, four loci still have different sign strings: f1v.1, f1v.2, f1v.3, and f1v.7. The other six agree only under that normalization; f1v.9 retains an alternative reading. These are editorial comparisons, not six certain manuscript readings. The raw alternatives remain intact. No reading was changed to make it resemble Latin.

Correction to the preceding comparison entry: fragment 204 on f102r1 does show a broad blue-green basal structure. A fold obscures part of its extent, not the whole structure. The earlier wording that the broad base was not preserved was too strong. The Czech and American English chapter now records the partially visible base explicitly. The earlier journal entry is retained below as part of the research history and should be read with this correction.

A further source check located Jorge Stolfi’s annotated 1998 file, which attributes a Fuchs page reference to Petersen. It is an indirect witness: neither Petersen’s original note nor the exact Fuchs edition and page has been verified here. The identification proposal remains a reported historical suggestion.

The new computational inventory is available as T-VAR-001; the image correction is recorded separately. None of these checks establishes a plant species, an anatomical organ, a cipher key, or plaintext.

A recurring plant is not yet an identification

The authenticated Yale image containing f102r1 has now been compared with f1v. The relevant drawing is fragment 204, at the right of the third row on the center panel. It shares a branching axis, broad simple leaves, and divisions near some leaf attachments with f1v. It does not preserve f1v’s terminal structure, alternating green and ocher, or broad base, and its leaf count and arrangement differ.

A second control comes from a fifteenth-century northern Italian herbal. On University of Pennsylvania, Oversize LJS 419, fol. 36r, a plant is openly joined to a female human figure. This establishes that anthropomorphic plant imagery could occur inside a medical-botanical codex. It does not prove that the Voynich image conceals an organ, nor that the two manuscripts share a workshop or model.

The structured comparison is recorded as T-IMG-002. Its result is deliberately narrow: fragment 204 is an internal motif parallel, not an exact copy or a botanical identity. The chapter remains in draft while uncertain transcription readings and the primary basis of older plant identifications are checked.

No plant species, organ, cipher key, or plaintext was established in this step.

A drawing is not an empty space

Work on f1v has begun with the original Yale image and four documented close-ups. The chapter remains in draft. Its two text blocks are encoded in IT 2a and ZL 3b as four and six lines, each interrupted by the plant drawing.

The first executed check found a processing trap. Removing the drawing marker as ordinary markup would fuse the units on either side. Our layout-preserving count gives 90 tokens for IT and 81 or 92 for ZL, depending on how uncertain spaces are treated. A deliberately faulty control loses exactly ten boundaries in every run. These are transcription units, not ten newly deciphered words.

The new parser keeps the two sides of each drawing interruption separate. The earlier f1r counts are unaffected: their saved inputs contain no such markers. The results and raw folio records and analysis script document the check. The two transcriptions are related research editions, not independent manuscript witnesses.

Test T-SEG-001 is a segmentation and software-regression check. It establishes no cipher key, Latin plaintext, botanical identification, or anatomical meaning. Comparison with related plant drawings and historical sources is still ahead.

What a space can change

The first folio chapter is complete in Czech and American English. It begins with the whole original page and follows four details: the paragraph endings, the red marks, the faint marginal alphabets, and a repeated sequence of signs. The cumulative edition now contains the introduction and f1r, with 28 endnotes and 28 source records.

Our first executed comparison concerns the units from which a reading would have to be built. Two specified transcriptions of f1r agree on four text blocks, but the total ranges from 203 to 215 tokens when uncertain spaces are treated differently. The repeated pair chol chol survives all four runs. Those letters are EVA transcription labels; they are not a Latin translation.

The green forms offer a second check on our interpretation. They show through from the plant drawing on f1v, rather than forming a separate green diagram painted on this side. Lisa Fagin Davis’s published discussion helps distinguish that physical relationship from the later writing in the margin. Before we propose a bodily meaning, we need to establish which side of the leaf carries the paint.

The comparison used IT 2a and ZL 3b, with the normalization and results recorded as T-LAT-001. It establishes no cipher key or plaintext. The next chapter examines f1v on its own side of the leaf.

