Live: https://ignatian-research.netlify.app/
A working corpus of the old Society of Jesus (1540–1773), organised in five lines. At the centre, the seven core works: the writings of Ignatius of Loyola himself — the Spiritual Exercises, the Constitutions with their Complementary Norms, the dictated memoir known as A Pilgrim's Testament, the Spiritual Diary, and the letters of 1524–1547 — and, around them, the texts in which his spirituality became the practice of an order: the Official Directory of 1599 on giving the Exercises, and the Memoriale of Blessed Peter Faber, the first companion, discernment kept as a daily journal from 1542 to 1546. Around those stand the programme modules that carry the corpus out into the world and its arguments: the school of discernment carried to its end — Nadal's annotations on the Examen, Rodríguez's Of Conformity to the Will of God, and the Abandonment to Divine Providence attributed to Caussade — Xavier's letters from India and Japan, the emblems of the Imago primi saeculi of 1640 with their Dutch mirror, Friedrich Spee (the Cautio Criminalis and the Trutznachtigall), and the counter-voices — the forged Monita secreta, Pascal's Provinciales, and the brief Dominus ac Redemptor of 1773 with which the corpus closes. The apparatus indexes them by their canonical numbering, traces the vocabulary that migrates between them, and lets you put a question to the corpus with the evidence attached.
It is a working instrument, not a publication. It was built for one reader's use and is opened here in case it is useful to others; the limits are stated at length rather than buried.
The site opens with a stand-alone scholarly introduction (Ignatiana: An Introduction to a Research
Apparatus for Ignatian Spirituality, September 2026), reachable from the navigation as
Introduction. It presents the five lines, the argument that runs through them, the threshold of
1773, and the apparatus's method and limits, with a full reference list; on the site its first
mention of each work links into the corresponding reader. The essay lives in
data/introduction.json (editorial matter, CC BY 4.0) with the author's manuscript deposited in
docs/; it is maintained alongside the site as modules ship.
This is the constraint everything else follows from. Ignatius died in 1556, Favre in 1546, and the writings of the early Society are long out of copyright. The translations that make them readable are mostly not.
| Work | Translation | Rights | In this repository |
|---|---|---|---|
| Letters and Instructions, vol. I | D. F. O'Leary, 1914 | public domain | full text included (24 letters) |
| The Official Directory of 1599 | Latin: Monumenta Ignatiana, 1919 · English: W. H. Longridge, 1919, or unofficial working translation made for this repository | all public domain (US, pre-1930); working translation dedicated CC0 | full bilingual text included (287 paragraphs; Longridge for 285 of them) |
| The Spiritual Exercises (parallel edition) | Spanish Autograph & Vulgata 1548: Monumenta Ignatiana, 1919 · English: Elder Mullan, 1914 | all three public domain (US) | full trilingual text included (370 paragraphs, Exx [1]–[370]) |
| The Spiritual Exercises — commentary | W. H. Longridge, SSJE, 1919, with his seventeen Additional Notes | public domain (US, pre-1930) | included as a switchable layer of the trilingual reader |
| Memoriale of Bl. Peter Faber | Latin: editio princeps (Bouix, Paris 1873) · English: unofficial working translation made for this repository | Latin public domain (US, 1873); translation and numbering dedicated CC0 | full bilingual text included (409 paragraphs, with the appendix of nine letters and counsels) |
| The Spiritual Exercises | George E. Ganss, 1992 | in copyright | derived data only |
| Constitutions & Complementary Norms | ed. John W. Padberg, 1996 | in copyright | derived data only |
| A Pilgrim's Testament | Parmananda R. Divarkar, 1995 | in copyright | derived data only |
| The Spiritual Diary | Joseph A. Munitiz, 1987 | in copyright | derived data only |
The trilingual Exercises edition does not replace Ganss: his 1992 translation remains a locked, unlockable work, and while it is not opened the concordance searches the public-domain edition instead, every hit exact to its Exx [n]. The [1]–[370] grid itself is editorial — it was introduced in 1928 and appears in none of the public-domain printings, so it was assigned here to the older texts and validated for completeness (370 of 370, no gaps, all three languages).
