Alchemical Hands in the Hypnerotomachia Poliphili

Marginalia, Scholarship & Reception

← All Documents

Concordance Methodology

docs/archive/HPCONCORD.md — 1,027 words

HPCONCORD: Concordance Methodology

How We Built the Marginalia-to-Image Concordance

Problem

James Russell's PhD thesis (Durham, 2014) references specific folios of six annotated copies of the Hypnerotomachia Poliphili by their signature marks (e.g., b6v, e1r, y7r). We have photographs of two of these copies — BL C.60.o.12 (196 images) and Siena O.III.38 (478 images) — but the photos are numbered sequentially (e.g., C_60_o_12-014.jpg, O.III.38_0022r.jpg), not by signature. The concordance bridges these two reference systems.

Step 1: Signature Map (build_signature_map.py)

The 1499 HP uses a standard Aldine signature scheme: gatherings of 8 leaves (16 pages), labeled a–z then A–F. Each leaf has a recto (front) and verso (back) side.

We built a complete mapping of 448 signature positions:


a1r → folio 1, recto    (image 001)
a1v → folio 1, verso    (image 001)
a2r → folio 2, recto    (image 002)
...
b1r → folio 9, recto    (image 009)
...
F3r → folio 227, recto  (image 227)

The key insight: the BL sequential photo numbers correspond to folio numbers (C_60_o_12-014.jpg = folio 14 = signature b6). The Siena facsimile uses explicit recto/verso notation (O.III.38_0014r.jpg / O.III.38_0014v.jpg).

Important caveat: The BL copy is the 1545 edition, which has a slightly different layout than the 1499. The signature scheme is the same, but page content may differ by a few leaves. Matches are therefore MEDIUM confidence until manually verified against the actual image content.

Step 2: Reference Extraction (extract_references.py)

We extracted 282 signature references from Russell's dissertation PDF using regex pattern matching:


# Matches: (a4r), (c7v), b6v:, etc.
SIG_PATTERN = re.compile(
    r'(?:\(([a-zA-Z]{1,2}\d[rv])\))'   # parenthesized
    r'|(?:\b([a-zA-Z]{1,2}\d[rv]):)'    # with colon
    r'|(?:\b([a-zA-Z]{1,2}\d[rv])\b)'   # bare
)

False positives (common English words + digit + r/v, like in1r) were filtered by stopword list.

Each reference was tagged with:

Step 3: Image Cataloging (catalog_images.py)

Scanned both image directories and inserted 674 image records:

Step 4: Matching (match_refs_to_images.py)

Joined signature references → signature map → image catalog to produce 610 matches:


-- Example match chain:
-- Russell p.157 references "b6v" in BL copy
-- signature_map: b6v → folio 14, verso
-- images: C_60_o_12-014.jpg (folio 14)
-- → MATCH (FOLIO_EXACT, MEDIUM confidence)

All automated matches are flagged needs_review = 1 for manual verification.

Step 5: Hand Attribution (add_hands.py)

Created 11 annotator hand profiles across the six copies, with two flagged as alchemists:

CopyHandAttributionAlchemist?School
BL C.60.o.12ABen JonsonNo—
BL C.60.o.12BAnonymous**Yes**d'Espagnet
BuffaloAAnon. (Jesuit?)No—
BuffaloBAnon. (Jesuit?)No—
BuffaloCAnonymousNo—
BuffaloDAnonymousNo—
BuffaloEAnonymous**Yes**pseudo-Geber
ModenaPrimaryBenedetto GiovioNo—
ComoPrimaryBenedetto GiovioNo—
VaticanPrimaryAlexander VIINo—
SienaPrimaryAnonymousNo—

Hand attribution rules map specific signature references to hands based on:

1. Explicit attribution in Russell's text (e.g., "Hand B labels b6v with alchemical ideograms")

2. Page range within the dissertation chapter (alchemical content starts p.155 in Ch. 6)

3. Single-hand copies (Modena, Como, Vatican, Siena) get all refs attributed to their primary hand

193 references were attributed to specific hands. 89 remain unattributed (mostly from intro/methodology chapters where Russell discusses copies in general terms).

Step 6: Bibliography and Timeline (add_bibliography.py)

Extracted 310 works from Russell's bibliography, of which 58 HP-specific entries were inserted into the database. Of these:

Created 29 scholar profiles and 39 timeline events spanning 1499–2020.

Database Schema Summary


manuscripts (2)          → physical copies with image directories
images (674)             → individual photographs
signature_map (448)      → signature → folio number mapping
dissertation_refs (282)  → folio references extracted from Russell
matches (610)            → ref ↔ image joins
annotator_hands (11)     → identified annotator profiles
bibliography (58)        → cited HP scholarship
scholars (29)            → scholar profiles
scholar_works (24)       → scholar ↔ work links
timeline_events (39)     → reception/scholarship timeline

Key Queries


-- All alchemist annotations with matched images
SELECT h.attribution, h.school, r.thesis_page, r.signature_ref,
       i.filename, r.context_text
FROM annotator_hands h
JOIN dissertation_refs r ON r.hand_id = h.id
JOIN matches m ON m.ref_id = r.id
JOIN images i ON m.image_id = i.id
WHERE h.is_alchemist = 1
ORDER BY h.manuscript_shelfmark, r.thesis_page;

-- Scholarship gap: works we need
SELECT author, title, year, journal_or_publisher, topic_cluster
FROM bibliography
WHERE in_collection = 0 AND hp_relevance = 'DIRECT'
ORDER BY year;

-- Timeline of HP reception
SELECT year, year_end, event_type, title, description
FROM timeline_events ORDER BY year;

Known Limitations

1. BL 1545 vs. 1499 offset: The BL copy is the 1545 edition; our signature map is based on the 1499. Content positions may differ by a few leaves. All BL matches are MEDIUM confidence.

2. Photo coverage: The 196 BL photos may not cover every annotated folio. Russell photographed selectively, focusing on the most interesting marginalia.

3. Hand attribution gaps: 89 of 282 references lack hand attribution. These are mostly introductory cross-references where Russell discusses multiple copies in general terms.

4. Buffalo images missing: We have no photographs of the Buffalo copy (5 hands including the pseudo-Geber alchemist). The matched BL/Siena images for Buffalo refs show the same folio in a different copy, not the Buffalo marginalia themselves.

5. Automatic matching assumes sequential numbering = folio numbering. This holds for both collections but should be verified against the first and last few images.