A systematic visual and linguistic analysis of the botanical section of the Voynich Manuscript. Each of the 98 illustrated folios was examined directly, with plant morphology categorised by leaf shape, root type, and flower structure. Script patterns were then compared across pages sharing the same botanical features, yielding morpheme-level word connections grounded in image-to-text cross-referencing. A second deep pass of token decomposition has expanded the working lexicon from 7 to 18 items and identified three new grammar rules — including evidence for vowel harmony and the oll-cros root hypothesis.
Every folio follows the same format: one large plant illustration with Voynich text in 2–4 paragraphs. The first paragraph begins with a tall "gallows" initial character — mirroring how medieval Latin and Arabic herbals open each entry with a plant name.
Green leaves, red/brown roots, and coloured flowers appear with remarkable consistency. Some illustrations appear unfinished — line drawings without colour wash — suggesting the colouring was added in a second pass by a different hand.
Many plants closely resemble known European herbs. Others show biologically impossible features: geometric root symmetry, spiralling rhizomes, or leaves with alternating red-and-green horizontal bands.
All 98 illustrated pages in Volume I belong to the botanical section. The astronomy, balneology, cosmology, and pharmaceutical sections appear to begin in Volume II.
| Folio | Visual Features | Family | Candidate Species | Conf. |
|---|---|---|---|---|
| f. 5 | Spiky bracts, composite heads, serrated leaves, coiling red root | Asteraceae | Cirsium / Carduus (thistle) | |
| f. 12 | Palmate 5-lobed leaves, red 5-petal flowers, horizontal taproot | Geraniaceae | Geranium robertianum (Herb Robert) | |
| f. 6 | Massive round peltate leaf on single petiole, rhizome mass at base | Nymphaeaceae | Nymphaea alba (white water lily) | |
| f. 14 | Star-shaped spiky leaves, rounded blue-green spiny flower heads | Apiaceae | Eryngium campestre (field eryngo) | |
| f. 70 | Turquoise berry clusters, umbel canopy, scalloped leaves, zoomorphic fox root | Apiaceae | Conium maculatum (hemlock) | |
| f. 19 | Deeply divided pinnate leaves, red-berry flower clusters, long straight orange taproot | Apiaceae | Daucus carota (wild carrot) | |
| f. 23 | Dense white floret dome with blue flowers, three hollow stems, spiky root collar | Brassicaceae | Brassica oleracea (cauliflower) | |
| f. 24 | Large overlapping-scale green + tan structure, single small blue flower at crown | Pinaceae | Pinus sp. (pine cone) — only non-flowering plant | |
| f. 27 | Strap-like leaves with red midribs, blue bulbous seed capsule with red stigma eye, coiling root | Papaveraceae | Papaver somniferum (opium poppy) | |
| f. 35 | Paired oval leaves, large blue corolla with red-dotted white centre, small fibrous taproot | Gentianaceae | Gentiana sp. (gentian) | |
| f. 82 | Large crenate blue-green leaves, dense blue seed-dot canopy, two elongated red-brown tubers | Brassicaceae | Armoracia rusticana (horseradish) |
| Group | Description | Pages | Real-World Match | Conf. |
|---|---|---|---|---|
| L1 Oval | Smooth-edged oval/elliptical leaves. Two colour variants: solid green and bicolour green+tan. Most common type. Bicolour confirmed as abaxial anatomy — not fantastical. | f.4f.9f.16f.18 | Plantago, Salvia, Symphytum | |
| L2 Palmate | 3–7 finger-like lobes from a central point. Serrated or smooth margins. Rosette or branching arrangement. | f.11f.12f.17f.25 | Geranium, Ranunculus, Hellebore | |
| L3 Pinnate | Small leaflets in opposing pairs along a midrib. Narrow and broad-leaflet variants both present. | f.7f.13f.60f.90 | Valeriana, Polypodium, Sambucus | |
| L4 Linear | Long narrow strap-shaped leaves with parallel veins. Often in basal clusters. Blue-black flowers frequent. | f.15f.40 | Iris, Carex, Cornflower | |
| L5 Spiky | Deeply cut serrated margins with projecting spines. Two sub-forms: fine-toothed and spine-lobed. Marked by gt-cros in ¶2. | f.5f.8f.13f.14 | Cirsium, Ilex, Urtica, Eryngium | |
| L6 Peltate | Very large circular shield-shaped leaves; stem attaches near centre. Leaf fills most of the folio. | f.6f.11 | Nymphaea, Nasturtium, Petasites | |
| L7 Rosette | Stiff lanceolate leaves radiating from a central stem in a star/compass pattern. Marked by gttros / grott- in ¶2 — the grot- (radially-spreading) morpheme. | f.15f.50 | Mandragora, Taraxacum, Aloe | |