Beginning with enciphered Latin

I have made the linguistic starting point explicit: this investigation will proceed from enciphered Latin, alongside the existing gynecological premise. Medieval medical Latin will provide the first comparison corpus. The language and subject remain separate questions: evidence for one does not automatically establish the other.

A candidate reading will need an explicit conversion rule, documented treatment of abbreviations, and grammatical connections across the text. The exact output of a key, any abbreviation expansions, the normalized Latin, and its translation will be recorded as separate stages. The same key must then work on passages held aside during its construction. These are planned tests; this update does not present a deciphered passage.

Useful comparisons include Michael Greshko’s Naibbe model, which uses both Latin and Italian, and the abbreviation reference tools Cappelli and Ad fontes. Their roles differ: a generative model does not identify the manuscript’s language, and an abbreviation dictionary does not supply a cipher key.

A beginning that can be tested

The new introduction is complete in Czech and American English. It sets out the author’s gynecological premise, the history needed to examine it, and the tests that a proposed reading must face. The chapter has 17 endnotes, an index, and a register of 18 sources. Two original Yale folio images introduce the visual inquiry.

The distinction between an observation, an interpretation, and a textual reading will guide each folio chapter. A plantlike form can suggest a bodily analogy, but the analogy needs evidence from recurring relationships, historically grounded comparisons, and the surrounding text. The next chapter begins with folio 1r, followed by 1v.

This is a working book edition. No verified plaintext has been established. The source register identifies which publications were available in full and which could be checked only through abstracts or bibliographic records.

Whose knowledge was being protected?

A gynecological subject would not by itself identify the manuscript’s authors or readers. Monica H. Green’s research on male authority in premodern gynecology makes that question more complex. Written medicine circulated through relationships of education, authority, and access to books.

Keagan Brewer’s open conference paper on concealment in late-medieval gynecology provides specific historical comparisons. Brewer and Michelle L. Lewis’s 2024 article proposes a reproductive interpretation of the Rosettes. Its abstract has been verified for this introduction; a detailed assessment of that proposal awaits a full-text review.

New models for the writing and the book

Michael A. Greshko’s Naibbe project, published in 2025, models how ordinary text can become ciphertext resembling aspects of Voynichese. It supplies a comparison for future testing, without translating the manuscript.

Colin Layfield and Lisa Fagin Davis’s 2026 study of singulion structure proposes an original sequence of individual bifolia. Its publisher-deposited abstract has been checked. The book will retain the present folio identifiers while that structural hypothesis awaits a full audit.

Neither model has been independently reproduced in this research step. Recent scholarship is being added to the evidence register with its limitations visible.

A gynecological hypothesis, folio by folio

I have chosen a gynecological reading as the starting point for the next stage of my Voynich book. I will examine the manuscript as a guide to women’s health, with its female figures at the center of the inquiry and its plant-like forms considered as a possible visual vocabulary of organs and bodily processes. This is a declared, subjective choice about where to begin. The manuscript’s writing has not yielded a verified reading in this investigation.

The question is whether this reading can hold together across the manuscript. A shape interpreted as a bodily structure on one folio must be compared with its other appearances. Historical parallels will need specific sources. Proposed readings of the writing will need rules that another reader can follow.

The book will place the evidence in front of the reader: a full-page reproduction at the opening of each folio chapter, enlarged details at the start of its subsections, and a discussion of the image and text together. Uncertain passages, competing explanations, and unsuccessful tests will remain part of the account.

A foldout can disappear from a chapter list

The first source check found a gap between an earlier chapter list and the available images. The list contained 183 page or panel entries, while five additional image files associated with folios 85–86 were stored separately. They included individual views and a stitched view of the Rosettes foldout. Those files were absent from that chapter list.

Five files do not necessarily mean five additional folios. A scan can show several panels, and another can show a different view of the same sheet. The next step is to reconcile the physical foliation, the image sequence, and the planned chapters so that the illustrated atlas accounts for each surviving surface without counting an overlapping view twice.

This is a finding about the project’s source inventory. It does not establish a meaning for the foldout or its text.