Date: 2026-07-20 · Authority: R4 owner authorization · Status: adopted in a candidate revision requiring independent review (historical wording; review discharged 2026-07-20 by the fresh-session review and protected merge — docs/project-closure/r4-fresh-fable-review/FABLE-REVIEW-SIGNOFF.md, Decision 0016) · Reopens nothing; the owner's adopted Atharī thesis is unchanged — only evidence status and attribution precision are corrected.
- Attribution defect. The corpus attributed the modern "concrete reason / ideal reason" contrast, and its application to Ibn Taymiyyah, to Evans 2010 and Turner 2022. R3 already recorded that the phrase pair could not be found in either. The supplied compilation shows the application in a different work by different authors.
- Status conflict. The Atharī paper's front matter said classical loci are
[via compilation]and "the printed editions were not independently opened", while the R3 sourcing closeout reported work-level verification, verbatim confirmation of the Majmūʿ vol-12 formula, and no phantom sources. Both cannot govern the same draft. - No single machine-readable source of source-truth: prose, two Markdown ledgers, and a claim-source matrix could drift independently.
references/source-status.yaml is authoritative; the Markdown ledgers are generated or validated from it (scripts/validate_source_status.py, in CI, offline). Each load-bearing claim row carries: claim ID; document/locus; claim level; source; source type; edition/publication metadata; DOI or stable identifier; exact page/volume where available; verification status; verification date; whether wording was directly checked; whether the page number is edition-dependent; school/tradition flag; support/complication/contradiction; notes.
Statuses: PRIMARY_TEXT_EXACT · PRIMARY_WORK_THEME · PRIMARY_LOCUS_EDITION_DEPENDENT · SECONDARY_VERIFIED · SECONDARY_RECONSTRUCTION · COMPILATION_MEDIATED · INFERENCE_CROSS_SOURCE · ORTHEMOLOGICAL_EXTENSION · UNVERIFIED_REMOVE_OR_DOWNSCOPE.
Blanket front-matter source claims are abolished: each Atharī claim carries its own status, and the paper's front matter points to the registry instead of asserting one status for all loci.
Verified this pass:
- Enis Doko and Jamie B. Turner, "Islamic Religious Epistemology," ch. 10 in The Cambridge Handbook of Religious Epistemology, eds. Fuqua, Greco & McNabb, Cambridge University Press, 2023, pp. 148–162, DOI
10.1017/9781009047180.013— the verified modern application to Ibn Taymiyyah.SECONDARY_VERIFIED. Doko must be credited; this is not a Turner-solo claim. - C. Stephen Evans, Faith Beyond Reason: A Kierkegaardian Account, 1998 — the reported source of the distinction. Two 1998 editions exist (Edinburgh UP; Eerdmans, Grand Rapids); Doko & Turner cite the Eerdmans edition and give "(1998, 94)". The page locus could not be independently verified — no accessible copy, snippet, or third-party quotation carrying the page was found, and the two editions may paginate differently. Status
SECONDARY_RECONSTRUCTION, cited as reported by Doko & Turner, with the page never asserted as directly verified. Research residual RR-2 carries the trigger. - Turner 2022, "Ibn Taymiyya on theistic signs and knowledge of God," Religious Studies 58(3):583–597, DOI
10.1017/S0034412521000159— retained only for the natural-signs/theistic-signs claims it actually supports, not as the source of the phrase pair. - Turner 2023, "Ibn Taymiyya's 'Common-Sense' Philosophy," in Pluralizing Philosophy's Past, Springer, pp. 197–212, DOI
10.1007/978-3-031-13405-0_14— a book chapter, 2023 (the corpus's "2022" was wrong). - Turner 2021, "An Islamic Account of Reformed Epistemology," Philosophy East and West 71(3):767–792 — distinct record.
- 2013 McGill PhD dissertation (Institute of Islamic Studies).
- 2020 Brill monograph Ibn Taymiyya on Reason and Revelation, IPTS 111, ISBN 978-90-04-41285-9 (hbk) / 978-90-04-41286-6 (open-access e-book), a substantial revision of the dissertation.
- 2018 "From Legal Theory to Erkenntnistheorie: Ibn Taymiyya on Tawātur as the Ultimate Guarantor of Human Cognition," Oriens 46(1–2):6–61, DOI
10.1163/18778372-04601002— independent article, previously cited via the compilation.
The uploaded compilation is never cited as though it were one authored monograph or the original publication of any included work. Its own Section 1 is the Doko & Turner chapter; claims read from it are labeled COMPILATION_MEDIATED until confirmed against the publisher's version.