The Directory's English side offers two texts. The default is W. H. Longridge's translation of 1919 (London: Robert Scott) — a human, public-domain translation whose chapter and paragraph numbering matches the Latin throughout (he omits only the title page and the Florentine censor's note at the end of chapter XL). The alternative deserves a plain label: a machine-generated working translation, made directly from the 1919 Latin and consulting no copyrighted translation — honest, consistent in its key terms, and without any ecclesiastical or scholarly authority. In either case, cite the Latin.
From the same 1919 Longridge volume, his running commentary on the Exercises — segmented into blocks
and keyed editorially to the [1]–[370] grid — and his seventeen Additional Notes are included as a
switchable layer of the trilingual reader (data/longridge_exx.json, loaded on demand).
For the four in-copyright translations this repository carries no running text at all: only paragraph-level citation anchors, counts, co-occurrence edges, name registers, and editorial matter written for this site. All rights in those translations rest with their publishers and translators — principally the Institute of Jesuit Sources and Inigo Enterprises.
Full-text functions work against a copy you own. Your PDF is read in the browser by pdf.js, stored in that browser's IndexedDB, and never uploaded. Only the passages local retrieval selects for a specific question are sent onward.
The programme modules are a different case: each is public-domain source in its own right, so its running text is shipped, complete, on its own citation grid. They are Pascal's Lettres provinciales (Letters V, VII, X, M'Crie's 1856 translation); the Monita secreta in Brownlee's 1857 bilingual printing; the brief Dominus ac Redemptor of 1773 (Latin after Theiner 1852, with a working translation); four letters of Francis Xavier (Coleridge's 1872 biography); eleven emblems of the Imago primi saeculi (Antwerp 1640) with the Dutch Af-Beeldinghe of the same year and page images; and Friedrich Spee — the Cautio Criminalis (two dubia, Rinteln 1631) and the Trutznachtigall (two poems, Cologne 1654). Where a module carries a working translation it is this site's own, made from the original and marked as unofficial; the sources are listed in SOURCES.md. These modules are part of the concordance and the citation-bound dialogue, but not of the linguistic statistics, which describe the core corpus only.
If you hold rights in one of these editions and consider anything here to exceed what derived data and scholarly citation permit, write to the address in the site's legal notice and it will be dealt with promptly.
A canonical citation index. 1,887 anchors tying paragraph numbers to the page on which they begin, so
that a concordance hit reads SpEx [23] or Const [134] rather than a page number nobody cites.
Coverage by work: Exercises 369 of 370, Constitutions 812 of 827, Pilgrim's Testament 101 of 101, Diary
486 of 490, and 119 across the six directory documents.
A concordance across the whole apparatus at once, with keyword in context, hit distribution per text, and a canonical citation on every line. The public-domain core (the Exercises, the Letters, the Directory of 1599 and the Memoriale) and every shipped programme module — from Pascal to the brief of 1773 and the Imago's emblems — are always searchable; the four in-copyright works join once you have opened your own copies.
The 1914 letters in full, searchable and quotable without any copy of your own, with recipient, place and date for each — as are all the programme modules.
A discernment lexicon: 27 terms of the vocabulary of consolation, desolation and election traced across the corpus, with their distribution — which is where the differences between the works become visible rather than merely assertable.
Keyness, registers and networks: for each core work, the lemmas that mark it against the rest by log-likelihood; persons, places, the memoir's itinerary, and a co-occurrence graph of the correspondence.
Five reading paths and a coda. Guided routes through the corpus, each station with a reason and a guiding question — the Four Weeks, the discernment of spirits, the election, governance by letter, and an iconographic path through the Imago's emblems; a closing note owns the corpus's boundaries and the way it was made.
A citation-bound dialogue. Questions are answered only from passages retrieved from the corpus (and, for the locked works, from your own copies), with a canonical citation required on every substantive claim. It refuses to assert what the passages do not carry. It is still a language model and can misread; every citation it gives can and should be checked.
