Botanical geography — preliminary literature and corpus findings

Preliminary finding: No reliable species or geography conclusion. The 134 external illustration entries map to source views, but exact native illustration boundaries are still unverified. Plant-part terminology and model confidence are not calibrated; species accuracy is unmeasured.

2026-09-30 · Preliminary preparatory study. No accepted species, historical range or manuscript-origin result. Visual Transcription Pass 3 remains the Current Project.

Question and result

Can botanical pictures supply defensible candidate identifications and geographic constraints? There are real comparison leads, but their presence does not establish species identities or origin. The most useful immediate work is source/instance reconciliation plus independently recorded plant-part observations, followed by calibrated comparisons with identified historical herbals. Geography should follow supported identifications, rather than choose them.

Lead verification and limits

O'Neill 1944: The publisher's public one-page original preview verifies the actual argument: Urtica on fol.25, Botrychium lunaria on fol.100v, Helianthus annuus on fol.93 (side unspecified), and Capsicum on fol.101v. O'Neill reports agreement of six unnamed botanists for sunflower and argues for a date after 1493 using his pepper introduction claim. These are the author's identifications and historical argument, not accepted labels or independently verified introduction history. Fol.101v also needs reconciliation with secondary retellings. No origin follows from this note alone. Original publisher preview, bibliographic record.

Tucker/Janick atlas: *Flora of the Voynich Codex: An Exploration of Aztec Plants* is ©2019, distinct from their 2018 *Unraveling the Voynich Codex*. The publisher states the authors propose a sixteenth-century New Spain origin, reports 131 large herbal images and 228 small plant-part illustrations, and advertises comprehensive identification. This verifies the publication and its claim, not the correctness or independent scholarly acceptance of identifications. Full diagnostic chapters have not been evaluated. Publisher.

European herbal affinities: Zandbergen's primary 2012 presentation compares Tractatus and alchemical herbals, including f035v against Paris BnF Latin 6823 f.60r and f007v against Latin 6823 f.81v / Sloane 4019 f.46r. Its author explicitly asks for specialist confirmation and reports no complete copy among 50 herbals inspected. These are concrete iconographic leads; style/detail resemblance does not establish plant species or production place. The presentation contains permission acknowledgements, so copying its external illustrations into a new public presentation is not automatically licensed. Primary slides, pp.14–25. Touwaide's 2015 pp.141–158 article and Toresella/Segre Rutz publications remain primary-text acquisition gaps; do not collapse secondary summaries into verified scholarly consensus.

Voynich Garden: Its own About page explicitly reports 134 whole-plant illustrations; the gallery independently exposes 134 unique instance IDs across 126 external page names. Its cleaned crops, AI renditions and voting percentages are different evidence types. Votes are not calibrated identification probabilities. The instance suffixes are external illustration IDs, not new manuscript folios. About, gallery. Only identifiers/counts were brought into the corpus mapping; no votes or proposed taxa were used for blind input.

Date and institutional context: Yale describes scientific material consistency with the fifteenth century and an undeciphered manuscript, while its page's headline uses a later production date. Do not turn that mixed summary into a precise drawing date or silently choose it to favor a flora. Direct dating evidence must be separated from ink/drawing timing in a later range stage. Yale.

Existing image/text work: Sterneck, Polish and Bowern's 2021 topic-modeling abstract reports joint association with scribe and illustrated subject. This motivates controlling scribal/layout effects; it does not prove plant names or translation. Only the abstract was assessed here. Primary preprint.

Corpus reconciliation

Canonical inventory: 213 Yale source views, all unique Yale IDs. Garden: 134 instances / 126 external page names. 127 instances map to an exact canonical source-page filename; illustration boxes remain unverified. Seven initially unmatched instances now have verified source-view crosswalks, while exact native boxes remain unaccepted. External f89v-a/b correspond to catalog f090r1/r2 on the combined Yale view; f90v-a to f090v1; f94v-a/b/c to f094v/f095r1/f095r2; f95r-a to f095v2. Original Yale partial-view ambiguity is retained. Quire 15 scan layout, Quire 17 scan layout. The local section metadata's 118 herbal *source views* is another unit, not a count of whole plants. The atlas's 131 cannot yet be reconciled row-by-row without its inventory. No folio or plant identity was invented. See `corpus_reconciliation.json`.

Preliminary visual pilot

Two Yale source pages, f093r and f011v, were selected deterministically and inspected as resized whole-page inputs by a local qwen3.8 vision model. Initial observations and a bounded repair were preserved before assessment. Four further source-view mapping inputs were tested. Species identification was deliberately not attempted; this is a feasibility pilot, not a botanical accuracy benchmark.

What needs to happen next

The next research stage is to verify native illustration boundaries, obtain accessible diagnostic primary texts, and calibrate a small pilot against identified dated herbals, mismatched alternatives, abstention and whole-folio holdouts before scaling.

