---
id: "20261121-0215-does-griewanks-2012-text"
title: "Capture: Does Griewank's 2012 text itself use 'priority but no paternity' language, or is that framing introduced by Liu?"
type: "capture"
status: "promoted"
promoted_to: ["30-notes/claim-griewank-2012-full-text-contains-no-priority-but-no-paternity-phrase.md","30-notes/claim-liu-priority-paternity-framework-is-symmetric-across-linnainmaa-werbos-rumelhart.md"]
not_promoted: ["Claim 2 as originally framed ('the phrase is Liu's own stated judgment about Linnainmaa, not a quotation or paraphrase of Griewank') — not promoted as its own note. Retrieve-before-write found this exact conclusion already load-bearing in the vault three ways over: claim-griewank-2014-full-text-contains-no-priority-but-no-paternity-phrase's body, entity-yuxi-liu.md's 2026-11-20 line, and claim-linnainmaa-priority-not-paternity's 2026-07-06 revisit section. A fourth note restating it would duplicate rather than extend. The capture's own genuinely new material — Liu's symmetric three-way framework across Linnainmaa/Werbos/Rumelhart, flagged in its own 'Further leads' as uncaptured — was promoted instead as claim-liu-priority-paternity-framework-is-symmetric-across-linnainmaa-werbos-rumelhart.md, which carries the re-confirmed quote and supersedes the need for a separate attribution note.","Griewank's Humboldt-Universität zu Berlin affiliation / contact email (Further leads, bullet 2) — a minor biographical/institutional detail, not load-bearing enough for its own claim-note. Appended instead as a dated line on entity-andreas-griewank.md.","Liu's citation-count claim ('basically never cited before the 2010s') and the symmetric framework's downstream implications for Werbos/Rumelhart entity pages (Further leads, bullet 3) — already tracked by claim-linnainmaa-post-2010-citers-mostly-generic-ad-not-priority-historiography.md and the open question-adjacent territory; not re-verified this session, so left exactly where the capture found it."]
origin: "batch"
writer_model: "claude-sonnet-5"
date_created: "2026-11-21T00:00:00.000Z"
provenance: "batch run 2026-11-21, harvested from topic 'Does Griewank's 2012 text itself use priority but no paternity language, or is that framing introduced by Liu?' [high], itself harvested from 10-inbox/raw/2026-07-01-griewank-2012-linnainmaa-priority-autodiff-reception-before-2010s.md, whose Claim 3 explicitly flagged this as unresolved: 'It is not established from available sources whether Griewank 2012 uses this formulation himself or whether Liu coined it independently.' Resolved by direct full-text re-read of Griewank 2012 via extract_pdf and direct re-read of Liu's essay via archive_page, both in this session."
derived_from: []
tags: ["griewank","linnainmaa","liu","automatic-differentiation","backpropagation","reverse-mode","history-of-ml","priority","primary-source-verification","negative-finding","eponymy"]
seek_code_commit: "d40ee39"
---


> [!info] Core question resolved
> **The framing is Liu's, not Griewank's.** A direct, full-text read of Griewank's 2012 essay "Who Invented the Reverse Mode of Differentiation?" (all 12 pages) finds that the words "priority" and "paternity" occur nowhere in it, individually or combined — confirmed both by manual read and by machine quote-matching against the extracted primary text. The phrase "priority but no paternity" is Yuxi Liu's own analytical judgment, stated in his own voice ("I think we can safely say...") in his 2023 essay "The Backstory of Backpropagation," immediately before — and as a separate claim from — his citation of Griewank 2012 for unrelated biographical content about Linnainmaa's thesis.

## Claim: Griewank's 2012 text contains neither the word "priority" nor the word "paternity," nor the phrase "priority but no paternity," anywhere in its 12 pages

**Claim type**: historical/textual-verification (a claim about what a specific primary document does and does not say)
**Sourcing floor**: Tier 1-2 required — met by direct full read plus machine-verified quote-matching against the primary document itself.
**Source tier achieved**: Tier 1

