Concordance Methodology

This page documents the actual methods used to build the concordance between Russell's dissertation references and the manuscript photograph collections. All statistics are drawn from the project database. Uncertainty and provisional status are marked explicitly throughout.

The Problem Russell Presents #

Russell's thesis references folios by signature (e.g., "b6v", "h1r") rather than by page number, since the 1499 edition has no printed pagination. Our manuscript photograph collections, however, use different naming conventions: the Siena images use folio numbers with recto/verso suffixes (e.g., O.III.38_0014r.jpg), while the BL images use sequential numbers (C_60_o_12-001.jpg through C_60_o_12-196.jpg). The core challenge is mapping Russell's 282 signature-based references to the 674 available photographs.

BL vs. Siena Image Sets #

Property	Siena O.III.38	BL C.60.o.12
Images	478	196
Naming	Folio number + r/v suffix	Sequential number only
Edition	1499 Aldine	1545 Aldine (second edition)
Folio mapping	Directly encoded in filename	Requires inference
Matching confidence	HIGH / MEDIUM	LOW (all provisional)

The BL copy is the 1545 edition, not the 1499. The signature map is based on the 1499 collation formula. While the 1545 edition largely follows the same structure, any pagination differences between editions would make sequential-number-to-folio inference unreliable. All BL matches must be treated as provisional.

How the Signature Map Works #

The signature map is a deterministic lookup table of 448 entries generated from the 1499 collation formula: a–z⁸ (omitting j, u, w), A–F⁸, G⁴. Each quire has 8 leaves (except G with 4), and each leaf has recto and verso sides. The map converts any valid signature (e.g., "b6v") to a sequential folio number and vice versa.

This map is generated by build_signature_map.py and is fully deterministic. It is correct for the 1499 edition. Its applicability to the 1545 edition is assumed but not verified.

How Dissertation References Were Extracted #

Russell's thesis PDF was processed by extract_references.py using PyMuPDF for text extraction and regular expressions for signature pattern matching. The script extracted 282 references, distributed across manuscripts:

Manuscript	References
Buffalo RBR	59
Inc.Stam.Chig.II.610	55
C.60.o.12	50
INCUN A.5.13	42
Modena (Panini)	29
O.III.38	28

How Image Filenames Were Parsed #

catalog_images.py parses image filenames from two manuscript collections into structured records with folio number, side (recto/verso), and page type. The Siena images encode folio information directly (O.III.38_0014r.jpg = folio 14 recto). The BL images use sequential numbers that do not directly encode folio information.

How Matching Was Performed #

1 Signature Lookup

Convert each dissertation reference's signature (e.g., "b6v") to a folio number using the signature map.

2 Manuscript Filter

Filter images to the manuscript shelfmark specified in the dissertation reference.

3 Folio Match

Match the computed folio number to images with the same folio number and side.

4 Fallback Cross-Match

If no direct match, attempt cross-manuscript matching (lower confidence).

5 Confidence Assignment

Assign HIGH (exact Siena match), MEDIUM (folio-based), or LOW (BL sequential inference).

Where Confidence Is High vs. Low #

Confidence	Matches	Description
HIGH	431	Siena images with explicit folio+side in filename, matched by signature lookup
MEDIUM	0	Siena images matched by folio number (cross-manuscript or side ambiguous)
LOW	0	BL images where sequential photo number is assumed to equal folio number

By Manuscript

Manuscript	HIGH	MEDIUM	LOW
Siena O.III.38	392	0	0
BL C.60.o.12	39	0	0

Why BL Matches Are Provisional #

Critical caveat: All 39 BL matches are classified as LOW confidence. The BL photographs are sequentially numbered (001–196) without folio labels. Our matching pipeline assumes that photo N corresponds to folio N, but this assumption is unverified. The BL copy is the 1545 edition, not the 1499 edition from which the signature map was built. Any differences in pagination, inserted leaves, or missing leaves between editions would invalidate this mapping. These matches require manual verification against the physical photographs.

What Human Review Is Still Needed #

Verify BL photograph-to-folio correspondence against physical or high-resolution images
Confirm that the 1545 edition follows the same collation as the 1499
Spot-check MEDIUM confidence Siena matches for side (recto/verso) accuracy
Review hand attribution for edge cases where multiple hands annotate the same folio
Validate marginal text transcriptions against original manuscript images

How This Methodology Supports Future Scholarship #

The concordance pipeline produces a structured, queryable dataset linking Russell's close reading to digital images. Once the BL matches are verified, this enables:

Folio-level browsing of annotations alongside manuscript images
Systematic comparison of annotation density across copies
Identification of folios that attracted multiple annotators
Cross-referencing between annotations and the dictionary of terms
Future multimodal analysis using computer vision on manuscript images

Related Dictionary Terms

Signature Folio Recto Verso Quire Collation Incunabulum Annotator Hand Marginalia 1499 Edition 1545 Edition

Alchemical Hands Essay James Russell Concordance Methodology (Doc) match_refs_to_images.py build_signature_map.py

Alchemical Hands in the Hypnerotomachia Poliphili