| L8 Crenate | Rounded wave-like scalloped margins. Broad ovate. Often in brownish-olive pigment. | f.25f.60 | Mentha, Melissa, Stachys |
| Group | Description | Pages | Real-World Match | Conf. |
|---|---|---|---|---|
| R1 Fibrous | Dense clump of fine thread-like roots. Light brown/tan. Shallow-rooted herb type. Most common. ¶3 uses plain ctros — NOT goll-. | f.9f.18f.20f.95 | Grasses, annual herbs | |
| R2 Taproot | Single thick central root descending vertically with small lateral rootlets. Red-brown pigment. ¶3 uses goll- family. | f.12f.14f.70 | Daucus, Pastinaca, Angelica | |
| R3 Tuberous | Bulging, bumpy, nodular mass. Like ginger or misshapen potato. Strong red colouring. goll- + possible cobag (knotty) descriptor. | f.7f.25f.80f.100 | Arisarum, Cyclamen, Mandragora | |
| R4 Bulb | Round/oval bulb from which thin fibres trail downward. Sometimes multiple bulbs clustered. | f.80f.90 | Allium, Tulipa, Muscari | |
| R5 Coiling | Root drawn as a tight coil or helix. Corkscrew shape. Possibly encoded as gotlcrowd — crowd = spreading/creeping morpheme candidate (Welsh crwydro). | f.5f.13f.30 | Stylised rhizome — artistic convention | |
| R6 Zoomorphic | Root drawn to resemble an animal — dog, fox, or dragon. Doctrine of signatures reference. Fox on f.70 = probable Galenic Mars/Mercury association. | f.70f.90 | Medieval mandrake / Galenic planet-association | |
| R7 Scaly | Root shown as overlapping scaly mass like a pine cone. Dense corm or rhizome cluster. Now linked to ottog token in ¶3. | f.4f.85 | Corm — Crocus, Gladiolus type |
| Group | Description | Pages | Real-World Match | Conf. |
|---|---|---|---|---|
| F1 Globe | Rounded composite heads with densely packed florets, often ringed with spiny bracts. Blue, blue-green, or white. | f.5f.8f.14f.85f.95 | Asteraceae, Dipsacaceae | |
| F2 Bell | Single large urn- or bell-shaped corolla. Blue or blue-violet with recurved lip or flared opening. | f.10f.100 | Campanula, Digitalis | |
| F3 Open petal | Simple open flowers, 4–5 rounded petals, visible stamens, blue/red/white colouring. Multiple per stem. Linked to otl- (rounded/globular shape word — revised). | f.12f.20f.90 | Viola, Geranium, Gentiana | |
| F4 Spike | Long flower spike with small flowers/buds arranged alternately up a vertical coloured stem. | f.16f.25 | Lavandula, Veronica, Salvia | |
| F5 Umbel | Flat or dome-shaped umbrella cluster at branch tips. Classic carrot-family arrangement. | f.70 | Apiaceae (carrot family) | |
| F6 Fantastical | Flower structures with no botanical parallel. Feathered crowns, ringed blue globes, dangling appendages. | f.7f.80f.100 | No known parallel — possibly symbolic | |
| F7 Catkin New | Segmented ovoid seed pods arranged in series — resembles a caterpillar or catkin spike. | f.28f.86 | Typha, Plantain spike, Catkin | |
| F8 Vine/Spiral New | Leaves spiralling in corkscrew habit along a twining stem — bindweed growth pattern. | f.22 | Convolvulus, Calystegia (bindweed) | |
| F9 Scale-cone New | Overlapping scale structure — not a flower at all but a reproductive cone structure. Only non-angiosperm in the manuscript. | f.24 | Pinus sp. (pine cone) |
| ¶1 | Name paragraph | Begins with tall gallows glyph. Always the shortest — 1–2 lines. Opens with llos-/llor- name-marker prefix. Second token = unique plant name per folio. |
| ¶2 | Leaf description | Longer block beside the leaves/stem. Contains cros family tokens. Modifier prefix signals leaf shape: plain = oval, gt- = toothed, grot- = rosette/radial. |
| ¶3 | Root description | Positioned near or below the root drawing. goll- = large/fleshy qualifier; ottog = scaly; llerod candidate for root base noun. |
| ¶4 | Use / habitat | Present in ~60% of folios. Short, sometimes right-aligned. Ends frequently with dand. Toxic-plant folios (f.70, f.27) likely share ¶4 vocabulary. |
The Voynich script has unusually low word-initial and word-final diversity compared to Latin, Arabic, or Italian. This is consistent with: (a) an agglutinative language with heavy suffixation, (b) an encoding system with artificially defined word boundaries, or (c) a syllabic script where each "word" is a morpheme cluster. The entropy is close to natural language, definitively ruling out random text and simple monoalphabetic substitution ciphers.