Only supported candidates proceed to circa-1400 native/cultivated/traded/copied-presence evidence. Modern distribution is insufficient. No geographic overlap or percentages are currently defensible. Image-text correlation stays a separately authorized later stage with frozen independent labels, predeclared similarity/controls/permutations, multiplicity correction, and v006 boundary/segmentation sensitivity. Preserve negative results and no-identification outcomes.

Local results and meaning

The local model can describe coarse layouts, but overstates leaf arrangements and uncertainty, and its proposed boxes may clip roots. Species accuracy is unmeasured; four completed mapping calls do not establish botanical accuracy. No species, historical range or geographic inference is accepted. A small controlled botanical pilot is the next research stage after boundary verification and accessible diagnostic evidence; mass classification and image-text correlation have not been run.

Explore All Results

All 134 external gallery identifiers; exact source page matching is separate from identifying plant-instance boundaries. Seven exceptional source-view crosswalks are resolved below; exact native boundaries remain unverified. These identifiers carry no species labels.

All local observations and repairs

Initial answers remain visible. No answer is an accepted plant identity. All attempts were saved and frozen before coordinator assessment.

B01_attempt1

Local model: qwen3.8:latest · 16.5 seconds

```json
{
  "roots": "Present. A dense cluster of numerous, thin, wavy, reddish-orange lines extending downwards and outwards from the base of the stem.",
  "stem": "Present. A single, central, vertical stalk, colored pale pinkish-red, running from the root cluster to the top of the plant.",
  "leaves": "Present. A pinnate arrangement of approximately 12-14 large, broad, green leaflets attached along the central stem. The leaflets are roughly oval or lanceolate, with smooth margins, and are arranged in opposite pairs.",
  "flowers_or_fruit": "Present. A large, terminal, diamond-shaped or rhomboid structure at the top of the stem. It is colored tan/brown and covered in a regular pattern of small, dark dots. The base of this structure flares out with fine, hair-like lines connecting to the stem.",
  "possible_composite_or_symbolic_features": "The overall structure is a single composite unit: roots, stem, pinnate leaves, and a distinct terminal head. The terminal head's dotted pattern and shape are distinct from the leaves. The geometry is highly stylized and symmetrical.",
  "unclear_features": "The connection between the terminal head and the stem is depicted with fine, radiating lines, making the exact attachment unclear. The internal structure of the terminal head is not visible, only the surface pattern. The leaf venation is not depicted.",
  "identification_status": "not_attempted"
}
```

B02_attempt1

Local model: qwen3.8:latest · 17.0 seconds

```json
{
  "roots": "Present. A central vertical root structure branches into multiple long, thin, wavy tendrils extending downwards and outwards. Several small, oval or circular shapes are attached to the branching points near the base of the stem.",
  "stem": "Present. A single, relatively straight vertical stalk connects the root system to the main body. It is depicted with fine parallel lines suggesting texture or striations.",
  "leaves": "Absent as distinct individual leaves. The upper portion is a dense, rounded mass composed of overlapping, scale-like or petal-like segments in alternating green and brownish-orange colors.",
  "flowers_or_fruit": "The large, bulbous upper structure resembles a fruit or a dense flower head. It is ovoid/teardrop shaped, tapering towards the stem. The surface is covered in a repeating pattern of overlapping scales. A small cluster of blue and white elements sits at the very top apex.",
  "possible_composite_or_symbolic_features": "The overall structure is highly stylized and symmetrical. The scale pattern is uniform and geometric rather than naturalistic. The combination of a bulbous, scaled body on a thin stalk with wavy roots suggests a symbolic representation or a stylized depiction of a specific plant part (like a pinecone, artichoke, or similar) rather than a full plant with separate leaves and flowers.",
  "unclear_features": "The specific nature of the small oval shapes on the roots is unclear. The blue/white element at the top is small and its exact form is difficult to discern. The boundary between the 'fruit/flower' mass and the stem is somewhat blended.",
  "identification_status": "not_attempted"
}
```

B02_repair1

Local model: qwen3.8:latest · 17.9 seconds

```json
{
  "roots": "A cluster of thin, wavy, root-like lines extending downwards from the base of the stem. These lines are outlined in a reddish-brown ink and are interspersed with several small, uncolored, oval or circular shapes.",
  "stem": "A single, vertical structure composed of multiple fine, parallel lines in reddish-brown ink. The lines converge at the base and fan out slightly at the top to support the upper mass.",
  "leaves": "A small, fan-shaped cluster of dark blue, pointed shapes located at the very apex of the illustration, emerging from the top of the main body.",
  "flowers_or_fruit": "A large, bulbous, ovoid mass forming the main body of the illustration. It is composed of a dense, overlapping grid of scale-like or petal-like units. These units are colored in alternating patterns of green and reddish-brown. The shapes are roughly oval or teardrop-shaped and are arranged in a tessellated pattern.",
  "possible_composite_or_symbolic_features": "The overall structure combines a root system, a multi-stranded stem, a large scaled body, and a distinct blue crown, creating a unified but geometrically complex form.",
  "unclear_features": "The specific nature of the overlapping units in the main body (whether they represent scales, petals, leaves, or fruit) is ambiguous based on geometry alone. The function of the small, uncolored ovals within the root system is unclear.",
  "identification_status": "not_attempted"
}
```

