About this map

Treatise Mapping asks which passages of Hume’s Treatise the secondary literature on personal identity actually reads, and which papers find their way to passages the rest of the literature passes by.

The corpus

100 papers from the PhilPapers category on Hume and personal identity, collected in 2020. Their PDFs and full texts are kept privately; this site shows only where each paper cites the Treatise, a short snippet around each citation, and bibliographic details.

Finding citations

Each PDF was converted with docling, with OCR only for scanned papers. A staged detector then looked for citations:

1. Citation patterns (2,966): Norton references like T 1.4.6.4, Selby-Bigge pages like SBN 252, and forms like “(T 252)”.
2. Quote matches (123): a bare page number after a quotation, accepted when the quoted words appear in the Treatise on that page.
3. Model checks (102): the remaining ambiguous page numbers, judged with their context by Gemini on Vertex AI. Most ambiguous numbers turned out to cite other scholars’ books, and were dropped.

3,191 citations were accepted. 27 papers have none, mostly book reviews whose page numbers refer to the book under review.

From citations to paragraphs

Each citation spreads one unit of weight over the paragraphs it covers: a Norton paragraph gets all of it, a section or a Selby-Bigge page shares it among its paragraphs. A paragraph’s shade on the map is the total weight it receives across the corpus.

Focus and surprise

Focus is the share of a paper’s citations on the 15 most-cited paragraphs, mostly T 1.4.6 and Appendix 10 to 21. Surprise asks how unlikely it was that a paper with this many citations would cite a given paragraph at all, and takes the highest value over the paragraphs it cites precisely (with at least 0.3 of a citation, which keeps page-range smearing out). Both are given for the 56 papers with at least 10 citations.

The two are shown separately on purpose. Every single-number novelty measure we tried, including the 2020 site’s “difference from the literature”, TF-IDF and KL divergence, ranked off-topic papers as the most novel.

Themes

Four citation themes come from non-negative matrix factorisation of section-level citation profiles, with 1.4.6 and the Appendix left out because nearly everyone cites them. See Themes.

Sources and versions

The Treatise text and its Norton and Selby-Bigge numbering come fromHume Texts Online, edited by Peter Millican and Amyas Merivale. References follow David Fate Norton and Mary J. Norton’s Clarendon edition (T) and L. A. Selby-Bigge and P. H. Nidditch’s edition (SBN).

The first version of this project (2020) used a simpler extraction pipeline and is archived, as it was, attreatise-mapping-project-ur.outsidethevillage.com.