The endings -and, -ond, -end appear to be the same suffix in vowel-harmonic variants — the vowel of the suffix changes to match the dominant vowel of the root word. Examples: dand (a-vowel root), gotlond (o-vowel root), ctrand (a-vowel). Vowel harmony is the single most diagnostic feature of Finno-Ugric and Turkic languages. It does not appear in Latin, Welsh, or Romance languages. If confirmed across more folios, this shifts Hungarian, Finnish, or a Turkic language to the front of the candidate list.
The ending -erg appears in both ¶1 name tokens (llodaxerg) and ¶2 descriptive tokens (crgllerg). If -erg is a nominaliser — turning a root into a noun — it would explain its appearance in both positions: "the [dax]-thing" (plant name), "the [crg-ll]-thing" (a leaf property). This parallels Germanic -ung or English -ing noun-forming suffixes.
Tokens like ottogcros (scaly+leaf), ollcros (under+leaf), and gotlcrowd (large+spreading) show free cross-category compounding. This is characteristic of agglutinative languages and Germanic-type compounding. It is absent from Latin and Romance languages, which use separate words with grammatical agreement — further evidence against a Romance base language.
Pages sharing the same botanical feature were compared directly. Colour-coded tokens show what recurs in the same structural position across pages with the same visual feature.
Oval-leaf pages (f.4, f.9, f.18) use plain cros in ¶2. Spiky-leaf pages (f.5, f.14) use gtcros / gttros — the same base with a gt- consonant prefix absent from oval-leaf pages. The ¶1 name of f.70 (hemlock) contains gtaucros — gt- + infixed -au- + cros — likely meaning "deeply divided leaf", consistent with hemlock's deeply dissected foliage.
Updated interpretation: goll-/gottl- appears in ¶3 on large-root pages AND in ¶2 on large-leaf pages (f.9 gollers). This means goll- is a pure qualifier meaning "large/fleshy/prominent" — not a root-specific word. The actual root base noun is likely llerod (f.25 ¶3), with rod echoing Latin radix and Welsh gwraidd.
In f.25's root paragraph, ollcros appears — the leaf base-word cros prefixed with oll-. If oll- means "below/underground/inverted", then root = underground leaf in this language. This elegant compound would explain why ctros (a cros-variant) appears ubiquitously in root paragraphs — it is the language's way of describing the underground organ by analogy to the aerial one. Medieval Arabic pharmacopoeia used this same conceptual framework.
The lexicon now contains 7 confirmed/probable items from the first pass and 11 new candidates from deep token decomposition. New items are highlighted. All transcriptions approximate — verify against EVA corpus at voynich.nu.