A few megabytes of JSON, all of it derived, editorial, or public-domain source text.
| File | Contents |
|---|---|
works.json |
the seven core works: translator, rights, citation form, body range, sections, measures |
programme.json |
the programme modules — line, status, source, and (for shipped ones) data file and citation form |
exercitia.json, directorium.json, memoriale.json |
the shipped full-text core editions (trilingual Exercises, bilingual Directory of 1599, bilingual Memoriale) |
pascal.json, monita.json, dominus.json, xavier.json, imago.json, spee.json, trutz.json, nadal.json, rodriguez.json, caussade.json |
the shipped programme modules, paragraph by paragraph on their own citation grids |
anchors.json |
1,887 canonical paragraph anchors, each with PDF page and printed page |
letters.json |
the 24 public-domain letters in full, with recipient, place, date |
introduction.json |
the introductory essay (editorial matter, CC BY 4.0); manuscript in docs/ |
introductions.json, sections.json |
editorial orientation, textual history, key passages, internal divisions |
lexicon.json |
27 discernment terms traced across the corpus |
glossary.json |
39 institutional, practical and philological terms |
keyness.json, terms.json, network.json |
log-likelihood profiles, distributions, co-occurrence graph |
persons.json, places.json, itinerary.json |
registers and the memoir's route |
discernment.json, corpus.json |
term-family distributions and corpus totals |
longridge_exx.json, longridge_dir.json |
Longridge's 1919 commentary and Directory translation |
anchors.json is the piece most likely to be useful on its own. Its shape is
{ workId: [ { n: paragraph, p: pdfPage, s: printedPage }, … ] } — a plain concordance between the
canonical numbering and the pagination of the reference editions, which is tedious to build and easy to
reuse.
Drop one or more PDFs into Open from your own copies. Recognition is automatic but not blind: each file is scored against all five reference editions on the evidence of its front matter — the title of the edition, the translator or editor, phrases proper to that volume, and phrases belonging to a different volume, which count against — corroborated by the page count of the reference edition. A work is opened only when one candidate wins by a clear margin; otherwise the panel asks which work it is, with the file already read.
This matters because the volumes quote each other. The Constitutions name the Exercises on their first pages, and Ganss's name stands in the front matter of three editions because the 1996 Constitutions print his translation — so a signature that counts shared phrases alone will file Padberg's Constitutions and Munitiz's Diary under the Exercises. Each work can also be opened from a named file directly, which overrides recognition, and removed again individually. Where a page count diverges from the reference edition the panel says so, rather than letting a divergent printing be silently mis-cited.
Editions the anchors were built against: Ganss (IJS, 1992); Padberg (IJS, 1996); Divarkar (IJS, 1995); Munitiz, Íñigo: Discernment Log-Book (Inigo Enterprises, 1987); O'Leary, ed. Goodier (B. Herder / Manresa Press, 1914). The Directory of 1599, the trilingual Exercises, Favre's Memoriale and every programme module need no anchors: their texts are shipped paragraph by paragraph, so a hit in them is exact to the paragraph without a page lookup. A different printing of a locked work will still search; its pages will not line up.
Set out in full on the site's Method, sources and limits page, and worth reading before citing anything. In brief: the 1914 letters are an optical scan whose errors are left visible rather than silently emended; name recognition is automatic and over-recognises liturgical capitalisation while under-recognising Spanish and Basque names; the Complementary Norms are not separated out as an independent citation series; the readability and complexity measures this site once displayed were removed by design, because in a corpus read through five different translators they profiled the translators, not the texts (the keyness profile, which compares like with like inside one reference corpus, is what remains); and the itinerary follows the memoir's own account, which is a narrative composed thirty years after the events and shaped for a purpose.
A static site with one Netlify Function. No build step, no dependencies — the folder deploys as it stands.
python -m http.server 8000 # then open http://localhost:8000
Everything except the dialogue works locally: retrieval, concordance, registers and charts all run in the browser.
To deploy your own copy, point Netlify at a fork, or drag the folder onto https://app.netlify.com/drop.
netlify.toml must stay in the deployment root.
The dialogue calls /.netlify/functions/dialogue, which uses Netlify's AI Gateway: Netlify injects
ANTHROPIC_API_KEY and ANTHROPIC_BASE_URL into the function, so no key of your own is needed and usage
is billed in Netlify credits. It requires at least one production deploy and AI features enabled for the
team, and it is bypassed if you set either variable yourself — set your own ANTHROPIC_API_KEY if you
would rather use your own Anthropic account. The model defaults to claude-sonnet-4-6; set DIALOG_MODEL
to change it, with fallback to claude-sonnet-4-5 and claude-3-7-sonnet-latest. Without any endpoint
configured the rest of the site is unaffected.