M01 — initial local mapping

20.1 seconds; raw response frozen before review

```json
{
  "plants": [
    {
      "order": 1,
      "geometry": "A central plant with a bulbous base, three spiky green leaves, and two long, curving stems ending in large blue spheres.",
      "bbox": [390, 100, 680, 950]
    },
    {
      "order": 2,
      "geometry": "A tall plant with a bulbous base, branching stems covered in numerous small green circles, and a long, winding root system extending to the right.",
      "bbox": [690, 40, 980, 950]
    }
  ],
  "other_regions": [
    {
      "description": "Top-left cluster of small botanical fragments and a green vessel.",
      "bbox": [80, 40, 380, 300]
    },
    {
      "description": "Left-side vertical arrangement of blue and green vessels.",
      "bbox": [80, 350, 150, 750]
    },
    {
      "description": "Bottom-left cluster of root-like structures and small green leaves.",
      "bbox": [130, 600, 380, 980]
    }
  ],
  "uncertainty": "The separation between the central plant and the right-hand plant is clear, but the root systems extend significantly, making the exact boundaries of the 'whole plant' slightly ambiguous. The left page contains multiple small, disconnected elements that are difficult to classify as single 'whole plants'."
}
```

M02 — initial local mapping

10.8 seconds; raw response frozen before review

```json
{
  "plants": [
    {
      "order": 1,
      "geometry": "A single central vertical stem with paired, downward-pointing triangular leaves. The stem branches at the top into three distinct flower heads, each with a dark center and radiating petals. The base of the plant transitions into a brown, root-like structure that extends horizontally to the left and right, with the right side ending in a clawed or hand-like shape.",
      "bbox_2d": [198, 25, 969, 975]
    }
  ],
  "other_regions": [],
  "uncertainty": "Low. The illustration is a single, continuous drawing of a plant-like entity with a clear stem, leaves, flowers, and root system. There are no separate, distinct plant drawings to separate."
}
```

M03 — initial local mapping

18.6 seconds; raw response frozen before review

```json
{
  "plants": [
    {
      "order": 1,
      "geometry": "A central rosette of broad, lanceolate leaves radiating from a base, with a single tall, slender stem rising to a small, dark, circular flower head. The root system consists of several thick, tuberous structures extending downwards.",
      "bbox": [80, 150, 380, 950]
    },
    {
      "order": 2,
      "geometry": "A cluster of three to four upright stems featuring pinnately compound leaves with rounded leaflets. Small, delicate flowers are visible at the tips of some stems. The base shows a dense, fibrous root structure spreading outwards.",
      "bbox": [390, 350, 680, 950]
    },
    {
      "order": 3,
      "geometry": "A dense, bushy plant with numerous upright, lanceolate leaves forming a thick clump. Several stems extend upwards with small, green, pod-like structures. The root system is depicted as a thick, reddish-brown mass with fine, hair-like roots extending from the bottom.",
      "bbox": [700, 330, 970, 950]
    }
  ],
  "other_regions": [],
  "uncertainty": "Low. The three plant illustrations are clearly separated by page breaks and distinct visual boundaries. No significant overlap or ambiguity in clipping."
}
```

M04 — initial local mapping

12.1 seconds; raw response frozen before review

{
  "plants": [
    {
      "order": 1,
      "geometry": "Central vertical stem with a large, branching, comb-like root system at the base. Mid-stem features two clusters of green, serrated leaves. The top consists of multiple thin, branching stalks terminating in small, round, uncolored buds or berries.",
      "bbox_2d": [268, 203, 920, 968]
    }
  ],
  "other_regions": [
    {
      "description": "Small fragment of a plant visible on the far left edge, showing a stem with two dark green leaves and small red berries.",
      "bbox_2d": [58, 448, 168, 626]
    }
  ],
  "uncertainty": "The left-edge fragment is clipped by the page boundary and its full extent is unknown."
}

Coordinator assessment after freeze

Feasibility response accepted; advisory only. Roots, stem and dotted terminal structure are observable. Oval/lanceolate and opposite-pair terminology is overly categorical for curved illustrated units. Counts and pinnate interpretation are not accepted labels.

Plant analogies introduced despite no-species prompt; absent-leaf judgment too categorical. No plant analogies or geographic inference; competing interpretations of upper units explicit. Feasibility response accepted with remaining caution: blue apex classified as leaves is an unverified part interpretation.

Review Needed

• Exact native illustration boundaries remain unverified for all 134 external instances
• Touwaide primary text and atlas diagnostic chapters remain unavailable
• Independent identified-herbal controls and botanical abstention calibration required before scaling

Scaling gate is not cleared. Species accuracy and independent botanical agreement are unmeasured.