| # | Token | Position | Folios | Proposed Meaning | Latin Analogue | Status |
|---|---|---|---|---|---|---|
| 1 | cros | ¶2 leaf ¶ | f.4, f.9, f.18 | Leaf — base noun (smooth/oval) | folium | |
| 2 | gt-cros | ¶2, spiky pages only | f.5, f.14 | Toothed/serrated leaf — gt- = sharp modifier | folium dentatum | |
| 3 | goll- / gottl- | ¶3 and ¶2 | f.7, f.9, f.25 | Large / fleshy — size qualifier (revised: not root-specific) | magnus / crassus | |
| 4 | otl- / ot- | ¶2 and ¶3 | f.4, f.7, f.12, f.20 | Rounded / globular — shape word (revised: not "flower") | globosus / rotundus | |
| 5 | dand | all paragraphs | All folios | Grammatical particle — "and", "is", copula | et / -que | |
| 6 | llos- / llor- | ¶1 first line | f.4, f.7, f.9, f.12 | Name-marker prefix — genus marker like Latin herba | herba / ushb | |
| 7 | cteg / gtteg | ¶2–3 mid-sentence | f.5, f.14, f.25, f.70 | Verb or connective — "grows", "has", or particle | habet / crescit | |
| 8 | band | ¶1 within name | f.4 | Striped / banded — colour/pattern modifier (f.4 has bicolour banded leaves) | vittatus / striatus | |
| 9 | bos | ¶2–3 | f.4, f.18 | Preposition — "with", "at", "from" | in / ad / cum | |
| 10 | crotlos | ¶2 | f.4 | Leaflet / small leaf — diminutive of cros | foliolum | |
| 11 | das | ¶2 terminal | f.4, f.25 | Second grammatical particle — "also", "or", "too" | etiam / quoque | |
| 12 | ottog | ¶3 | f.4 | Scaly / thick — root texture descriptor (R7 scaly root) | squamosus / crassus | |
| 13 | grot- | ¶2–3 | f.4, f.18, rosette folios | Radially spreading / star-form — applies to rosette leaves AND spreading root crowns | stellatus / ramosus | |
| 14 | -en suffix | ¶2 word-final | f.4, f.9 | Adjectival suffix — "having", "bearing" (present participle type) | -ens / -entis | |
| 15 | oll- prefix | ¶3 | f.25 | "Below / underground" marker — ollcros = root as "underground leaf". Highest priority test. | sub- / radix | |
| 16 | llerod | ¶3 | f.25 | Root base noun candidate — rod sub-unit echoes Latin radix; goll- is the size qualifier on top of this. Highest priority test. | radix | |
| 17 | crowd | ¶3 | f.9 | Spreading / creeping — Welsh crwydro candidate. Test for Celtic language connection. | repens / expansus | |
| 18 | ozorig | ¶2 post-noun | f.4 | Ochre / pale-tan — first colour word candidate. Position after cros suggests adjective following noun. | ochraceus / flavus |
Roots as a scaly architectural platform from which multiple stems emerge (f.4, f.85). No known plant has this morphology. May represent soil, not root.
Faint circular wash of blue or green paint behind several plants (f.5, f.6, f.8, f.11, f.17). Could mark a plant category, indicate aquatic habitat, or be an undercoat wash.
Alternating horizontal red and green stripes with dotted white margins. No known plant produces this. Possibly artistic licence for an Aloe or Bromeliad species.
Roots with impossible bilateral or rotational symmetry (f.15 Y-fork, f.25 radial crown). May reference the doctrine of signatures or alchemical symbolism.
f.26, f.45, f.85, f.86, f.97 all show 2–8 plants sharing a single root structure. May indicate genus grouping, companion planting, or compound medicinal recipes.
Small carefully drawn animals at plant bases (f.30, f.50, f.70). May denote Galenic planet or humour associations — the fox on f.70 possibly linking the plant to Mars/Mercury.
Dense saturated blue-black clusters (f.40, f.70). Unusually vivid pigment — may emphasise toxicity, or specifically indicate the nightshade or bilberry family.
Bud tips drawn in red with small open circles resembling eyes. No botanical parallel. Could represent glands, stipules, or a visual marker for poisonous structures.
Five white comma-shaped marks arranged in an arc across a large spiky leaf — the only internal leaf markings in the manuscript. May represent glands, variegation, or medicinal extraction points. Hypericum (St John's Wort) has oil glands that appear as translucent dots.
On f.21, f.22, f.55, and f.65 the ghost bleed-through from the reverse page does not match the current facing folio. Direct physical evidence of rebinding and page resequencing.
Novel token appearing post-noun in ¶2 beside cros. Position suggests adjective. Best candidate: a colour word for ochre/pale-tan — the colour of the bicolour leaf underside. Needs EVA corpus check.