No key is stored in this repository, and none has ever been committed to it.
index.html shell and navigation
app.js router, data, all views
corpus.js pdf.js reading, identification, index, KWIC, collocation, BM25
viz.js charts and force-directed graph, no external library
dialogue.js dialogue module and PDF export
data/*.json derived data, editorial matter, and shipped editions
assets/imago/ page images of the Imago emblems and the Dutch mirror
assets/spee/ facsimile pages of the Trutznachtigall poems
tools/ build scripts for the shipped editions, and the CI checker
vendor/ pdf.js 4.6.82, jsPDF 2.5.2 — vendored, so no CDN is contacted
netlify/functions/ the dialogue function
Four kinds of material, licensed separately, because one blanket licence would misdescribe at least two of them. The mapping is in LICENSES.md; in short:
- Code — MIT.
- Editorial texts (introductions, glossary, discernment lexicon, section notes, this README) — CC BY 4.0.
- Derived datasets, the anchor table above all — CC0 1.0, a public-domain dedication. A concordance between canonical paragraph numbers and printed pagination is a compilation of facts: laborious to establish, discovered rather than invented. It is released free because a citation index that others cannot build on is of little use.
data/letters.jsonis the 1914 O'Leary translation. It is in the public domain in its own right, and no licence here applies to it — it was never the author's to license.data/exercitia.jsonholds three public-domain texts — the Spanish Autograph and the Vulgata as printed in the Madrid 1919 Monumenta Ignatiana, and Mullan's 1914 translation — plus this repository's editorial work of segmentation and numbering, which is dedicated to the public domain under CC0 1.0 like the other derived data.data/directorium.jsonis two things at once. Its Latin is the 1599 Directory as printed in the Madrid 1919 Monumenta Ignatiana, public domain in the United States; a faithful transcription adds nothing licensable. Its English is the working translation made for this repository, and whatever rights that could attract are dedicated to the public domain under CC0 1.0, so that nobody ever has to ask.data/memoriale.jsonholds Favre's Memoriale as printed in 1873, public domain by age; the working translation and this repository's paragraph numbering are dedicated CC0. The provenance caveat is stated in the file itself and on the method page: the autograph is lost, the 1873 Latin is the Society's transmitted version, and the canonical MF numbers of the 1914 critical edition could not be reproduced.data/longridge_exx.jsonanddata/longridge_dir.jsoncarry Longridge's 1919 commentary, Additional Notes and Directory translation — public domain by age; the editorial keying to the [1]–[370] grid and the paragraph alignment are dedicated CC0 like the rest of the derived data.- The programme modules (
data/pascal.json,monita.json,dominus.json,xavier.json,imago.json,spee.json,trutz.json,nadal.json,rodriguez.json,caussade.json, with the images inassets/imago/andassets/spee/) carry public-domain source text — printings of 1631 to 1919 and, for the images, of 1640 — whose faithful transcription and reproduction add nothing licensable; the editorial segmentation, the working translations made for this repository, and the descriptions are dedicated CC0. Each file states its own source; the whole list is in SOURCES.md. data/pilgrim_profile.json,data/exercised_self.jsonanddocs/carry the author's essays The Pilgrim's Profile (2026) and The Exercised Self and the Ecstatic Image (2026), presented on the site's separate author page (#/author). They are interpretation, not edition, and are © the author, all rights reserved — deliberately outside the open licences of this repository.vendor/carries the upstream licences of pdf.js (Apache-2.0) and jsPDF (MIT).
None of this grants any right in the four in-copyright translations the apparatus describes. No part of their text is in this repository.
Pantaleon Fassbender, Ignatiana: A Research Apparatus for Ignatian Spirituality, https://ignatian-research.netlify.app/, archived at https://doi.org/10.5281/zenodo.22682750 (accessed …).
See CITATION.cff for a machine-readable form. When citing a passage found through the concordance,
cite the printed edition, not this site — the anchors exist precisely so that you can.
Corrections are welcome, and errors in the anchor table especially so: a wrong anchor produces a wrong citation, which is the one failure mode this instrument must not have. Open an issue with the work, the paragraph number and the page you find it on.
Operated by a private individual and unaffiliated with the Society of Jesus, the Institute of Jesuit Sources or any publisher. Privacy notice and legal notice are on the site.