[[entity-andreas-griewank|Andreas Griewank]]'s essay "Who Invented the Reverse Mode of Differentiation?" (*Documenta Mathematica*, Extra Volume ISMP, 2012, pp. 389–400) was re-fetched directly via `extract_pdf` from its publisher host (ems.press, TLS verified, matching the sha256 already on file from an earlier vault session) and read in full — prologue, "Relevance to Optimization," "Round-off Analysis à la Linnainmaa," "Jacobian accumulation," "Temporal complexity," "The memory issue," "Gradients and adjoint dynamics," "Summary and outlook," and the reference list. The words "priority" and "paternity" do not appear anywhere in this text. Candidate quote-matching against the extracted text with `quote_check` for "priority but no paternity," "paternity" alone, and "priority" alone all returned `grounded: false`. Griewank's own attribution language in this essay is instead: Linnainmaa is credited as the person from whose "point of view" the essay chooses to tell the reverse-mode story (for stated scope reasons — "we do not wish to specifically address electronic circuits or chemical processes"), and the essay opens by *declining* Trefethen's sole-credit framing of Griewank himself, listing at least eight earlier or parallel claimants (Ostrowski, Hachtel et al., Linnainmaa, Bennett, Werbos, Baur–Strassen, Speelpenning, Griewank's own 1987 rediscovery) rather than adjudicating a single "priority" verdict.

This corroborates, from the 2012 primary side, what a parallel full read of Griewank's 2014 paper already established for that later text ([[claim-griewank-2014-full-text-contains-no-priority-but-no-paternity-phrase]]): Griewank's own vocabulary across both of his historical essays is hedged and descriptive ("apparently the first author," "facilitating the rebirth," narrative "point of view"), never the compact "priority/paternity" formulation.

| Field | Value |
|---|---|
| source_url | https://ems.press/content/book-chapter-files/27379 |
| source_sha | 3b3b8fc861df955d8da60d652af820021c684867c0c23d6dad9551733aefa134 |
| source_title | "Who Invented the Reverse Mode of Differentiation?" |
| source_author | Andreas Griewank |
| source_date | 2012 |
| source_venue | Documenta Mathematica, Extra Volume ISMP (2012), pp. 389–400 (European Mathematical Society) |
| source_tier | 1 |
| source_quote | "(negative finding — no quote exists; `quote_check` against the full extracted text returned grounded:false for "priority but no paternity", "paternity", and "priority" individually)" |
| method | Full manual read of all 12 pages of `extract_pdf` output, cross-checked with machine quote-matching via `quote_check` |

## Claim: The phrase "priority but no paternity" is Yuxi Liu's own stated judgment about Linnainmaa, not a quotation or paraphrase of Griewank

**Claim type**: historical/textual-verification (attribution of authorship for a specific phrase)
**Sourcing floor**: Tier 3-4 acceptable for this kind of uncontested textual-authorship claim; the claim here rests on a direct, receipt-backed read of the document whose authorship is in question.
**Source tier achieved**: Tier 4 (Liu's essay is a personal/self-published blog per the standing vault rating), directly quoted and archived

[[entity-yuxi-liu|Yuxi Liu]]'s essay "The Backstory of Backpropagation" (yuxi.ml, published 2023-12-26, modified 2024-11-22) was re-fetched directly via `archive_page` and read in full. In the section headed "Seppo Linnainmaa," Liu writes, in his own first-person analytical voice: "A bibliography search shows that his thesis were basically never cited before the 2010s. It is one of those papers that people cite for ritual completion only. **I think we can safely say that his thesis has priority but no paternity.** (Griewank 2012) describes in detail what is in the thesis, according to which he developed it to analyze the cumulative effect of round-off error in long chains of computer computations." The phrase is explicitly framed as Liu's own conclusion ("I think we can safely say") drawn from a citation-count observation, and the Griewank citation that immediately follows it is attached to a separate, distinct claim — what the thesis technically contains — not to the priority/paternity phrase itself.

The same essay shows the phrase is part of a symmetric three-way analytical framework Liu applies himself, not a term he found in any source: he writes that Paul Werbos "has priority and paternity," and that David Rumelhart "has no priority, but paternity" — a parallel structure built around Linnainmaa, Werbos, and Rumelhart that only makes sense as Liu's own conceptual apparatus for sorting the three men's claims, not as something borrowed from Griewank's essay (which does not discuss Rumelhart at all — see [[claim-reverse-mode-multiple-independent-discovery]]'s 2026-09-11 audit note on this point).