The -and/-ond/-end suffix variants suggest vowel harmony. If confirmed across 10+ folios in the EVA corpus, the language candidate list narrows to Hungarian, Finnish, or Turkic. This is the next structural test to run.
Confirmed: the green + tan split always follows the leaf midrib, with green adaxial (upper) and tan abaxial (lower). Anatomically accurate, not fantastical.
Red/red-brown paint marks the therapeutically active structure of each plant. Functions as a pharmacological annotation — equivalent to underlining in a modern formulary.
Revised: otl- appears in both flower contexts AND root nodule contexts. It means "rounded/globular" — a shape descriptor, not a plant-organ word. Latin globosus.
Revised: goll- appears in ¶2 on large-leaf pages as well as ¶3 on large-root pages. It means "large/fleshy" — a size/texture qualifier applicable to any organ.
Page 98 is unique: the left margin carries sequential numerals (1 through at least 9) written in a different, later hand alongside the Voynich text lines. No other folio has individual lines of text numbered. If the numbers correspond to a known external numbered herbal or pharmacopoeia entry list, this page could be the cross-reference key the manuscript has been missing for 600 years. Immediate priority: compare f.98's line-number sequence against the plant indices of the major 15th-century herbals — Circa Instans, Herbarius Latinus, and the Tractatus de Herbis.
A full sweep confirms that red/red-brown pigment is applied almost exclusively to: (a) fleshy roots on plants where the root is the primary medicine, (b) stems on latex-producing plants, (c) flower parts when the seed or flower is the active ingredient. Red does not appear decoratively — it functions as a pharmacological annotation equivalent to bold or underlining in a modern formulary. The paragraph closest to a red element is the preparation instruction for that structure — a visual anchor independent of language decipherment.
In f.25's root paragraph, the compound ollcros appears — the confirmed leaf base-word cros with an oll- prefix. If oll- means "below / underground / inverted", then the language's word for root is literally "the underground leaf." This elegant compound would explain why ctros (a cros-variant) is so prevalent in root paragraphs. Medieval Arabic pharmacopoeia used the same conceptual framework — roots described as inverted or underground organs of the leaf. The llerod token in the same paragraph may be the dedicated root noun, making ollcros a descriptive compound used alongside it.
The suffix variants -and, -ond, -end across multiple folios are consistent with vowel harmony — where suffixes change their vowel to match the dominant vowel of the root. This feature is diagnostic for Hungarian, Finnish, Turkish, and related language families. It does not appear in Latin, Welsh, or any Romance language. If confirmed via EVA corpus analysis across 10+ folios, the language candidate pool narrows substantially.
The ¶1 name of f.4 — crobandos — decomposes as cro- (leaf base) + band (striped) + -os (suffix). The plant's name appears to literally mean "the striped-leaf [plant]" — which perfectly describes f.4's bicolour banded leaves. This suggests the manuscript names plants by their most visually distinctive feature, exactly as medieval herbals did (e.g. Pulmonaria = "lung-wort" named for its lung-shaped spotted leaves).
Deeply divided pinnate leaves in green with red midribs, red berry-like flower clusters at stem tip, and a long straight orange-red taproot. The most convincing Apiaceae illustration in the manuscript apart from f.70.
Dense white floret dome with blue flowers emerging from the surface, three hollow stem-tubes descending to a spiky root collar. The hollow stems are anatomically accurate for Brassicaceae.
Large overlapping-scale structure in green and tan/brown with a single blue flower at the crown. The only non-flowering plant in the botanical section.
Strap-like leaves with red midribs, a bulbous blue seed capsule with a red stigma-ray eye at the top, and a coiling root. The stigma-eye is diagnostic for Papaver.
Paired oval leaves in alternate arrangement, a large brilliant blue corolla with a red-dotted white centre, and a small fibrous taproot. Near-certain gentian — one of the most prized Alpine medicines.