Direct claims about ʿaql ṣarīḥ, fiṭrah, necessary knowledge, mental universals vs external particulars, tawātur, corruption mechanisms, and objective truth with varied acquisition paths are verified against a primary locus where accessible, else labeled SECONDARY_RECONSTRUCTION — with no paraphrase carrying false precision. The Darʾ locus queue is research residual RR-3.
The corrected 20:11–12 distinction stands and the whole registry is revalidated after any change. For hadith and classical quotations, exact wording verification is distinguished from work/theme-level verification by status, and edition-dependent paginations are flagged per row rather than globally.
The Atharī paper's blanket front-matter claim is replaced by a per-claim pointer; no creed position changes. PRIMARY_TEXT_EXACT is reserved for loci whose wording was directly checked (the Qurʾānic loci, Muslim 2708, the Majmūʿ vol-12 formula). CI validates the registry and the prose-to-registry agreement; no live web request runs in CI.
The body above stands. Two corrections:
The body said references/source-status.yaml is authoritative for every load-bearing claim. It is not: the registry carries a selected set of rows while the manuscript and the sourcing ledgers carry many more claims. Rather than leave an overstatement standing (or manufacture rows to match a slogan), the registry is rescoped to what it actually covers, and the coverage is now machine-declared in the file itself under covered_claim_families: the concrete/ideal-reason chain (CIR-*), the El-Tobgui records (ELT-*), the enumerated Atharī per-claim rows (ATH-*), the sequential latent-variable related work (LAT-*), and this project's own extension (EXT-*). A companion not_covered block names the surfaces that carry the rest — docs/sourcing/SOURCING-LEDGER.md, the companion ledger, and references/quran-loci.yaml — and states explicitly that absence from this registry is not a claim that a source is unverified; it means the registry does not adjudicate it. scripts/validate_source_status.py enforces the scope: every row id must belong to a declared family, and no row may sit outside one.
Stated in the registry header and enforced by the validator's own docstring: the checks are offline record-shape and internal-agreement checks. They confirm required fields, vocabulary membership, status/evidence-access consistency, family membership, and agreement with paper prose. They do not verify that a source is true, that a quotation was actually read, or that a DOI resolves. No check in this file is evidence about the world.
CIR-1 was SECONDARY_VERIFIED while its own notes recorded that publisher metadata was verified (Crossref, Cambridge Core) but the chapter wording was read only through the supplied compilation and the publisher's text was not opened (403). Those are two claims with two evidence accesses, so they are now two rows: CIR-1 covers bibliographic existence/authorship/DOI/pages at SECONDARY_VERIFIED (with an explicit status_scope field saying so), and CIR-1W covers the exact in-chapter application/wording at COMPILATION_MEDIATED, with a recorded promotion trigger (direct publisher full-text access) and an explicit statement that no corpus claim rests on it alone.
The body above stands except one sentence of the Consequences: "PRIMARY_TEXT_EXACT is reserved for loci whose wording was directly checked (the Qurʾānic loci, Muslim 2708, the Majmūʿ vol-12 formula)."
The Majmūʿ al-Fatāwā vol-12 words/voice formula does not hold PRIMARY_TEXT_EXACT. The R4 fresh-session review downgraded row ATH-3 to COMPILATION_MEDIATED (the registry contradicted the companion's own [via compilation] label on the same locus, and no printed edition or authoritative full-text rendering was opened; the weaker claim was taken — see the R4 amendment above and docs/project-closure/r4-fresh-fable-review/POST-SUBSTITUTION-HUNK-REVIEW.md §C4). This sentence survived that downgrade and thereby restated the superseded status; the independent post-merge audit (finding B2) caught the contradiction.
Current statement: PRIMARY_TEXT_EXACT is reserved for loci whose wording was directly checked — presently the Qurʾānic loci (references/quran-loci.yaml, row ATH-1) and Muslim 2708 (row ATH-2). Row ATH-3 stands at COMPILATION_MEDIATED with its recorded promotion trigger (direct edition or authoritative full-text access). The R5 session likewise did not open such a text and made no upgrade. The adopted Atharī doctrinal position is unchanged; this is an evidence-access correction only. scripts/validate_source_status.py now checks the registry-to-prose agreement bidirectionally (a row claiming exact against compilation-labeled prose, AND prose or a decision claiming exact against a weaker registry row), with the registry controlling for its declared families.