# Assessment of prior findings against the transcription

Date: 2026-10-22
Data: Zandbergen–Landini transliteration ZL3b (13 May 2025, IVTFF 2.0, EVA alphabet), cross-checked against Takahashi (IT2a) and Currier–D'Imperio (CD2a). Scripts: `data/analysis/`.

## Summary

The research record as of March 2026 (FINDINGS_TRACKER.md, RESEARCH_LOG.md, the Picatrix note and the draft chapters) rested on no machine-readable text. The `data/`, `transcription/` and `research/statistical` folders were empty; the only primary material on disk was twelve low-resolution folio images. Every lexical reading in the tracker was produced by a language model describing glyphs in those images. When those readings are checked against the standard transliterations, the specific claims fail, while three underlying structural observations turn out to have a real basis. Details below, one line per claim.

## Claims that do not survive

| ID | Claim | Test | Result |
|----|-------|------|--------|
| Z-01..Z-09 | Nine zodiac names read from labels on ff. 70–73 (`qotain` Pisces, `otchor` Aries, `shcamed` Gemini, `qodar` Scorpio, `nottbas` Taurus, `occtird` Libra, `obecand` Capricorn, `olpidag` Sagittarius, `Spato` Virgo) | Search all 37,691 tokens of ZL3b | `shcamed`, `nottbas`, `occtird`, `obecand`, `olpidag`, `spato` do not occur anywhere in the manuscript in any transliteration; the letters b, p-after-s, and word-initial n/s-p are not even possible EVA sequences in those positions. `otchor` occurs 14× — in the herbal and recipe sections, never on a zodiac page. `qotain` occurs 60× — in the biological and recipe sections, never on f70v2 (Pisces). `qodar` occurs 12×, never on f73r (Scorpio). |
| "Almanac prediction confirmed" | `nottbas`, `occtird`, `obecand` found as section headers on ff. 106r, 108r, 110v | Same | None of the three strings exists. ff. 103–116 contain starred paragraphs, no headers. |
| PL-01..PL-06 | Planet roots `Merc-`, `Satur-`, `Iupit-`, `Ven-`, `Sol-`, `Lun-` on ff. 87–94 | Same | No such strings. The glyph inventory has no sequences that transliterate as these. The Picatrix "cross-check" therefore matched a table of medieval planet correspondences against labels that do not exist. |
| M-04 | Lexeme `follan`, balneological, p < 0.001 | Same | Does not occur. |
| ST-01 | Index of coincidence 0.065–0.072 by section, "consistent with English/Latin" | Computed | IC on the EVA character stream is 0.077 (A 0.082, B 0.077); the reference texts give 0.070–0.074. The figures in the tracker were not computed from anything. The comparison is also weak evidence either way, since IC depends on how the script is transliterated. |
| S-02 | Currier A/B "resolved" as register difference | — | Not a finding; a hypothesis with no test attached. The data show A and B differ in vocabulary (25 % of tokens lie in words significantly skewed between Herbal A and Herbal B) *and* in scribe (Davis: A by scribes 1 and 4, B by 2, 3, 5), so the two are confounded in this corpus. What the text can show is reported in the book, chapter 6. |
| S-03, CL-02 | Seven-planet pharmaceutical classification, Picatrix match "near-perfect" | — | Built on PL-01..06. Withdrawn. |
| S-04, S-05 | Rosette as al-Idrisi six-zone map; bathers as 360 decan figures with `oll-` degree labels | Counted | The rosette foldout has nine circles, not six. Nymph labels in the biological section (type Ln) number 64 tokens, not 180–220; "oll-" as a prefix (EVA `ol-`) is a general feature of the text, not of these labels. |
| "89 % corpus coverage", "47/47 paragraph-initial positions" | — | No record of what was counted. |

## Observations with a real basis (restated correctly)

| Prior label | What is actually in the data |
|-------------|------------------------------|
| M-01 "article prefix 4oll-/foll-" | Currier `4O` = EVA `qo`. 15.5 % of running-text words begin with `q`; 99 % of all `q` are word-initial and 97 % are followed by `o`. `qo-` is the single most constrained position in the script. Whether it is an article, a prefix, or a spelling convention is not determinable from distribution alone; it is almost absent from labels (0–3 %), which is the one fact worth building on. |
| M-02 "entry particle Tor-/For-" | Currier `P`/`F` = EVA `t`/`k` (gallows). 84 % of paragraph-first lines begin with a gallows glyph, against 9 % of other lines (ZL paragraph markers, n = 730 vs 3,304). This is Currier's and Tiltman's old observation and it is robust. |
| M-03 "classifier oll-" | EVA `ol` is the second most frequent word (520×) and a frequent word-initial string; `ol-`-initial compounds (`olkeedy`, `olchedy`, …) are concentrated in Currier B. Nothing ties it to celestial content. |
| S-01 botanical entry formula | Paragraphs do have a structure: gallows-initial first line, shorter final word, `y`-final words dominating line-internally, `m`-final words concentrated at line end (15.6 % of line-final words end in `m` against 0.9 % elsewhere). This is line-level, not entry-level, structure. |

## What this means for the book

Nothing from the tracker is cited in the new manuscript except through this memo. The book is rebuilt from the transliteration, the published scholarship, and the object itself, and it claims only what those support.