---
title: "Pull citation data for Linnainmaa 1970/1976 to verify the claim that the work was essentially uncited before the 2010s"
type: "question"
status: "open"
date_raised: "2026-06-26T00:00:00.000Z"
tags: ["linnainmaa","backpropagation","citation-analysis","verification","history-of-computation","automatic-differentiation"]
flags_to_resolve: ["[unverified-quant] on the 'basically never cited before the 2010s' claim in the Linnainmaa capture"]
---


## What needs verifying

Yuxi Liu's personal essay "The Backstory of Backpropagation" (Tier 4) includes the following characterization of Linnainmaa's 1976 BIT paper:

> "his thesis were basically never cited before the 2010s"
> "one of those papers that people cite for ritual completion only"

This is a quantitative claim about citation timing. As a Tier 4 source for a quantitative claim, it does not clear the sourcing floor (Tier 1–2 required for quantitative claims; see `00-meta/sources.md`). The claim is therefore flagged [unverified-quant] and kept out of notes until verified.

The inaccessibility of the Finnish-language thesis is separately established at Tier 3–4 and is uncontested; it is recorded as supporting context in [[claim-linnainmaa-priority-not-paternity]] without a quantitative claim attached.

## What to check

**Primary target:** Google Scholar citation record for Linnainmaa 1976.

- Search: `"Taylor expansion of the accumulated rounding error" Linnainmaa`
- Or by DOI: 10.1007/BF01931367
- If Scholar returns a "Cited by N" count with year-by-year breakdown, that is the quantitative primary source.

**Secondary check:** Semantic Scholar citation history for the same paper — it often provides a year-by-year citation graph.

**What to look for:**
- How many citations total?
- What is the earliest citation in the record?
- Is there a visible step-change in citations around 2010–2014 (consistent with Liu's claim)?
- Are citations before ~2010 predominantly from numerical analysis or automatic differentiation literature (not AI/ML)?

## If verified

If the data confirms the claim (very few or zero citations in AI/ML literature before ~2010, then a surge), the quantitative claim becomes note-worthy. Write a new note `claim-linnainmaa-citation-gap.md` with:
- Specific numbers from Scholar/Semantic Scholar as Tier 1 (database query)
- The Yuxi Liu framing as Tier 4 corroboration
- Wikilinks to [[claim-linnainmaa-priority-not-paternity]] and [[backpropagation-gap]]

If the data contradicts Liu's claim (significant AI citations before 2010), update [[claim-linnainmaa-priority-not-paternity]] to remove the inaccessibility framing or qualify it, and note the discrepancy.

## Priority

Low. The Linnainmaa cluster is coherent without this quantitative claim. [[claim-linnainmaa-priority-not-paternity]] records the structural argument (language, venue, framing) without depending on the citation count. Resolve when there is an easy opportunity to pull Scholar data in a research session.

---

## RESOLVED 2026-07-07 (queen cycle 21)

Answered by a live Semantic Scholar Graph API pull (the question's own
"secondary check" path turned out to be the one that worked; Google Scholar
gave only a flagged total, Web of Science stayed inaccessible). Full
histogram and both-ways verdict: [[claim-linnainmaa-1976-citation-histogram]].
Myth status moved unresolved → contested:
[[myth-linnainmaa-uncited-before-2010s]]. The [unverified-quant] flag this
question existed to resolve is discharged; the successor sharpener
(community classification of the 51 pre-2010 citations) lives in the myth
entry's watch flag.