Methodology

Edition 2026-08. This page states the boundary of the project, the vocabulary it uses, the citation standard it holds itself to, and where it knows it is weak. It exists so that “who are you to adjudicate this” has an answer other than silence.

What this is

This project records what named figures — theologians, pastors, ecumenical councils and confessional documents — held about specific doctrines, with a sourcing tier on every claim and a verbatim quotation wherever one could be obtained and lawfully published.

It is not a comparison of denominations as monoliths. That genre is crowded and mostly well served. The disagreements inside a tradition are the interesting part, and no existing tool lets you ask what Aquinas, Calvin, Wesley and Heiser each held about a single question.

The inclusion boundary

The gate is the Nicene Creed of 381 — affirmation of its content, not subscription to the document, so non-creedal traditions such as Baptists and Churches of Christ are assessed on what they hold rather than on whether they would sign a text.

The Trinity is an absolute requirement. Anything that does not affirm that Jesus is God is outside the boundary.

Chalcedon is a locus, not a gate. This admits the Oriental Orthodox — Coptic, Armenian, Ethiopian, Eritrean, Syriac — who affirm the Trinity and the deity of Christ and dispute how the union is worded.

Modern non-trinitarian groups are outside the boundary: Latter-day Saints, Jehovah’s Witnesses, Oneness Pentecostalism. This is a statement about the scope of this project against a stated criterion. It is not a verdict on anyone’s salvation, and it should not be read as one.

Historical heresies are labelled where doctrine requires it, not catalogued. Arius appears because the Nicene Creed is unintelligible without him. There is no systematic heresy index here.

The problem inside the gate, named before a critic names it

381 is precisely the text on which East and West differ. The Nicene Creed without the filioque is the Orthodox position; every Western tradition in this corpus recites the interpolated version liturgically. So the inclusion criterion adjudicates a schism-defining dispute inside the criterion itself. That is a real weakness in the gate and it is stated here rather than hidden. It is not a reason to abandon the gate; it is a reason to state it precisely and to map the dispute as a locus rather than pretend it is settled.

The best external warrant for the gate: in April 2025 twelve Christian World Communions jointly declared that “The Nicene Creed presents the most authoritative statement of trinitarian Christian faith.” Signatories included the Anglican Communion, the Baptist World Alliance, the Ecumenical Patriarchate, the Lutheran World Federation, the Pentecostal World Fellowship and the World Methodist Council.

The one editorial position this project takes

Chalcedon is held to be correct — one person, two natures, fully God and fully man, without mixture. This is an editorial stance rather than neutral rendering, and it is said openly here instead of being smuggled into the descriptions. It is not a denominational marker: Catholic, Orthodox and every magisterial Protestant tradition hold it. The Oriental Orthodox position is mapped fairly and in its own terms.

Everywhere else, the aim is to state each figure’s position as they would recognise it.

Sourcing tiers

Every position carries one, and it renders on every cell.

Tiers are never upgraded without doing the work. As of this edition, 89% of researched positions carry a verbatim quotation and 1% of established ones do — which is why the tier is a usable proxy for “can I check this myself” rather than a label.

Quotation, and the copyright asymmetry

Two in five positions carry the figure’s own words. Where a quotation exists it is the point of the cell and it renders above the summary.

The rest are honest summaries, and the reasons differ. Calvin can be quoted at length; a living author cannot — that is the copyright asymmetry, and it is real. But it is not the only reason a column is unquoted, and it would be dishonest to imply it were: twelve figures currently carry no quotation at all, and eight of them died long before copyright could apply. For those, the tier says it plainly — an established position is the standard scholarly account, not a primary-source pass. Where a figure has its own reason, it states it on its own page.

A design that treated quotation as the only marker of quality would make those figures look worthless, so the interface shows how widely a figure was read alongside how much of them is quoted.

Video transcripts count as primary sources

For living figures who publish hours of themselves speaking, an automatic transcript of a clear recording is strong provenance, not weak. It is the person’s own words, and a reader can check any quotation against the recording in one click — which is a better standard of verification than a page reference in a book they do not own.

Two constraints, both narrow and both stated wherever they apply:

1. Automatic captioning fails on proper nouns and technical theological terms — precisely the words that carry doctrinal weight. Those corrections are bracketed, and the research files say where.
2. Captions do not label speakers. On a multi-host programme, attributing a co-host’s view to the named figure would be worse than recording no position at all: confidently wrong, and dressed as sourced. Where the speaker is not unmistakable, the claim is attributed to the programme or dropped.

An earlier version of this project treated auto-captions as too unreliable to quote at all. That was the wrong call — it discarded the best available evidence about what a living preacher actually teaches, in order to avoid an error affecting a handful of words.

Graded states, and which of them are derived

Positions are not checkmarks. The vocabulary, borrowed from the PhilPapers survey, is: affirms, leans toward, permits as adiaphora, rejects, no formal position, internally contested, anachronistic, and no stance fits.

A state marked with a degree sign (°) is derived, not authored. It was read off the stance, the sourcing tier and the correction note rather than stated by the corpus. Derivation is a reading of the data, and presenting a reading as an authored claim is exactly the kind of quiet overreach this project argues against — so it is marked wherever it appears. As the corpus is revised, explicit states replace derived ones and the mark disappears.

Blank cells are content

Some positions carry no stance, and they mean at least four different things:

• No position located — searched, not found. An honest gap.
• Held open on purpose — the figure engaged the question and concluded it should not divide anyone. That is a position, not an absence.
• The question is not framed in these terms — anachronistic or foreign to them, rather than unanswered.
• No stance fits — every available option would misdescribe them, so none was assigned.

None of these ever renders as an empty cell or a dash. A blank that reads “we did not do the work” when it means “he deliberately declined” is a misrepresentation.

Corrections

The note field carries caveats and corrections, and several exist specifically to fix documented misreadings — that a figure is a polytheist, that he denies the Trinity. These render visually distinct from the summary, because a correction that reads as body text stops doing its job.

Scripture

Scripture references are the passages a figure is known to lean on for that position, not exhaustive proof-text lists. The text shown is public domain: the Berean Standard Bible (dedicated to the public domain in 2023), the World English Bible, and the American Standard Version (1901). No modern copyrighted translation is used.

What is known to be weak

• The corpus is Reformation- and Reformed-weighted. Every machine-readable confessional dataset that exists is Reformed-first, and this corpus inherits that. Orthodox and Pentecostal coverage is the thinnest and will be the most criticised.
• Every cell is an interpretive act. Unlike a survey, the figures here did not self-report — most of them are dead. This is the entire risk surface, and it is why per-cell quotation matters more than volume.
• The corpus is small and one person maintains it. The coverage page shows exactly where it is thin.

Disclosure: how this corpus was assembled

Research and drafting were AI-assisted, using a documented workflow, and every entry was reviewed by the maintainer before publication. Sourcing tiers record how far each claim was verified: researched means a primary source was read, and established means it was not re-verified for this entry.

This is disclosed unprompted because the alternative — being found out — is the difference between a transparent method and a machine putting words in Aquinas’s mouth. If you find an error, it is an error, and it should be reported and fixed.

Corrections and disputes

If a position misrepresents a figure or a tradition, say so. Corrections that identify the claim, the figure and a primary source are actionable and welcome.

Citing this project

Every page carries a “cite this” affordance naming the current edition. Editions are frozen and remain resolvable — see how to cite.