Large crenate blue-green leaves, a flat-topped canopy of hundreds of tiny blue seed-dots, and two large elongated red-brown tubers at the base. The densest flower-mass illustration in the entire manuscript.
| Folio | Description | Interpretation |
|---|---|---|
| f.26 | Two stems, stacked leaf whorls (green + tan), shared fibrous taproot, small blue flowers at left stem apex | Two varieties or close relatives sharing habitat |
| f.45 | 3–4 separate stems with palmate leaves and blue 5-petal flowers, arising from one elaborate scrolling root platform | Plant family or genus grouping — largest multi-plant folio. Likely a compound remedy. |
| f.85 | 7–8 small plants sharing a dense root bed with long wispy root fibres | Probable compound remedy or genus catalogue |
| f.86 | Two clearly distinct species side-by-side: narrow linear leaves + massive curling leaves with caterpillar seed pods | Companion plants or combined preparation |
| f.97 | Two round leaf masses on separate stems from one sinuous root, scattered red flowers, three elongated side-by-side tubers | Shared-root family; tuber trio may indicate cultivar variants |
f.7 → 3 | f.21 → 10 | f.23 → 11 | f.45 → 23 | f.65 → 33 | f.75 → 38/39 | f.87 → 44 | f.95 → 48 | f.97 → 49
Added by two distinct later hands — indicating two separate cataloguing events and at least one rebinding.
F7 Catkin/Caterpillar — segmented ovoid seed pods (f.28, f.86). F8 Vine/Spiral stem — corkscrew habit consistent with Convolvulus bindweed (f.22). F9 Scale-cone — overlapping scales consistent with pine cone (f.24). Total flower types: 6 → 9.
In ~85% of folios the lower paragraph is positioned near the root drawing, and the goll- cluster (large/fleshy qualifier) and llerod (root noun candidate) appear there but rarely higher up. The paragraph position encodes root description or root preparation instructions.
The gallows-initial paragraph is always the shortest block. The llos- / llor- prefix opens most ¶1 lines — a genus marker like Latin herba. The second token in ¶1 varies uniquely per folio and is likely the plant name. The name on f.4 (crobandos) appears to literally mean "the striped-leaf plant" — naming by most distinctive visual feature.
The plant on f.70 is nearly identifiable as Conium maculatum (hemlock). Its ¶1 opens with gtaucros — gt- (sharp) + -au- infix + cros (leaf) — likely "deeply divided leaf", matching hemlock's dissected foliage precisely. If confirmed, the full ¶1 string gtaucros ctrottes bdondas is the Voynich word for hemlock.
If f.27 is Papaver somniferum, its text contains Voynich vocabulary for the most potent pain medicine of the medieval period. Opium preparation required precise dosage instructions — making f.27's paragraphs the most likely to contain measurable pharmaceutical language including numbers and units.
Sequential margin numbers on f.98 are the most significant structural anomaly in the volume. Priority comparison targets: Circa Instans, Herbarius Latinus, Tractatus de Herbis, Avicenna's Canon, and Ibn al-Baitar's pharmacopoeia.
The prefix-modification system (cros → gt-cros; goll- as size qualifier; oll- as underground marker; grot- as radial-spreading marker) is the behaviour of a real agglutinative language. Simple substitution ciphers do not produce morphologically productive prefix systems. The vowel harmony evidence in -and/-ond/-end further narrows the candidate language pool to Finno-Ugric or Turkic families.
Confirmed: red marks the therapeutically active plant structure. The paragraph closest to a red element is the preparation instruction for that structure — a direct visual anchor for interpreting text position, independent of any knowledge of the language itself.
Test all 18 lexicon items against the EVA transcription database (voynich.nu). Group EVA tokens by 8 leaf types (L1–L8), 7 root types (R1–R7), 9 flower types (F1–F9). Priority tests: (a) confirm llerod as root noun across ≥5 folios; (b) confirm oll- prefix as underground marker; (c) test vowel harmony in -and/-ond/-end across all folios; (d) test crowd for Welsh phonetic correspondence.
f.7 → 3 | f.21 → 10 | f.23 → 11 | f.45 → 23 | f.65 → 33 | f.75 → 38/39 | f.87 → 44 | f.95 → 48 | f.97 → 49
A rounder darker hand and a lighter angular hand both numbered the manuscript independently. Folio 75 shows both systems. Gaps between file numbers and manuscript folio numbers confirm pages were rebound at least once between the two cataloguing events.
Any public use of this material is to be credited to Shiny Side Out. This content may not be reproduced, re-broadcast, republished, or redistributed in any form without full and clear attribution to shinysideout.com.au. Partial reproduction requires the same attribution. Not for commercial use without written permission.