| Field | Value |
|---|---|
| source_url | https://yuxi.ml/essays/posts/backstory-of-backpropagation/ |
| source_sha | 622b14e6d32c830950943db62ccb01afe8b3bc2883127024149327ac5d2a08d5 |
| source_title | "The Backstory of Backpropagation" |
| source_author | Yuxi Liu |
| source_date | 2023-12-26 (modified 2024-11-22) |
| source_venue | Yuxi on the Wired (yuxi.ml), personal essay site |
| source_tier | 4 |
| source_quote | "I think we can safely say that his thesis has priority but no paternity." |
| source_quote_2 | "He [Werbos] has priority and paternity." |
| source_quote_3 | "He [Rumelhart] has no priority, but paternity." |
| method | Full re-read of `archive_page` output; `quote_check` against the archived text confirmed the first quote `grounded: true` |

## Vault connections

This capture closes the exact gap flagged at the end of the 2026-07-01 capture that originated this line of inquiry (preserved there as "the key open question this would settle: does Griewank himself use 'priority but no paternity' language, or is that Liu's framing?"), and sharpens [[claim-linnainmaa-priority-not-paternity]] — whose own title and frontmatter quote the phrase from Liu's essay (Tier 4) rather than from Griewank — by establishing affirmatively, on a now-complete read of Griewank 2012, that the phrase is absent from the primary text rather than merely unconfirmed. It is a companion negative finding to [[claim-griewank-2014-full-text-contains-no-priority-but-no-paternity-phrase]] (same absence, confirmed independently for Griewank's 2014 paper) and to [[claim-griewank-2012-does-not-mention-amari]] (same exhaustive full-text-scan method applied to a different name). It also sits alongside [[claim-griewank-2012-prologue-complicates-trefethens-sole-credit-attribution]], which already established that Griewank's own prologue resists a single-inventor framing — consistent with an author who would not reach for a tidy two-word epigram like "priority but no paternity" in the first place.

> [!note] Seek's commentary:
> The tell was always going to be in the citation placement, and it's exactly where you'd hope: Liu drops the epigram, *then* cites Griewank — for a different fact entirely. That's not how you cite the source of a phrase; that's how you cite the source of a detail, right after stating your own verdict. Once you see Liu running the same priority/paternity split on Werbos and Rumelhart too, the case closes itself — nobody borrows a rhetorical device and then reuses it three times as their own organizing principle. It's his scale, and he's just telling you where he got the raw material to weigh each man on it.
> — Seek

## Further leads

- Liu's symmetric three-way framework (Linnainmaa: priority, no paternity; Werbos: priority and paternity; Rumelhart: no priority, paternity) is itself a distinct, uncaptured claim about how Liu organizes the whole backpropagation-origins story — worth its own atomic note separate from the Linnainmaa-specific one already in the vault.
- The EMS Press copy of Griewank 2012 lists his affiliation at time of writing as Institut für Mathematik, Humboldt-Universität zu Berlin (contact: griewank@mathematik.hu-berlin.de) — a small biographical/institutional detail not previously recorded in [[entity-andreas-griewank]].
- Liu's essay states he personally "checked all his [Linnainmaa's] English papers during the 1970s" and that only the 1976 *BIT* paper "has been cited non-negligibly before 2000" — a citation-history claim already flagged elsewhere in the vault as resting on Liu's own unverified bibliography search ([[claim-linnainmaa-post-2010-citers-mostly-generic-ad-not-priority-historiography]] and the open [[question-verify-linnainmaa-citation-count]] territory) rather than a documented count; not re-verified in this session.

## Entity candidates

- Seppo Linnainmaa — person — the foundational figure whose priority claim this whole phrase is about; already has an entity note ([[entity-seppo-linnainmaa]]); flagged first since he is the figure the "priority" half of the phrase is measured against, not an afterthought.
- Andreas Griewank — person — author of the Tier 1 primary text this capture's negative finding rests on; already has an entity note ([[entity-andreas-griewank]]).
- Yuxi Liu — person — confirmed in this capture, by direct quotation of his own words, as the originator of the "priority but no paternity" phrase and of the broader priority/paternity framework applied to Linnainmaa, Werbos, and Rumelhart alike; still no entity note found in the vault for this name despite being load-bearing in multiple existing claim-notes' frontmatter — worth creating.
- Paul Werbos — person — subject of Liu's parallel "priority and paternity" verdict in the same essay; already has an entity note per other vault notes citing him, not re-checked in this session.
- David Rumelhart — person — subject of Liu's parallel "no priority, but paternity" verdict in the same essay; likely already has an entity note per other vault notes citing him, not re-checked in this session.
